A Meta-Analysis for Exploring the Diverse Causes and Effects of Stress in Teachers
Author(s): Cameron Montgomery and André A. Rupp
Source: Canadian Journal of Education / Revue canadienne de l'éducation , 2005, Vol. 28,
No. 3 (2005), pp. 458-486
Published by: Canadian Society for the Study of Education
Stable URL: https://www.jstor.org/stable/4126479
REFERENCES
Linked references are available on JSTOR for this article:
https://www.jstor.org/stable/4126479?seq=1&cid=pdf-
reference#references_tab_contents
You may need to log in to JSTOR to access the linked references.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms
Canadian Society for the Study of Education is collaborating with JSTOR to digitize, preserve
and extend access to Canadian Journal of Education / Revue canadienne de l'éducation
This content downloaded from
147.251.22.8 on Thu, 21 Nov 2024 07:58:51 UTC
All use subject to https://about.jstor.org/terms
A Meta-analysis for Exploring the Diverse Causes
and Effects of Stress in Teachers
Cameron Montgomery & Andre A. Rupp
This study provides a correlational meta-analysis of 65 independently written or
published studies on teacher stress between 1998 and 2003. We measured the
relationships between teacher stress and numerous other constructs including coping,
burnout, emotional responses, personality mediators, personal support,
environmental structure, and background characteristics. A theoretical-empirical
model of construct relationships investigated across studies was developed and n =
2,527 correlational effect sizes were used to estimate the empirical relationships
between the operationalized theoretical constructs. Results showed that the strongest
association of teacher stressors exists with negatively oriented emotional responses
confirming the central role of teachers' coping mechanisms, personality mediators,
and burnout potential according to our model of the stress cycle.
Key words: stress, coping, teacher bum-out, teacher emotional response
Cette recherche fournit une meta-analyse correlationnelle de 65 etudes, ridigees ou
publiees entre 1998 et 2003, sur le stress chez les enseignants. Les auteurs ont mesure
les relations entre le stress des enseignants et de nombreux autres construits, dont
I'adaptation au stress, I'dpuisement professionnel, les reactions emotionnelles, les
personality mediators, le soutien personnel, la structure du milieu et les antecedents.
Un module theorico-empirique des relations entre les construits a ete blabord et des
valeurs d'effets correlationnels de n = 2527 ont ete utilisees pour determiner les
relations empiriques entre les construits theoriques operationalis&s. Les resultats
demontrent que, pour les facteurs de stress chez les enseignants, I'association la plus
importante se trouve du c6te des reactions emotionnelles negatives, ce qui confirme le
r61e cl' des mecanismes d'adaptation au stress des enseignants, des personality
mediators et du potentiel d'dpuisement professionnel selon le module du cycle du
stress qui a etd utilise.
Mots cl&s: stress, adaptation au stress, meta-analyse, enseignants.
CANADIAN JOURNAL OF EDUCATION 28, 3 (2005): 458-486
This content downloaded from
147.251.22.8 on Thu, 21 Nov 2024 07:58:51 UTC
All use subject to https://about.jstor.org/terms
A META-ANALYSIS FOR EXPLORING THE DIVERSE CAUSES AND EFFECTS OF STRESS 459
Over the past ten years educational research has established that hig
teacher stress is associated with psychological distress, which may b
mediated through different coping mechanisms and personality trait
(Chan, 1998). Specifically, poor active coping abilities or an over-reliance
of passive coping strategies may lead to negative emotional response
and, consequently, teacher burnout. A plethora of research on the
different sources of stress and their eventual consequences in teachers
and student teachers exists, and researchers have used varying methods
in explaining the intricate relationships between sources of psychologica
stress and other intricately related constructs such as coping
mechanisms, personality traits, emotional responses, environmental
effects, and burnout.
This study investigates and summarizes the correlational evidence of
the relationships between psychological stress and related constructs
through a meta-analytic lens to synthesize and, more importantly,
understand some general trends in the current research literature on
teacher stress and to conceptually represent a model of teacher stress
based on this literature. Specifically, this study focuses on the recent
research literature on teacher and student teacher stressI between 1998
and 2003. In doing so, this study updates recent international research
data on teacher stress and constitutes a modest step toward allowing
other researchers to understand relationships between stress and other
constructs within the stress cycle. This, in turn, is an attempt to aid them
in responding to Guglielmi and Tatrow's (1998) paramount call for a
shift toward more theory-based investigations that test causal models of
teacher stress and health-related outcomes with more sophisticated
research designs and measurement strategies, which can be facilitated by
having an empirical sense of which construct relationships merit closer
investigation.
TOWARD A THEORETICAL-EMPIRICAL MODEL OF STRESS
Numerous constructs have been identified as causes and effects of stress
in different populations over the years. Moreover, numerous mediating
factors may influence the relationships between variables in th
nomological networks in which researchers place constructs. In
This content downloaded from
147.251.22.8 on Thu, 21 Nov 2024 07:58:51 UTC
All use subject to https://about.jstor.org/terms
460 CAMERON MONTGOMERY & ANDRF R. RUPP
preparation for developing a meta-analytic framework, w
discussed, and synthesized existing theoretical stress models
empirical relationships and represent these studies into a fin
of the stress cycle that we will present later. Because our m
focuses on a variety of relationships among numerous const
than on a few isolated relationships between few constru
larger framework, we wanted to have a conceptual mode
stress to augment and refine, based on the empirical re
studied in the research we surveyed. Before presenting
however, we have described our understanding of stress and
related constructs, based on well-established existing re
theory.
Definition of Stress
Our understanding of stress originated in the empirical research of
Derogatis (1987), who conducted his most recent research using the
Derogatis Stress Profile (DSP), a psychological questionnaire to measure
individuals' stress dispositions. Derogatis based this questionnaire on
Lazarus's (1966) social interaction theory of stress which consequently
led us towards Lazurus's more recent research and theories of stress and
coping. Lazarus and Folkman (1984) define stress as a particular
interaction between the person and the environment, appraised or
evaluated by the person as being taxing or exceeding his or her persona
resources, and, as a consequence, disrupting his or her daily routine
According to Derogatis, stress may be defined as a state of psychologica
pressure influenced by three main sources or domains: personality
mediators (constructs of time pressure, driven behaviour, attitud
posture, relaxation potential, and role definition); environmental factor
(constructs of vocational satisfaction, domestic satisfaction, and health
posture); and emotional responses (constructs of hostility, anxiety, and
depression). Derogatis accordingly explains that these three source
must be studied interactively to develop a comprehensive account o
psychological stress.
