American Sociological Association is collaborating with JSTOR to digitize, preserve and extend access to
Journal of Health and Social Behavior.
http://www.jstor.org
Journal
of Healthand Social Behavior1990, Vol. 31 (September):260-276
Drawingon bothtensionreduction
and social learningtheories,we hypothesized
thatworkstressorslead to increaseddistress,whichin turnpromotes
problematic
alcoholuse amongvulnerableindividuals.Vulnerable
individuals
are hypothesized
to possess few personal and social resourcesfor respondingadaptivelyto
work-related
stressorsand distressand to holdpositiveexpectancies
foralcohol's
effects.
We testedour modelin a randomsampleof574 employed
adults,usinga
combination
ofpath analyticand hierarchicalmoderatedregressiontechniques.
Results revealed no supportfor a simple tensionreductionmodel of work
stress-induced
drinking
and onlylimitedsupport
for a social learningmodel.We
concludethata muchmorecircumscribed
viewoftheetiologicroleofworkstress
inproblematic
alcohol use is indicated.
The beliefthatworkstresspromotesheavy
or problematic
alcoholuse is widelyaccepted
(e.g., Gupta and Jenkins1984; Herold and
Conlon 1981; Milleret al. 1988; Trice and
Roman1978). Indeed,efforts
to reducework
stress or to provide alternativeformsof
copingwithstressare currently
advocatedas
measurestoprevent
alcoholabusein thework
force(e.g., Shain 1983; see Weiss 1986 for
opposingviewpoints).
Yet despitewidespread
acceptanceof the belief that work stress
promotesproblematic
alcoholuse, few studies provideconvincingempiricalsupportfor
thisnotion.Existingresearchfailsto provide
an adequate theoretically
groundedtest of
workstress-induced
drinking.Thus the purpose of thispaper is to proposeand testa
comprehensive
modelof workstress-induced
drinking.
Theoretical
Perspectives
on
Stress-Induced
Drinking
At leasttwo majortheoretical
perspectives
articulatea causal link betweenstressand
alcohol-related
pathology.Tensionreduction
theoryproposesthatalcoholreducestension
and, more important
forour purposes,that
people drinkalcohol forits tension-reducing
* This researchwas supportedby National properties
(CappellandGreeley1987). "TenInstitute
of AlcoholAbuse and AlcoholismGrant sion" refershere to variousnegativeemo#AA05702, awarded to the second author. tional states that plausiblycould serve as
Portionsof these resultswere presentedat the aversivesourcesof motivation,
suchas fear,
annual meetingof the AmericanPsychological anxiety,distress,or depression.Thus negaAssociation,held in New York City in August tiveemotionsplaya keyintervening
rolein a
1987. The authorswishto thankJeremy
Skinner, putativecausal sequence
linkingexposureto
WilliamGeorge,Dean McFarlin,GerardConnors,
conditionsand increaseddrinking.
Jim Neff, and JennyCrockerfor theirhelpful stressful
conditionsare hypothesized
commentson an earlierdraftof thismanuscript. Thatis, stressful
Correspondence
thisarticleshouldbe to increasenegativeemotions,whichin turn
concerning
addressedto the firstauthorat 348 Park Hall, increasealcoholuse.
The social learningperspective
PsychologyDepartment,
StateUniversity
on alcohol
of New
Yorkat Buffalo,Buffalo,NY 14260.
use and abuse complementsand extends
260
261
262
Work
Stressors
Hia
Dist
stress
Alcohol
Expectancies
Hib
Alcohol
Use/Abuse
263
previousresearchsuggests
study:social support(which may be con- (H2c). In contrast,
copingdoes not
ceived usefullyas copingassistance;Thoits thatactive,problem-focused
betweengroupswithandwithout
1986), copingresources,copingstyles,and discriminate
copingresponses.Accordingto theproposed clinically significantevidence of alcohol
effect pathology(Billingsand Moos 1983; Moos et
model, the natureof each moderator
dependsto a greatextenton the a priori al. 1981). Thus we do not anticipatethat
coping will moderatethe
specificationof the moderatorvariable as problem-focused
adaptiveor maladaptive.That is, adaptive relationshipof work distress to alcohol
to buffer outcomes.