This content downloaded from
147.251.22.8 on Thu, 21 Nov 2024 07:58:51 UTC
All use subject to https://about.jstor.org/terms
A META-ANALYSIS FOR EXPLORING THE DIVERSE CAUSES AND EFFECTS OF STRESS 461
Sources of Teacher Stress
Research has shown that teachers are exposed to a number of sources of
stress. Kyriacou (2001) reports that the main sources of teacher stres
stem from teaching students who lack motivation; maintaining disciplin
in the classroom; confronting general time pressures and workload
demands; being exposed to a large amount of change; being evaluated
by others; having difficult or challenging relationships with colleagues,
administration, or management; and being exposed to generally poo
working conditions. The author warns that the sources of stress
experienced by a particular teacher will, of course, be unique to him or
her and will depend on the precise complex interaction between his o
her personality, values, skills, and circumstances. Moreover, coping
mechanisms, personality traits, or the environment can interactivel
influence the degree to which stressful situations are being perceived
and influence the teacher's emotional and cognitive well-being.
Stress and Coping
Our approach to investigating coping mechanisms was grounded in
Lazarus and Folkman's transactional model of stress and coping which
focuses on how problematic events trigger stressful episodes. Lazaru
and Folkman (1984) and, more recently, Admiraal, Korthagen, and
Wubbels (2000) believe that daily events predict changes in
psychosomatic health better than major life events. Lazarus and Folkman
claim that when confronted by a given event, an individual engages in a
process of primary appraisal whereby the event may be seen as stressful
or benign, depending on the individual and the situation. Next, the
individual will engage in a process of secondary appraisal. In this
process, an individual will engage in the cognitive evaluation of his or
her personal and environmental resources to deal with the stressful
event. In other words, primary appraisal refers to the appraisal of the
stressful character of the situation, whereas secondary appraisal refers to
the evaluation of an individual's capacity to confront the situation. Both
types of appraisals are cognitive processes that depend, to a large extent,
on the appraising individual. Finally, their theoretical model predicts
This content downloaded from
147.251.22.8 on Thu, 21 Nov 2024 07:58:51 UTC
All use subject to https://about.jstor.org/terms
462 CAMERON MONTGOMERY & ANDRt R. RUPP
that individuals will use cognitive and
adaptation to deal with a given stressful eve
Chan (1998) points out that evidence of str
health and psychological distress, are not
stressors but are also determined by a host
of which are generally collected under t
mechanisms or strategies (Lazarus & Folkm
1985). Several classifications of coping
processes have been proposed in the lite
behaviour may be directed at managing or
causing the distress or at regulating the
problem. The former is referred to as prob
latter as emotion-focused coping (Admir
2000). Problem-focused coping behaviour co
problem-solving strategies such as defini
alternative solutions, weighing alternativ
benefits, selecting one of them, and taki
coping behaviour consists of positive rea
well as defensive strategies such as avo
distancing. Individuals will use emotionwhen
they believe that they cannot do anything to modify
environmental conditions. By contrast, they will utilize problem-focused
coping behaviour when they see conditions as changeable. Similarly,
Kyriacou (2001) defines direct action techniques as things that a teacher
can do to eliminate the source of stress, and mental or physical palliative
techniques as things that they can do to lessen the feeling of stress.
Stress and Personality
The literature also provides support that an individual's personality
characteristics influence the degree to which he or she seeks social
support when confronted by a stressful event (Houston & Vavak, 1991;
Kobasa, Maddi, Ouccelli, & Zola, 1985; Watson & Clark, 1984).
Guglielmi and Tatrow (1998) claim that personality traits, especially
Type-A personality, and demographic characteristics, such as gender,
age, and ethnicity, further mediate the ability to establish and maintain
This content downloaded from
147.251.22.8 on Thu, 21 Nov 2024 07:58:51 UTC
All use subject to https://about.jstor.org/terms
A META-ANALYSIS FOR EXPLORING THE DIVERSE CAUSES AND EFFECTS OF STRESS 463
supportive social networks and facilitative cognitive appraisal of
stressor. Stress is not viewed as being due exclusively to situation
personal characteristics but rather to the interaction between them. T
findings reveal that the seeking of social support and the engagemen
successful coping strategies can render an objective situation
demanding, threatening, or harmful to an individual.
Stress and Burnout
Individuals may experience burnout as a result of stress itself, a sudden
breakdown of their mediating coping mechanisms, or an ineffectiveness
of their mediating coping mechanisms over a long period of time
(Guglielmi & Tatrow, 1998; Vandenberghe & Huberman, 1999). Burnout
has traditionally been viewed as having three components: emotional
exhaustion, depersonalization, and lack of personal accomplishment.
Burnout is most frequently measured using Maslach's Burnout Inventory
(Maslach & Jackson, 1981). A feeling of burnout is not a direct effect of
repeated exposures to stressful situations, however. Burnout is mediated
through various active and passive coping mechanisms, as discussed
above, and is a result of the accumulation of positively and negatively
oriented emotional responses that have arisen through coping
mechanisms.
A THEORETICAL-EMPIRICAL MODEL OF TEACHER STRESS
Based on the above definition of stress as well as discussio
lectures of recognized scientific theories on stress, an over
literature on teacher stress, including both qualitative and qu
methodologiesi, and an extensive review of theoretical tea
models, including their related constructs and definitions
overview, we developed a model of key construct relation
model includes as its central component a representation
intra-individual process of reacting to stressful events (see Fig
The most important premise of the model is that a teache
to outside events, is the core agent throughout the entire mod
stressors are present at the beginning of what we consid
This content downloaded from
147.251.22.8 on Thu, 21 Nov 2024 07:58:51 UTC
All use subject to https://about.jstor.org/terms
CAMERON IMONTGOMERY ~ ANDR~ R. RUPP
??~I
i5' '
r '?' ? C. - ?111 ?1
?'?:??
r!?:: ?
?r ?r~i? ?r?
-?
n-?)~~? ? ??. I,
V ??
c ?~i?:?~~; ~ r?
,,r ?!:::
.?. :i?? ,-L? c. ?.'; 1.
1? :?;.t ?:rb? d 1.??''' ~''
:?.?-? ..i
d'' - i~!i'? r .'
?~% a r?-:.f::i:;:
''
`/?: : r- r-'??;:?'''
: j.i;
f 1r..'~'' c c.r.. t: 'Z\: ?
jf ? r4-? r;? ' ?
??: i. " :?' !?~ '?
j~
t4: \I.'i-- (-.1
::1; ?3?~ BFr ?i?.? ?,
?; ?c.;-,
?i:? 1. t'
:: I.
.. :: I ?
?:? 1Ua?,r
fi P ?;r P ,??-. I?. .. r;54 i''?'I ??, f S .... : ?
C , ? r iil%
'' .??.. r-? :.I
j?'I ? :!i"i ; i !5..4
I?' ?'~ ~
CIL ,?? ?:?... c ii 1. ?.~?, ,, ?; : ? IS.. r. .-p~
r?.?
L '?;'r .~?:? .r'
F~?'~T?i. F''-: ''
gf .. -.I ?.. :r
?: . i?? ? r???';: ?1?.. i?i
Lt::f' :?;.~L%!~,.?.
" il?-L I ~
.. L?: r (-? :ii` r.$. .,
,?, ??- . I,~%?';%':I i ..~5 .? ~?:??` ~ ;;.r ~.?,/Y ??
.i ???~
.? 't, t ~: '? ''?.cr?~;!?:.~
it. ;r
:n n n:t' s',-; ?I '.~cr
I `
?~?? ' :.
?, i??.lr
L52 ~~I
I
`~"~ I'
i It
,, c?fi?:
' f.-1`??`
: ? d .. f
j
.5.P
;C
rr:
ts . .
a -~~?:~?:?-~~,??~'''.,..
I'
';? rl?
1:
9% '?
.r.
jII1?
..
ii- ?
--
i ?,
..
?. i
This content downloaded from
147.251.22.8 on Thu, 21 Nov 2024 07:58:51 UTC
All use subject to https://about.jstor.org/terms
A META-ANALYSIS FOR EXPLORING THE DIVERSE CAUSES AND EFFECTS OF STRESS 465
stress cycle. As for the other labels, the labelling process of the categories
captures the general patterns of construct relationships and maps onto
similarly signified latent variables in structural equation models (Chan,
1998). Specifically, we developed this model after reviewing and
discussing major characteristics in both qualitative and quantitative
research. Next, we qualitatively categorized individual variables and
variable relationships in the empirical literature (i.e., in quantitative
articles) on teacher stress surveyed for this paper, the reason why the
stress cycle in the diagram is semantically tied to an environment. Apart
from the sources of teacher stress (external stressors), the remaining part
of the model appears to display universally applicable relationships that
may be used in other domains for research.
According to this model, intra-individual processes consist primarily
of the experience and evaluation / appraisal of external stressful events
that have their sources in different aspects of a teacher's professional /
vocational life such as students, administration, colleagues, general work
demands, and characteristics of the school environment (external
stressors). In addition, external stressful events in a teacher's personal /
domestic life such as problems in the relationship with a life partner or
financial debt may also influence his or her overall emotional, cognitive,
and behavioural state. Once an individual has appraised these events, he
or she engages in active coping or passive coping strategies, perhaps
both. The former can take the form of cognitive strategies (e.g., changing
perspective, exerting self-control, rationally distancing oneself),
behavioural strategies (e.g., setting limits for work, seeking advice from
others, engaging in relaxation exercises), and emotional strategies (e.g.,
being calm, thinking positively) and also manifest themselves in
individuals' physical responses or health posture. In contrast, passive
coping mechanisms such as resignation, drinking, wishful thinking, and
avoidance are characterized by a lack of direct engagement with the
stressful event toward its resolution. As a result of the application, or
lack thereof, of these coping mechanisms, or sometimes directly as a
consequence of the stressful events, an individual experiences a host of
emotional responses, which are either positively oriented such as hope,
enjoyment, or passion, or are negatively oriented such as anxiety,
frustration, depression, or even suicidal ideation. In addition, an
This content downloaded from
147.251.22.8 on Thu, 21 Nov 2024 07:58:51 UTC
All use subject to https://about.jstor.org/terms
466 CAMERON MONTGOMERY & ANDRF R. RUPP
individual may experience strong feelings of satisfaction or
dissatisfaction with his or her job and life situation more generally,
which may influence this individual's commitment level to his or her
work. Finally, then, individuals may experience different levels of
emotional exhaustion, depersonalization, or personal accomplishment as
facets of burnout. It is, of course, debatable whether the degree of
satisfaction is causally prior or posterior to experiences of burnout,
which will likely depend on an individual and a situation; we chose to
place satisfaction prior to burnout because we viewed it as
predominantly connected / related to emotional response.
The entire intra-individual situational process described so far is
further mediated by relatively stable personality traits, so-called
personality mediators (e.g., driven behaviour, attitude posture,
relaxation potential, type-A behavior) that influence the strength of the
relationships depicted in the core of the model. In fact, certain
individuals may be more predisposed to external stressors because of
their personality. Personality, therefore, accounts for what we consider
to be the inner stressors in our model. Moreover, the relationships are
mediated by the degree to which individuals feel supported, both in
their vocational environment (e.g., by their bosses and colleagues) and
their domestic environment (e.g., spouse and friends); these sets of
factors are depicted as outward circles around the core path diagram.
Further outward in the model, the relatively stable structural
characteristics of the environment are represented such as the teachers'
grade level, their average class size, or the type of school in which they
are teaching. Similarly, background variables such as sex, educational
qualifications, and years of experience are individuals' stable
characteristics that may have some influence on the intra-individual
process of dealing with stressful events.
This model does not account, per se, for the fluidity of individuals'
responses to stressful events to distinguish between state and trait
components of stress, a measure that would require the collection of
longitudinal data on cohorts. Moreover, because the focus of this paper
is the estimation of average correlational effect sizes for each of the
displayed paths and hypothesized relationships, we are necessarily
restricted to statements that aggregate effects over teacher groups and
This content downloaded from
147.251.22.8 on Thu, 21 Nov 2024 07:58:51 UTC
All use subject to https://about.jstor.org/terms
A META-ANALYSIS FOR EXPLORING THE DIVERSE CAUSES AND EFFECTS OF STRESS 467
situations. At the same time, our study is an important empirical step
toward investigating in a relatively comprehensive manner the systemic
nature of construct relationships between stress and other widely
studied constructs. This model may, therefore, be applicable to other
contexts or domains (e.g., managerial, medical, social) when trying to
understand stress and its relationship with other constructs.
The organization of the variable sets in Figure 1 represents, partially,
our primary focus on the core relationships between stress, the
engagement of coping mechanisms, the experience of emotional
responses and satisfaction, and the experience of burnout. In addition, it
also describes the relative importance we ascribe to the different classes
of constructs and concepts that are represented in the model. These
factors lead us to the following hypotheses.
Research Hypotheses
We hypothesize that the relationship between stress and coping
mechanisms as well as between coping mechanisms, emotional
responses, and burnout is stronger than the influences that background
variables have on the coping process. Similarly, considering the outer
rings, we hypothesize that environmental structure variables will display
weaker relationships with the intra-individual variables than the support
structure variables or the personality mediator variables. At the same
time, we believe that personality mediators and support variables
display strong influences with the intra-individual relationships
depicted in the core.
METHOD
The following section of this article describes our methodolog
to code variables and to compute average effect size measu
meta-analysis.
Collection of Articles
We located relevant recent articles on stress though a compute
international databases housed in North America such as P
This content downloaded from
147.251.22.8 on Thu, 21 Nov 2024 07:58:51 UTC
All use subject to https://about.jstor.org/terms
468 CAMERON MONTGOMERY & ANDRF R. RUPP
Eric, Sociofile, Canadian Periodical Index, Index-Cpi.Q, Inf
Dissertations, Current Contents, Ontario Scholars Portal, Francis
(International Humanities and Social Sciences), and Merlot as well as
through searches of the internet via Google, Metacrawler, and Yahoo.