resourcesandskillsarehypothesized
the negative emotion-alcoholrelationship, Our model suggeststhatindividualswho
whereas maladaptiveattributeswould be hold strongpositiveexpectanciesfor alcothis hol's effects
shouldbe morelikelyto drinkto
expected to exacerbateor strengthen
to drinkheavily,and
relationship.
cope withworkdistress,
Social support is widely regarded as ultimatelyto experiencealcohol problems.
adaptive(e.g., Broadheadet al. 1983; Leavy Thus we hypothesizethat positivealcohol
1983) and thuswould be expectedto buffer expectancieswill exacerbatethe relationship
relationship
(H2a). betweenworkdistressand alcoholoutcomes
theworkdistress-alcohol
positiverelationship
Coping resourcesare generalizedattitudes such thata significant
and skills thatare consideredadvantageous should be obtainedamong high-expectancy
theyincludeattitudes individualsbut not among low-expectancy
acrossmanysituations;
abouttheworld individuals(H2d).
aboutself(esteem),attitudes
skills(intelli- Finally,we proposethatindividualswho
(beliefin mastery),intellectual
skill(social compe- are deficientin adaptiveresourcesand skills
gence),and interpersonal
and who
tence)(Menaghan1983). Thus,highlevelsof (or highin maladaptiveattributes)
copingresourcesalso would be expectedto are also highin positiveexpectanciesshould
relationship be mostlikelyto experienceadversealcohol
bufferthe workdistress-alcohol
outcomesas a resultof elevationsin work
(H2b).
Copingstylesare typicalor habitualways distress.Thus, positiveexpectancies,adapof approaching
(i.e., copingwith)problems, tiveresources,and workdistresswill interact
to cope, alcoholuse, and
whereascopingresponsesare specificactions to predictdrinking
(overtor covert)made in specificsituations alcohol problems such that high-distress,
thatare intendedto reducea givenproblem low-resource,high-expectancy
individuals,
(Menaghan1983). In the presentstudywe will exhibitthe greatestinvolvementwith
contrast
twobroadformsof copingstylesand alcohol(H2e).
coping
responses-activeor problem-focused
versus avoidantor emotion-focused
coping
(Folkmanand Lazarus 1980). Priorresearch METHOD
whorelyon avoidant
suggeststhatindividuals
formsof emotion-focused
coping are more Sample
likely to experiencenegativealcohol outlife
Respondentsin the presentstudy were
comes in response to stress-inducing
events(Billingsand Moos 1983; Moos et al. drawnfroma randomsamplesurveyof 1,933
householdresidentsin Erie County,New
1981). Althoughneitherstressor-by-coping
interactionsYork.Designatedrespondents
19 yearsof age
nor negativeemotion-by-coping
were testeddirectlyin these studies,these and older were identifiedin a three-stage
data nonethelessimplythathigh-avoidance probabilitysample designedto yield equal
of two racial groups(blacks
copersare morelikelyto relyon alcoholas a representation
or managingthedistress and nonblacks)and threeeducationallevels
meansof alleviating
generatedby stressfullife events.Thus we (less thanhighschool, highschool, at least
anticipatethatcoping stylesand responses somecollege).The overallsamplecompletion
indicativeof avoidance of emotion will rate in the presentstudywas 78.3 percent;
were
exacerbatethe relationshipbetween work most (84.5%) of the noncompletions
distressand alcohol outcomessuch that a refusals.