We also located articles in data banks housed outside North America
such as the FisBildung in Germany and Repere in Canada and Franc
We employed various permutations of keywords to track art
Every search included the keywords "teachers" and "stress," while
searches included keywords such as "student teacher," "pre-ser
"burnout," "coping," "anxiety," and "depression." We used
reference lists in all primary articles to perform a search of addit
references within the sampling frame of interest until we found no
articles. We selected studies that investigated teacher stress in some
although, in several cases, the title did not necessarily indicate it (e.
included studies on teacher burnout, teacher coping, and teacher h
if they were also investigating the relationship between these cons
and teacher stress). Finally, of the 211 studies initially identifi
kept only articles that contained quantitative research, thus exclud
conceptual overview and synthesis articles as well as qualitative res
because we could not use meta-analytic procedures for these studie
addition, we eliminated articles that employed complex multiv
measures such as cluster structures because it was impossi
unambiguously define or assign effect sizes as measuring a si
variable or a single relationship between pairs of variables.
We included research from refereed journal publications (n =
dissertations (n = 13), and refereed conference proceedings (n = 1) f
total of n = 65 independently written or published studies. Eleven
studies were published in a language other than English (2 we
French, 2 were in Japanese, and 7 were in German); the Japanese s
were translated by trained bilingual graduate students, wherea
French and German studies were not translated because both authors are
multilingual.3
Sample Characteristics
Table 1 shows some important characteristics of the samples included
across the studies.
This content downloaded from
147.251.22.8 on Thu, 21 Nov 2024 07:58:51 UTC
All use subject to https://about.jstor.org/terms
A META-ANALYSIS FOR EXPLORING THE DIVERSE CAUSES AND EFFECTS OF STRESS 469
Table 1
Sample Characteristics across Studies
Study Country Teacher Sample Age Age Female Male Exp. Exp.
Size Mean (SD) (%) (%) Mean (SD)
Abel &
Sewell
(1999) USA Second. 98 --- --- 86 14 15.35 8.56
Adams
(2001) USA Second 364 --- --- 64 36 --- ---
Alkhrisha USA and
(2002) Jordan Other 228 --- --- --- --- --- ---
Anopchand
(2000) USA Second 143 --- --- --- --- --- ---
Antoniou et
al.(2000) Greece Other 110 --- --- 39 61 --- ---
Avramidis et
a1.(2000) UK Primary 135 --- --- --- --- --- ---
Bechen
(2000) USA Both 300 --- --- --- ---
BoehmKasper
&
Weishaupt
(2002) Germany Second 1091 --- --- 55 45 --- ---
Cains &
Brown
(1998) UK Primary 76 --- --- --- --- 1.00 ---
Center &
Callaway
(1999) USA Other 151 --- 10 85 15 --- 8.60
Center
&Steventon
(2001) USA Other 149 --- --- --- --- --- ---
Center
(2001) USA Other 35 37 14 94 6 10.93 7.80
Chan (1998) China Second 415 31 8 55 43 7.78 7.46
Hong
Chan (2002) Kong Second 83 24 3 72 28 --- ---
Hong
Chan (2003) Kong Other 83 24 3 72 28 --- ---
Cheuk et al.
(2002) China Second 180 --- --- --- --- --- ---
Chi On Hong
(2001) Kong Both 543 --- --- --- --- ---
Conley &
Woosley
(2000) USA Both 371 --- --- 73 27 --- ---
Cousin
(2000) USA Both 166 --- --- 90 10 5.00 ---
Davis &
Wilson
(2000) USA Primary 704 --- --- --- --- --- ---
DeJesus &
Conboy
(2001) Portugal Both 25 42 --- 88 12 18.00 ---
Friedel &
Dalbert
(2003) Germany Primary 108 43 8 --- 1 22.52 8.00
Griffith et al.
(1999) England Both 780 --- --- 76 24 --- ---
This content downloaded from
147.251.22.8 on Thu, 21 Nov 2024 07:58:51 UTC
All use subject to https://about.jstor.org/terms
470 CAMERON MONTGOMERY & ANDR R. RUPP
Study Country Teacher Sample Age Age Female Male Exp. Exp.
Size Mean (SD) (%) (%) Mean (SD)
Griva &
Joekes
(2003) UK Second 166 38 --- 52 48 12.40 ---
Hawe et al. New
(2000) Zealand Primary 353 --- --- 85 15 --- ---
Hemmings
& Hockley
(2001) Australia Primary 54 --- --- --- --- --- ---
Ishofsky
(1998) USA Both 43 --- --- --- --- --- ---
Jacobsson et
al. (2001) Sweden Both 928 43 --- --- --- --- 16.40
Kittel &
Leynen
(2003) Belgium Second 128 44 9 --- --- 19.90 9.80
Kumarakula
singam
(2002 USA Primary 227 --- --- 100 --- 13.31 10.62
Lafave
(2000) USA Primary 53 --- --- --- --- --- ---
Mearns &
Cain (2003) USA Both 86 40 11 76 24 13.88 11.20
Montgomery
& Leonard
(2003) Canada Both 106 --- --- 72 28 --- ---
Montgomery
(2001) Canada Both 458 --- --- 80 20 --- ---
MurrayHarvey
et al.
(2000) Australia Both --- --- --- --- --- --- ---
Neuenschwa
der (2003) Germany Second 146 45 --- 54 46 ---
Pack (2000) Canada Other 40 --- --- --- --- 19.00 ---
Pascual et al.
(2003) Spain Second 198 44 28 53 47 --- ---
Peklaj &
Puklek
(2001) Slovenia Primary --- --- --- --- ---
Petrie (2001) USA Primary 382 --- --- --- --- --- ---
Pisanti et al.
(2003) Italy Second 169 48 8 64 42 21.00 ---
Pithers & Australia
Soden &
(1998) Scotland Other 332 --- --- --- 0.00 0.00
Pomaki et al.
(2003) Greece Second 215 44 6 60 40 17.39 6.39
Rasku &
Kinnunen
(2003) Finland Second 373 46 10 69 31 18.40 10.70
Root (2001) USA Primary 74 --- --- --- --- ---
Ryo & Koji
(2003) Japan Both 710 --- --- --- --- --- ---
Sakata et al.
(1999) Japan Second 212 37 --- 45 55 13.41 ---
Sann (2003) Germany Second 297 48 9 46 54 20.30 9.10
Schmitz &
Schwarzer
(1999) Germany Other 140 --- --- 58 42 --- ---
Schmitz
(2001) Germany Other 132 --- --- 58 42 --- ---
This content downloaded from
147.251.22.8 on Thu, 21 Nov 2024 07:58:51 UTC
All use subject to https://about.jstor.org/terms
A META-ANALYSIS FOR EXPLORING THE DIVERSE CAUSES AND EFFECTS OF STRESS 471
Study Country Teacher Sample Age Age Female Male Exp. Exp.