Of the 1,933 respondents,916 were
significantpositive relationshipshould be
observedamonghigh-avoidance
copers,but employedat least 20 hours per week. Of
not necessarilyamonglow-avoidancecopers these, 175 respondentswho abstainedand
264
thearithmetic
meanof
167 who drankless thanonce a monthwere formedby computing
eliminated,yieldinga final sample of 574 theirrespectiveitems.
whohaddrunkat leastonce
covariates.Sex (1 =
adultrespondents
Sociodemographic
a monthoverthepastyear.To avoid biasing male,2 = female),age (in years),race (1 =
our test of the proposed hypotheses,we nonblack,2 = black), education(in years),
2
who drank and occupationalstatus(1 = white-collar,
limitedthesampleto respondents
drinkers, = blue-collar)wereused as covariatesin the
regularly.
Abstainers
and infrequent
thoughunlikely
to use alcoholto cope, almost present study. Race and education were
degreesof controlledbecause of the highlystratified
certainly
wouldexperience
varying
job stress.Thustheirinclusionwouldtendto samplingdesign used in the presentstudy;
sex, age, and occupationwere controlled
attenuate
thepredicted
effects.
in patterns
differences
Respondentsin the presentstudy were becauseofdocumented
ofthese
sociodemographicallyheterogeneous.Ap- ofalcoholuse andabuseas a function
characteristics
proximatelyhalf were white (51.6%); 47 sociodemographic
(e.g., Hilpercentwere marriedor living in stable ton 1987; Olkinuora1984).
Workstressors.We assessedtwostressors:
relationship;
51 percentwere female.They
rangedin age from19 to 69 (mean = 36.5, workpressureand lack of job control.Work
sd = 10.8). Nearly73 percentof thesample pressure(WP; 11 items) assessed the frehad completedhighschool;an additional22 quencywithwhichindividuals
perceivedhigh
percenthad completedfouror moreyearsof job-relateddemands resultingfrom heavy
Lack of job
college. White-collaroccupations(profes- workloadand responsibilities.
sional,technical,managerial)werefollowed control(LJC; six items) assessed the frequency with which individualsperceived
by 53 percentof thesample.
constraints
on theirabilityto function
autonomously and to influenceimportantjob
Bothscales werederivedempiriparameters.
Procedures
cally froma pool of 17 itemstakenfrom
Data were collected by a corps of 27 several previouslypublished measures of
in thesummerand fallof 1986. workstressors
interviewers
(viz., House et al. 1979; Insel
Interviewers
receivedfive days of intensive and Moos 1974; Pearlinand Schooler1978).
techniques, We used a four-point
trainingon generalinterviewing
responseformat
(almost
of the survey,and study- neverto almostalways).
administration
specific procedures.Interviewswere conwereintended
to
Althoughitemsoriginally
ductedin respondents'
homesaccordingto a measurethreestressor
dimensions
(workload,
and lack of job control),both
highly structuredinterviewschedule that responsibility,
and self-adminis-principalcomponentsand factor analyses
containedbothinterviewerteredportions.The completeinterviewre- revealedtwo robustfactorsin whichitems
90 minutesto adminis- constituting
theworkloadandtheresponsibilquiredapproximately
werepaid $25 fortheirtime. ity dimensionsloaded togetheron the first
ter.Respondents
factor.Because theobtainedtwo-factor
strucacrossboth
tureprovedto be largelyinvariant
sexes and races and because it yieldedthe
Measures
most reliable indices, we employed the
We employeda totalof 24 indicesin the empiricallyderivedscales in all subsequent
presentstudyto assess six broadcategoriesof analyses(see appendixforindividualitems).
variables: 1) sociodemographiccovariates Alcoholoutcomemeasures.We examined
(five indicators),2) work stressors(two threealcohol-related
outcomesin thepresent
outcomes(three study: 1) the self-reported
measures),3) alcohol-related
frequencywith
used alcohol to cope, 2)
measures),4) workdistress(one measure),5) whichrespondents
personaland social resources(12 measures), averagedaily consumption
over the past 12
and 6) alcohol expectancies(one measure). months,and 3) the numberof drinking
These measuresare describedbelow; esti- problems experiencedover the past 12
matesof internal
reliability
(coef- months.