Size Mean (SD) (%) (%) Mean (SD)
Schmitz
(2004) Germany Other 103 --- --- 73 27
Schonfeld
(2000) USA Both 184
Schonfeld
(2001) USA Both 184 --- --- 100
Sumsion &
Thomas
(2000) Australia Both 13
Thomas et
al. (2003) Australia Primary 102 43 8 100 --- 17.18 8.26
Van der
Doef &
Maes (2002) Holland Second 454 44 8 36 64 16.30 8.20
Van Dick et
al. (1999) Germany Both 424 47 8 56 44 19.80 8.80
Van Zyl &
Pietersen South
(1999) Africa Second 66
Vandoan
(1998) USA Primary 246 --- --- 36 64 5.00
Verhoeven
et al (2003) Holland Second 1878 --- --- 42 58 16.80
Whitehead New
et al. (2000) Zealand Primary 386 --- --- 87 13
Winzelberg
& Luskin
(1999) USA Primary 21 25 --- 71 29 0.00 0.00
Yagil (1998) Israel Primary 69 31 8 69 ... ... ...
Yoon (2002) USA Primary 113 --- -- --- --- 12.00 ---
Hong
Yu (2002) Kong Second 128 --- --- 49 50 64.83 35.17
The mean sample size across studie
deviation of 303.4, which arose from a fe
and we computed this standard deviat
reported the standard deviation. Using
percentage of females in the samples was
deviation of 18.1 per cent, and the me
samples was 38.3 per cent with a stand
Moreover, the mean average age of tea
years with a standard deviation of 5.8
reported these data, and the mean a
experience was 15.6 years with a standa
on 26 studies. We noted that 17 studie
elementary school teachers or student
(30.8%) used only secondary school teache
out of the 65 studies (26.2%) used both ty
11 studies used teachers at other levels or did not provide this
This content downloaded from
147.251.22.8 on Thu, 21 Nov 2024 07:58:51 UTC
All use subject to https://about.jstor.org/terms
472 CAMERON MONTGOMERY & ANDRF R. RUPP
information. Therefore, the participants in the studies were largely
female, a reflection of the teaching profession internationally, generally
older (i.e., between 30 and 50 years of age), with substantial experience
(i.e., around 10 - 20 years), and generally elementary or secondary school
teachers.
Coding of Studies
We coded studies according to characteristics of the sampling frame and
stages, the experimental design structure, the population(s) sampled,
and the statistical methodologies utilized. We entered the information
from the studies in three different ways. First, because of the variability
in the measures employed across all studies, we created a data file that
contained each bivariate relationship in each study as a separate entry.
For each relationship, we entered the two variables as they were labelled
in the study along with the statistical technique used to compare them,
the effect size measure that was reported, and the degrees of freedom,
sample size, and p-value for the associated test. For studies that did not
report effect sizes or provided insufficient information, we computed
data manually.
We then reduced the original data file to a smaller number of studies
(n = 65) because for some studies only complex statistical models were
employed whose parameter estimates or effect sizes could not be
transformed to a correlation metric. The resulting data file contained k =
2,527 entries where, on the original metric; k = 2,061 effect sizes were
Pearson product-moment correlation coefficients; k = 62 were Spearman
correlation coefficients; k = 6 were Pearson point-biserial correlation
coefficients; k = 134 were independent-samples t-test statistics; and k =
264 were F-test statistics largely representing independent-samples t-test
statistics because they had 1 numerator degree of freedom. Because
constructs involved in the bivariate relationships varied, we developed
more global coding categories to summarize the effects of different
relationships that represent a single path in the theoretical-empirical
model depicted in Figure 1. Indeed, the statistical analyses presented in
the following sections are aimed at estimating the average zero-order
correlations between the variable pairs.
This content downloaded from
147.251.22.8 on Thu, 21 Nov 2024 07:58:51 UTC
All use subject to https://about.jstor.org/terms
A META-ANALYSIS FOR EXPLORING THE DIVERSE CAUSES AND EFFECTS OF STRESS 473
Creation of Variable Categories
After having entered all explanatory and response variables from all
studies within our sampling frame, we recoded the variables according
to the theoretically derived categories depicted in our model (see Figure
1), based on a detailed analysis of the study design descriptions. Hence,
the highest-order categories corresponded to different sources of stress,
different types of active and passive coping, different types of emotional
responses, different facets of burnout, and different types of support, as
well as personality mediators, environmental structure variables, and
background variables.
We were able to classify most variables into one of these categories.
For example, if one study used the phrase 'educational qualifications'
and another used the phrase 'educational level' while, at the same time,
both studies measured the number of years that student teachers had in
previous teaching-related experiences, we renamed both variables as
'years of experience' and assigned it to the category 'background
characteristics.' Similarly, we relabelled categories such as 'financial
stressors' and 'money stressors' as 'degree of financial stress' and
assigned them to the category 'stress' nested within 'domestic.' We
categorized constructs such as depression, suicidal ideation, and anxiety
as negative emotional responses rather than as internal stressors. In
cases where we disagreed over classifications, we reached a final
classification by going back to earlier studies that contained a theoretical
model with these variables in them or by consulting other empirical
research articles outside of our sampling frame.4
Statistical Analysis
We analyzed the correlational data using the methods outlined in Fern
and Monroe (1996), Lipsey and Wilson (2001), Hedges & Olkin (1985),
and Hunter and Schmidt (1990). The structure of effect-size statistics
varies, of course, by the statistical model for which the effect size is
reported as a measure. Nevertheless, most effect sizes can be readily
transformed from one metric to another (e.g., Fern & Monroe, 1996).
Because most of our effect sizes were zero-order Pearson productThis
content downloaded from
147.251.22.8 on Thu, 21 Nov 2024 07:58:51 UTC
All use subject to https://about.jstor.org/terms
474 CAMERON MONTGOMERY & ANDRF R. RUPP
moment correlation coefficients based on two variables measured on
interval scales, we chose to transform the remaining effect sizes ont
correlation metric, consequently eliminating a few effect sizes such
unstandardized and standardized regression coefficients in mu
linear regression models or structural equation models because t
represent partial effects and are computed with other variables inc
in the model.
For each correlation coefficient so obtained, we recorded the number
of samples on which they were based. Because of the variability in the
measures we investigated, these samples are not, technically, statistically
independent because multiple correlation coefficients from the same
study that index different bivariate relationships between different
construct pairs were included in the computation of the same average.
However, whether the violation of the statistical independence would
lead to serious biases in the computation of the average effect size
measures and their standard errors is debatable because it depends on
the amount of dependence. To ensure that sample sizes were noninflated,
we recorded only the total sample size for all independent
samples across the studies.
For each study, we recorded the reliabilities of the psychometric
measures because our original intention was to correct the involved
coefficients for unreliability. Unfortunately, empirical reliability
estimates were not reported for more than half of the scales and
subscales used across the studies; we did not complete this correction.
However, the median reliability for 380 reported subscale reliabilities
was 0.81 (5th percentile = 0.60 and 95th percentile = 0.92) and all reported
reliabilities were internal consistencies measured by Coefficient a. Based
on these data, we estimated the effect that corrections for unreliability
resulting in disattenuated coefficients would have on individual
correlational effect sizes to get a sense of the effect of unreliability
correction if all reliability values had been available. We considered the
25th, 50th, and 75th percentile of the observed reliability distribution as
three representative conditions and assumed that both variables on
which an observed correlation coefficient was based have the same
reliability. For these cases, Table 2 shows how the difference b
This content downloaded from
147.251.22.8 on Thu, 21 Nov 2024 07:58:51 UTC
All use subject to https://about.jstor.org/terms
A META-ANALYSIS FOR EXPLORING THE DIVERSE CAUSES AND EFFECTS OF STRESS 475
uncorrected and corrected effect sizes varies, depending on the size o
the observed correlation and the reliabilities of the variables involved.
For small effect sizes, this difference is generally no larger than about
0.05. For medium correlational effect sizes and a typical scale reliability
of 0.81, this difference is no larger than about 0.10 and is, of course, eve
less if the variable reliabilities are higher, and more if the variabl
reliabilities are lower. For large correlational effect sizes and a typic
scale reliability of 0.81, this difference is generally no larger than abou
0.15. If one compares these idealized cases with results of corrections fo
unreliability and range restriction in other meta-analyses from journals
such as Review of Educational Research, one can relatively safely assume
that the difference between the uncorrected and the corrected correlation
coefficients would generally be around 0.05 or 0.10 with corrected values
being, of course, higher than the uncorrected ones. Because most of the
observed correlations in this study were low, as will be discussed below,
the difference between uncorrected and corrected correlations in this
meta-analysis would most likely not be striking; we therefore did n
any comparison.
Table 2
Theoretical Values of Correlational Effect Sizes Corrected for Unreliability
Effect Size r a = .72 a = .81 a = .86
rc (rc - r) rc (rc - r) rc (rc - r)
0.00 0.00 0.00 0.00 0.00 0.00 0.00
Small 0.05 0.07 0.02 0.06 0.01 0.06 0.01
0.10 0.14 0.04 0.12 0.02 0.12 0.02
0.15 0.21 0.06 0.19 0.04 0.17 0.02
0.20 0.28 0.08 0.25 0.05 0.23 0.03
0.25 0.35 0.10 0.31 0.06 0.29 0.04
Medium
0.30 0.42 0.12 0.37 0.07 0.35 0.05
0.35 0.49 0.14 0.43 0.08 0.41 0.06
0.40 0.56 0.16 0.49 0.09 0.47 0.07
Large 0.45 0.63 0.18 0.56 0.11 0.52 0.07
0.50 0.69 0.19 0.62 0.12 0.58 0.08
0.55 0.76 0.21 0.68 0.13 0.64 0.09
This content downloaded from
147.251.22.8 on Thu, 21 Nov 2024 07:58:51 UTC
All use subject to https://about.jstor.org/terms
476 CAMERON MONTGOMERY & ANDRF R. RUPP
Effect Size r a = .72 a = .81 a =.86
rc (rc - r) rc (rc - r) rc (rc - r)
0.60 0.83 0.23 0.74 0.14 0.70 0.10
0.65 0.90 0.25 0.80 0.15 0.76 0.11
0.70 0.97 0.27 0.86 0.16 0.81 0.11
0.75 1.00 0.25 0.93 0.18 0.87 0.12
0.80 1.00 0.20 0.99 0.19 0.93 0.13
0.85 1.00 0.15 1.00 0.15 0.99 0.14
0.90 1.00 0.10 1.00 0.10 1.00 0.10
0.95 1.00 0.05 1.00 0.05 1.00 0.05
1.00 1.00 0.00 1.00 0.00 1.00 0.00
Notes: r = Observed correlational effect size, rc = Corr
unreliability, a = Coefficient a used as a measure of
r r
The correction is based on
a/ a2 a
that both variables have the same reliability.
Next, we constructed confidence intervals for the correlation
coefficients. The correlation coefficients were first transformed to the
standard normal metric using Fisher's z transformation, which possesse
superior statistical properties, and the standard errors, weights, and 95
per cent confidence interval limits were subsequently computed in that
metric. Finally, we transformed the confidence interval limits back to th
original correlation metric (see, e.g., Lipsey & Wilson, 2001, p. 72, for th
formulas). The following section now presents a summary of the major
associations found between indicators of the constructs included in our
model.
RESULTS
We computed average correlation coefficients for all pairs of variab
subsumed under the higher-order construct classification catego
depicted in Figure 1; Table 3 shows the resulting average correlat
effect sizes.
This content downloaded from
147.251.22.8 on Thu, 21 Nov 2024 07:58:51 UTC
All use subject to https://about.jstor.org/terms
A META-ANALYSIS FOR EXPLORING THE DIVERSE CAUSES AND
Table 3
Correlation Matrix with Average Correlational Effect Sizes
Stress Active Passive Emotional Burnout Personality Support
Coping Coping Responses Mediators
Stress 1 N= 6280 N = 1761 N = 6254 N = 4453 N = 7941 N = 57
K= 179 K= 25 K= 128 K= 124 K= 99 K=
Active Coping .2025 1 N = 1761 N = 3274 N = 4217 N = 5562 N = 4
(.1942, K=41 K=92 K= 123 K=77 K=5
.2107)
Passive Coping .0751 .1544 1 N = 1626 N = 1626 N = 1540 N = 1
(.0475, (.1327, K = 17 K = 27 K=7 K=8
.1025) .2759)
Emotional .2512 .0500 .0950 1 N = 2447 N = 5402 N = 3910
Responses (.2413, (-.0628, (.0611, K = 69 K = 79 K =
.2611) .1615) .1287)
Burnout .2673 .2685 .0858 .3977 1 N = 4821 N = 3700 N
(.2565, (.2575, (.0583, (.3846,.4106) K=49 K=
.2780) .2793) .1131)
Personality .2535 .1624 .1031 .3033 .2746 1 N = 4837 N
Mediators (.2433, (.1489, (.0531, (.2080, .3930) (.2600, K
.2635) .1760) .1527) .2890)
Support .2604 .1503 .0608 .2671 .2357 .2302 1 N
(.2475, (.1355, (.0105, (.2495, .2844) (.2188, (.2095
.2731) .1650) .1108) .2525)
Environment .1914 .2067 n/a .2772 .1781 .2225
Structure (.1806, (.1867, (.2488, .3052) (.1609, (.2047, .2
.2021) .2265) .1951) .3683
Background .1120 .0909 .1292 .0631 .0728 .1199
(.1035, (.0820, (.1085, (.0506, .0756) (.0642, (.1037, .1361) (.043
.1205) .0999) .1497) .0813) .0794
Notes. Values in the lower off-diagonal are avera
the total sample size of all independent samples
were observed for this construct pair.