consistency
ficientalpha), zero-ordercorrelations,and
Drinkingto cope (Polich and Orvis 1979)
statistics
are providedin Table 1. was assessed by six items.On a four-point
descriptive
all scaleswere scale, respondents
of
Exceptwherenotedotherwise,
reportedthe frequency
4a
4b
4c
5a
Sb
6a
6b
6c
6d
7a
7b
7c
(78)
22
(80)
10
25
16
32
-13 -03
(68)
57
10
(65)
25
(52)
-13
04
08
21
(82)
Stressors
1. Workpressure
(81)
2. Lack ofjob control
17
MediatingVariable
3. Workdistress
37
Moderating
Variables
4. Social support
a. Appraisal
02
b. Belonging
-02
c. Tangible
03
5. Copingresources
a. Mastery
-04
b. Self-esteem
02
6. Copingstyles
a. Anger-in
13
b. Anger-out
14
c. Anger-reflect -04
d. Activecoping
09
7. Copingresponses
a. Activebehavioral 13
b. Activecognitive
09
c. Avoidance
03
8. Positivealcohol
11
expectancies
OutcomeVariables
9. Alcoholconsumption-02
10. Drinking
problems -04
11. Drinking
to cope
06
Demographic
Variables
12. Sex (1 = male,
2=female)
06
13. Race (I = nonblack,
2=black)
-18
14. Age (years)
-06
15. Education(years)
15
16. Occupation
(I = whitecollar,
-20
2=bluecollar)
(68)
51
(86)
-10
-18
-07
-14
-19
-11
(71)
52
49
(68)
53
(72)
- 25
-23
- 26
-21
26
27
25
32
22
22
(76)
61
(85)
35
22
29
17
-16 -21
-25 -19
-19
-06
09
20
-24 -14
-08 -10
11
08
21
29
-25
-14
28
20
-31
-23
32
29
(69)
55 (70)
-31 -48
-17 -18
-01 -01
-02 -07
22
12
15
11
02
01
-21 -27
06
10
01 -02
-18 -24
07
-02
-29
-02 -01
-04 -09
32
20
-16
-13
-19
-27
29
20
01
-08
-26
-05
-09
-26
-03
-11
-24
-08
-13
-31
04
13
24
03 -02
13 -10
18 -07
00
-08
-07
-05
01
04
00
-01
08
07
16
23
16
26
46
4
3
04
-05
-11
-04
10
08
-16
00
06
04
08
03
-28
02
07
17
11 -09
17
13 -07 -07
01 -12 -02
06
14 -15
00 -08 -14
12
08 -08
15
19 -22
07
21
10
03
-15
-09
-03
-15
-02
-10 -11
01
09
17
03
-15
-04
00
-13
-08
23
-04 -06
03 -24
18 -01
07
-04
-18
-05
-14
-19
-10
05
-05
-10
-31
01
07
20
09
14
11
-06
06
01
00
-08
04
11
04
-04
0
-06
Note: N = 574. Except for sociodemographic variables, all measures are scored so that highernumbers indicate more of the me
alpha) are on the diagonal. Decimals are omitted. All correlations '
266
could socializeor
drinkingto manage or cope with negative withwhomtherespondent
emotions(e.g., to forgetyour worries,to relax).
Copingresourcesexaminedin thepresent
cheerup whenyou'rein a bad mood).
duringthe study included masteryand self-esteem.