This content downloaded from
147.251.22.8 on Thu, 21 Nov 2024 07:58:51 UTC
All use subject to https://about.jstor.org/terms
478 CAMERON MONTGOMERY & ANDRF R. RUPP
The upper off-diagonal of Table 3 shows the total samp
independent samples (N) and the number of effect size
based the averages (K), whereas the lower off-diago
average effect size statistics on a correlation metric with t
95 per cent confidence intervals. The total number o
between variables from the nine categories (i.e., stress
passive coping, emotional responses, burnout, personal
support, environmental structure, background characteri
across the 64 studies was k = 2,023. The average absolute c
was 0.19 with a standard deviation of 0.08. All average cor
significant at an individual a = .05 level because the cor
per cent confidence intervals do not contain 0, with the ex
average correlation between active coping and emotio
(rF = 0.05; 95% CI = [-0.06,0.16]) and the average correl
background characteristics and environmental structu
0.06,0.33]).
According to Cohen (1988), correlations of less than or
can be considered small whereas correlations of more than 0.40 can be
considered large although no definitive classification and labelling exi
therefore, the linguistic descriptors in the following are somewh
imprecise. Results from Table 3 indicate that average correlation
between external stressors such as student misbehaviour and workload
demands and others constructs are, generally, weak to moderate w
typically only about 5 per cent and 15 per cent of shared varian
between variable pairs. External stressors are most highly correla
with burnout (F = 0.27; 95% CI = [0.26,0.28]), support variab
(F = 0.26; 95% CI = [0.25,0.27]), personality mediator variables (F = 0.25
95% CI = [0.24,0.26]), and emotional response variables (F = 0.25; 95% C
= [0.24,0.26]) and less strongly correlated with active coping (F = 0.
95% CI = [0.19,0.21]), environmental structure variables (F = 0.19; 95%
= [0.18,0.20]), and background variables (F = 0.11; 95% CI = [0.10,0.1
the lowest average correlation was found for passive coping (F = 0
95% CI = [0.05,0.10]). In other words, external stressors may b
moderately influencing burnout directly, with support from family
other colleagues playing a mediating role with similarly moderat
effects.
This content downloaded from
147.251.22.8 on Thu, 21 Nov 2024 07:58:51 UTC
All use subject to https://about.jstor.org/terms
A META-ANALYSIS FOR EXPLORING THE DIVERSE CAUSES AND EFFECTS OF STRESS 479
We have attributed an important central role to emotional responses
because they are moderately to highly correlated with several other
variables. The majority of emotional response variables that have been
empirically investigated are negatively orientated (e.g., distress, anxiety,
depression), implying, for example, that the average correlations
between external stressors and these responses primarily reflect an
exposure to stressful events and lead to negative experiences for
teachers, be they mediated through coping mechanisms or not, and that
these responses, in turn, may lead to different types and magnitudes of
burnout. With emotional response variables, average correlations were
moderate to high in the frame of reference of the average correlations
observed here. Specifically, we observed a high average correlation
between emotional response variables and burnout (r = 0.40; 95% CI =
[0.38,0.41]), showing that the degree in which teachers emotionally
respond to stressful events and how satisfied they are as a consequence,
either mediated through coping mechanisms or not, has a strong
influence on the degree of burnout they experience. We observed
moderate average correlations between emotional responses and
personality mediator variables (F = 0.30; 95% CI = [0.21,0.39]),
suggesting that the way teachers emotionally respond to a variety of
stressful situations is closely tied to the relatively stable personality traits
that mediate their responses according to the stress cycle. Emotional
response variables were also moderately correlated with environmental
structure variables ( F = 0.28; 95% CI = [0.25,0.31]), and support variables
(r= 0.27; 95% CI = [0.25,0.28]), indicating that these factors also
influence quite strongly how people emotionally respond to stressful
events, which, in turn, influences the degree of burnout they experience.
Put simply, this shows that the subjectively perceived quality of the
environment and the support structures available to individual teachers,
both at home and at work, are important for dealing with stressful
situations. In this context, future induction studies considering the
relationship between mentoring and stress may reveal relevant findings
because the former may act as a buffer against burnout. Finally, as seen
with active and passive coping, the average correlation between
emotional responses and background variables (r = 0.06; 95% CI =
[0.05,0.08]) was low.
This content downloaded from
147.251.22.8 on Thu, 21 Nov 2024 07:58:51 UTC
All use subject to https://about.jstor.org/terms
480 CAMERON MONTGOMERY & ANDRE R. RUPP
In terms of active coping, average correl
to moderate, and we found that activ
correlated with burnout (r = 0.27; 95%
weaker correlations between active coping
variables (r- = 0.21; 95% CI = [0.19,0.23]), pe
(r = 0.16; 95% CI = [0.15,0.18]), support
[0.14,0.17]), and passive coping (r- = 0.15;
extremely weak correlations exist be
background variables (r = 0.09; 95% C
emotional response variables (F = 0.05; 95%
one actively copes through exercise (e.g., o
behavioral strategies, cognitive planning, o
emotional responses in the face of vario
determine if one will indeed feel emotiona
or not accomplished. Yet, at the same ti
engages in active coping strategies does
background characteristics and does no
emotional responses to stressful events
themselves, moderately correlated to st
burnout as a result. In terms of passive cop
generally low. Interestingly, the construct
with passive coping was background char
[0.11,0.15]), showing that, for example, th
people with different levels of teaching ex
though that, in itself, does not strongly i
they internally mediate the stressors.
In terms of personality mediators, avera
moderate as shown by, for example, the
personality mediator variables and bu
[0.26,0.29]), support variables (F = 0.23
environmental structure variables (rF =
Moreover, the average correlation bet
variables and background variables is
[0.10,0.14]), showing that personality trait
with stable background characteristics such
This content downloaded from
147.251.22.8 on Thu, 21 Nov 2024 07:58:51 UTC
All use subject to https://about.jstor.org/terms
A META-ANALYSIS FOR EXPLORING THE DIVERSE CAUSES AND EFFECTS OF STRESS 481
In terms of support variables, average correlations varied by the
construct they were related to. For example, although a moderate
average correlation exists between support variables and environmental
structure variables (rF = 0.33; 95% CI = [0.30,0.37]), only a low average
correlation exists between support and background characteristics
(F = 0.06; [0.04,0.08]), showing that the perception of available support
for individual teachers does not vary systematically with background
characteristics such as sex or educational level. Moreover, the average
correlation between background characteristics and environmental
structure variables is low (F = 0.14; 95% CI = [-0.06,0.33]), but is, at the
same time, the highest correlation between background characteristics
and other constructs.
Findings for Research Hypotheses
Based on the results presented so far, one can observe that the empirical
support for our hypotheses was, generally, only weak. Specifically, only
moderate associations exist between external stressors and burnout,
personality mediators, negatively or positively oriented emotional
responses including satisfaction and dissatisfaction, and support
variables as well as between emotional responses and burnout. Although
background characteristics did not seem to strongly affect other
components within the stress cycle, the structure of the teaching
environment (environmental structure) and personality mediators were
generally relatively highly correlated with other components within the
stress cycle, given the average correlations observed in this study. Within
the core of the model, it became clear that only some support occurred
for the influence of active coping strategies as effective mediators for
influencing the emotional responses and that no support occurred for the
effectiveness of passive coping strategies generally. As expected, stress
levels correlated positively with other variables in the stress cycle, but
not always as strongly as expected.
To summarize some of the central average correlations discussed
above, Figure 2 shows the theoretical-empirical model described earlier
with selected effect sizes indicated. Note that the numbers in Figure 2 are
This content downloaded from
147.251.22.8 on Thu, 21 Nov 2024 07:58:51 UTC
All use subject to https://about.jstor.org/terms
CAMERON MONTGOMERY & ANDRF R. RUPP
4 14JIL
rr
t--;..
1 I' ' ' " " i
i ii
j?? & L
??F~
ii
7i
Itl?1. I?
5
~ ?~?
L 6
I,:II
This content downloaded from
147.251.22.8 on Thu, 21 Nov 2024 07:58:51 UTC
All use subject to https://about.jstor.org/terms
A META-ANALYSIS FOR EXPLORING THE DIVERSE CAUSES AND EFFECTS OF STRESS 483
selected average zero-order correlations as discussed above and not
coefficients in a structural equation model, which is beyond the scope o
this paper.
One may gather from the results that follow that emotional
responses, personality mediators, support variables, and burnout play,
not surprisingly, a central role in the manner in which teachers respond
to external stressors. In particular, this may point to the importance to
examine emotions in conjunction with external stressors in future studies
to better understand the effects of external stressors on negative
emotions and, possibly, resulting in burnout. Indeed our results confirm
Goleman's (1995) theory of emotional intelligence because the results
presented here suggest that emotions have a more central role for
understanding the intricate relationship between stress, burnout,
personality, and support variables.
There are, of course, limitations to this study. If the correlation
coefficients were disattenuated and corrected for range restriction, their
absolute values would change making the corrected observed effect sizes
somewhat higher. In addition, it might be argued that there could be
some disagreement concerning the labelling of some of the categories if
the same studies were handed to other researchers, but the consensus
approach with multiple researchers used here coupled with a continual
integrative look at the literature to derive at final classifications makes us
confident that the classifications would be rather stable.
Implications
The results of the present study may be useful for researchers in other
fields trying to better understand the relationship between externa
stressors, negative emotions, personality mediators, support variable
and, most importantly, burnout. Indeed, understanding and uncovering
negative emotions related to external stressors is the first step towards
better performance, a higher degree of professional satisfaction, and
consequently, a higher level of teacher retention.
This content downloaded from
147.251.22.8 on Thu, 21 Nov 2024 07:58:51 UTC
All use subject to https://about.jstor.org/terms
484 CAMERON MONTGOMERY & ANDRF R. RUPP
CONCLUSION
Our study highlights the importance of considering the re
between stress and negative emotions, the latter leading po
burnout which is costly for both individuals and society. The
our study and our model should give educators and oth
professionals in the managerial or medical field a clearer vision
coping, and burnout.
NOTES
1 We use the term teacher stress and not student teacher stress
because most of the studies found concerned the former.
2 Only research with quantitative methodology/data was used to
refine our model after we initially conceived/constructed our model,
based on our overview of qualitative and quantitative studies.
3 The studies used in this meta-analysis are available from the
authors.
4 In the future, it would be important to investigate the inter-rater
reliability of the classifications because the current ones are based only
on the expertise of the two principal researchers.
REFERENCES
Admiraal, W. F., Korthagen, F. A J., & Wubbels, T. (2000). Effects of st
teachers' coping behaviour. British Journal of Educational Psychology, 70
Chan, D. W. (1998). Stress, coping strategies, and psychological distress a
secondary school teachers in Hong Kong. American Educational R
Journal, 35(1), 145-163.
Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd
Hillsdale, NJ: Erlbaum.
This content downloaded from
147.251.22.8 on Thu, 21 Nov 2024 07:58:51 UTC
All use subject to https://about.jstor.org/terms
A META-ANALYSIS FOR EXPLORING THE DIVERSE CAUSES AND EFFECTS OF STRESS 485
Derogatis, L. R. (1987). The Derogatis stress profile (DSP): Quantification of
psychological stress. Adv. Psychosom. Med., 17, 30-54.
Fern, E. F., & Monroe, K. B. (1996). Effect-size estimates: Issues and problems.
Journal of Consumer Research, 23, 89-105.
Goleman, D. (1995). Emotional intelligence. New York: Bantam Books.
Guglielmi, R. S., & Tatrow, K. (1998). Occupational stress, burnout, and health i
teachers: A methodological and theoretical analysis. Review of Educational
Research, 68(1), 61-99.
Hedges, L. V., & Olkin, I. (1985). Statistical methods for meta-analysis. San Diego
CA: Academic Press.
Houston, B. K., & Vavak, C. R. (1991). Hostility: Developmental factors,
psychological correlates, and health behaviors. Health Psychology, 10, 9-17.
Hunter, J. E., & Schmidt, F. L. (1990). Methods of meta-analysis. New York: Rus
Sage.
Kobasa, S. C., Maddi, J., Ouccelli, M., & Zola, M. C. (1985). Effectiveness of
hardiness, exercise and social support as resources against illness. Journal of
Psychosomatic Research, 29, 525-533.
Kyriacou, C. (2001). Teacher stress: Directions for future research. Educational
Review, 53(1), 27-35.
Lazarus, R. S. (1966). Psychological stress and the coping process. New York:
McGraw-Hill Book Company.
Lazarus, R. S., & Folkman, S. (1984). Stress, appraisal, and coping. New York:
Springer.
Lipsey, M. W., & Wilson, D. B. (2001). Practical meta-analysis. Thousand Oaks,
CA: Sage.
Maslach, C., & Jackson, S. E. (1981). The measurement of experienced burnout.
Journal of Occupational Behavior, 2, 99-113.
This content downloaded from
147.251.22.8 on Thu, 21 Nov 2024 07:58:51 UTC
All use subject to https://about.jstor.org/terms
486 CAMERON MONTGOMERY & ANDRE R. RUPP
Scheier, M. F., & Carver, C. S. (1985). Optimi
and implications of generalized outcome ex
219-247.
Vandenberghe, R., & Huberman, A. M. (Eds.). (1999). Understanding and
preventing teacher burnout: A sourcebook of international research and practice.
Cambridge, UK: Cambridge University Press.
Watson, D., & Clark, L A. (1984). Negative affectivity: The disposition to
experience-aversive emotional states. Psychological Bulletin, 96, 465-490.
This content downloaded from
147.251.22.8 on Thu, 21 Nov 2024 07:58:51 UTC
All use subject to https://about.jstor.org/terms