Average alcohol consumption
past 12 monthswas estimatedfromstandard Mastery(Pearlinet al. 1981; seven items)
questionsemployedin the assessed the extentto whichindividualssee
quantity-frequency
National Health and Leisure Time Survey themselvesas controllingthe forces that
affecttheirlives. Self-esteem
refers
(Wilsnacket al. 1984). We convertedre- strongly
of self-worth;
we
sponses into averagenumberof drinksper to one's generalperception
day: one drinkwas definedas 12 ouncesof assessed this resource with Rosenberg's
scale.
beer, fourounces of wine, or one ounce of (1965) 10-itemself-esteem
In this studywe used two measuresof
.5 ouncesof
hardliquor(i.e., approximately
absolutealcohol). (See Stacyet al. 1985 for activecoping style("JohnHenryism"[JH]
and two meathereliability
and valid- active coping, anger-reflect)
evidencesupporting
sures of avoidant coping style (anger-in,
measuresof alcoholuse.)
ityof self-report
We assessed drinkingproblemswith 17 anger-out).JH activecopingstyle(Jameset
style
itemsdesignedto yielda DSM-III diagnosis al. 1983; 12 items)is a stress-coping
of alcoholabuse and dependence,takenfrom characterizedby the belief that one can
the National Instituteof Mental Health control one's environment,coupled with
Diagnostic Interview Schedule (Robins, directand active effortsto do so. Anger
Helzer, Croughan,Williams, and Spitzer copingstylesassess how individualscharacreact or behave when theyfeel
1981;also see Robins,Helzer,Croughan,and teristically
data). angryor upset. Anger-reflect
(Harburgand
Ratcliff
1981forvalidityandreliability
problemsinclude going on Gleiberman1986; four items) assesses the
Representative
control
typically
"bingesor benders,"losinga job becauseof extentto whichrespondents
drinking,and having "blackouts" or the their anger in an effortto address the
"shakes." In thepresentstudywe computed underlyingproblemor the cause of their
the anger.Anger-in(Spielberger
et al. 1985; six
an indexofdrinking
problemsbycounting
totalnumberof problemsthathad occurred items)assesses the extentto whichrespondentstypically
suppressor avoiddealingwith
one timeor morein thepast 12 months.
Workdistress.Workdistresswas assessed theirangryfeelings.Conversely,anger-out
et al. 1985; six items)measures
witha six-itemscale developedby Kandelet (Spielberger
intensity
scale, the degree to which individualsengage in
al. (1985). On a four-point
respondentsstatedhow botheredor upset, aggressive behaviors when motivatedby
frustrated,
pleased(reversescored),etc. they angryfeelings.
job
Coping responseswere assessed by the
felt when thinkingof theirday-to-day
Health and Daily Living Coping Response
experiences.
rated
Personal and social resources. We in- Index(Moos et al. 1986). Respondents
cluded four broad categoriesof resource the extentto which theyused each of 32
variablesas covariatesand examinedthemas coping responsesto deal with a recently
eventor situation.We
potentialmoderatorsof the work distress- experiencedstressful
indices: acalcohol relationship:1) social support,2) derivedthreemethod-of-coping
coping resources,3) coping styles,and 4) tivebehavioral,activecognitive,and avoidcopingresponses.All psychosocialresource ance coping. Active behavioralcoping (13
Likert-typeitems)assesses the extentto whichresponmeasuresemployeda four-point
responseformatexcept for the coping re- dentsengagedin directactionand problem
sponse indices, which used a three-pointsolving.Activecognitivecoping (11 items)
assessestheuse of cognitivestrategies
aimed
format.
or redefining
in a
thesituation
Social supportwas measuredby three at minimizing
subscales(fiveitemseach) of the morepositivelight.Avoidancecoping(seven
abbreviated
SupportEvaluationList (Cohen items)assessesrelianceon avoidance,denial,
Interpersonal
forcoping
andHoberman1983): theperceivedavailabil- or tensionreductionas strategies
event.2
ityof 1) tangibleassistanceor materialaid, 2) withthestressful
We assessed
Positivealcoholexpectancies.
appraisalsupport(i.e., the availabilityof a
andtrusted
advisor),and 3) belong- expectancieswith a compositeof six subconfidant
of someone scales takenfromthe abbreviated
versionof
ing support(i.e., theavailability
267
Presure
.276
Work
-.096
Alcohol
Distress
Consumption
023
Lack of
Dnnking
Problems
IndirectEffectsofWorkStressors on AlcoholOutcomes
Alcohol Outcomes
RESULTS
TensionReductionModel of theWork
Relationship
Stress-Alcohol
Alcohol
Consumption
Dnnking
Problems
Work
Pressure
.020
-.026
.006
Lack of
Job Control
.027
-.035
.008
Work
Stressors
Drinking
to Cope
268
low (- 1 sd
variables, abovethemean)and at relatively
action)termsand theirconstituent
we centeredall variables(or deviatedthem below the mean) levels of the relevant
variable.
from their means) before computingand moderator
We believe thatthe presentdata analytic
testingthe hypothesizedinteractions(see
is conservative
withrespectto TypeI
Finneyet al. 1984fora discussionofthisand strategy
otherissues relatedto the estimationand errorwithoutunduly sacrificingstatistical
power. We use an inherently
conservative
effects).
of modifier
interpretation
procedurein whichthevariables
We estimatedseparateequationsforeach hierarchical
interestmustmake significant
moder- of theoretical
of thefivecategoriesof hypothesized
contributions
aftera largenumators and for each of the three alcohol independent
outcome measures. Thus a total of 15 ber of potentiallyconfoundingsociodemoequationswas estimated.Analyses graphicand psychosocialresourcevariables
regression
arecontrolled,
andwe probeinteractions
only
in Table 2.
are summarized
We conductedfollow-upanalyses only when the entireblock of producttermsis
at the conventional
significant
p
when the block of interactionsattained statistically
significanceat the conventionalp < .05 < .05 level.
to cope relationlevel. We probed significantinteractions Workdistress-drinking
describedbyCohenand ship. As shownin the firsttwo columnsof
following
procedures
Cohen (1983). Specifically,we generated Table 2, copingresources,copingstyles,and
separateregressionlines fromthe overall alcohol expectanciesinteractedsignificantly
thedistress- with work distress in the predictionof
equationto represent
regression
to cope. The magnitude
of all effects
at relatively
high(+ 1 sd drinking
alcoholrelationship
TABLE 2. Summary of Hierarchical Moderated Regression Analyses Predicting Alcohol
Outcomes from Work Distress, Personal and Social Resources, and the Distressby-ResourceInteractions
Drinkingto Cope
Moderator
Variables
AR2
AlcoholConsumption
AR2
Drinking
Problems
AR2
DF
Step 1:
.147**
143**
21,552
.318**
All maineffects
Eq. 1, Step2:
.004
.002
3,549
.007
WD x social support
.042
1,549
-.044
-.056
Appraisalsupport
- .056
- .051
- .020
1,549
Belongingsupport
.048
1,549
.056
.016
Tangiblesupport
Eq. 2, Step2:
.002
.000
2,550
.008*
WD x copingresources
.009
.039
-.009
1,550
Self-esteem
- .008
- .100*
1,550
- .050
Mastery
Eq. 3, Step2:
.009
4,548
.014*
.009
WD x copingstyles
.016
.062
-.023
1,548
Anger-in
1,548
-.018
.016
.098*
Anger-out
.088*
1,548
.100*
-.010
Anger-reflect
.012
1,548
.038
.095*
JHactivecoping
Eq. 4, Step2:
.005
.012*
3,549
.008
WD x copingresponses
- .027
- .018
1,549
.030
Behavioralcoping
- .024
.091*
.050
1,549
Cognitivecoping
-.008
.123**
.031
1,549
Avoidancecoping
Eq. 5, Step2:
.001
.004
1,551
WD x expectancies
.011**
- .037
.108**
.064
1,551
Positiveexpectancies
* p<.05; **p<.01.
Note: WD = Workdistress.Main effectsforthefollowing21 variableswereenteredin a blockon thefirststep:
sex, age, race, education,occupationalstatus,tangiblesupport,appraisalsupport,belongingsupport,mastery,
activebehavioralcoping,active cognitivecoping,
active coping,anger-in,anger-out,anger-reflect,
self-esteem,
avoidancecoping,positivealcohol expectancies,workpressure,lack of job control,and workdistress.Results
these21 maineffectsmaybe otainedfromthefirstauthoron request.
concerning
269
270
directeffectof .17
we obtainedan estimated
Moreover,
problems.8
sd increasein drinking
althoughthepresenceof multiplevulnerability factorsmay predisposean individualto
outcomes
experiencenegativealcohol-related
in responseto highlevels of workstress(cf.
Bry et al. 1982), few individualsin this
sample of employedpersons
representative
DISCUSSION AND CONCLUSIONS
were actually high on more than one
factor.For example, only 33
This researchtesteda model of stress- vulnerability
in relationto theexperience people (about6% of the sample)were even
induceddrinking
of work stressorsand negativeemotional one-halfof one standarddeviationabove the
sequelae. Overallwe foundno supportfora meanon bothavoidancecopingandexpectansimple tension reductionmodel of work cies; only six (less than 1%) were a full
drinking.Of three separate standarddeviationabove the mean on both.
stress-induced
provedequallyrare.Thus
alcohol outcomesexamined,not one was Othercombinations
by exposureto work it seems plausiblethatthe most vulnerable
predictedsignificantly
in the
in factareunderrepresented
stressorsand the experienceof work-relatedindividuals
negative emotion after both sociodemo- workforce,perhapsbecausetheylack many
graphicand psychosocialresourcevariables of the skillsrequisiteto successfulemployand
we obtained ment.Hencetheimpactof workstressors
Furthermore,
were controlled.7
only qualifiedsupportfor a social learning work distress on alcohol outcomes may
be limitedby therelativecompedrinking.On necessarily
model of workstress-induced
theone hand,data showingenhancedvulner- tence of those who are able to secure and
employment.
withlimitedcoping maintain
abilityamongindividuals
ofpeoplewho
abilitiesand strongpositivealcoholexpectan- Becauseofthesmallnumber
logic of the appearlikelyto experiencenegativealcohol
cies supportthe fundamental
social learningperspective.On the other outcomesin responseto work stress and
of
hand, these effectswere neitherrobustnor because of thegenerallysmallmagnitude
ourfindings
suggestthat
consistentacross multipleindicatorsof a theobtainedeffects,
programsaimed at reducingand
or acrossmultipleoutcome prevention
givenmoderator
managingwork stressmay not be a costmeasures.
way to addressalcoholproblemsin
Indeed, the only consistentpatternof effective
was obtainedfor the workforce.Althoughsuch an approach
effects
interaction
significant
drinkingto cope. This patternof findings may prove usefulin reducingotherstressmay be relatedproblems(e.g., depression,absenteesuggeststhatindividualdifferences
it seems unlikelyto
fordrink- ism, low productivity),
morelikelyto governmotivations
impacton alcoholpathology.
or have significant
ingthanto influencealcoholconsumption
targetedto individuals
problemsper se. Instead,interventions
the experienceof drinking
withboth knownor suspectedto have alcohol-related
is consistent
Such an interpretation
social learningtheory(Abramsand Niaura problemsmay be more appropriate(e.g.,
Archer 1977;
1987) and empiricalresearch(Cooper et al. constructiveconfrontation,
motivations Trice and Roman 1978). In addition,strate1988), whichshowthatdrinking
of gies aimed at alteringnormsthatpromote
maymediatetheinfluence
at leastpartially
of the
or at structural
features
on alcohol heavydrinking
individual
differences
predisposing
thatallow heavydrinking
workenvironment
use and abuse.
Despite limited support for theoretical maybe useful(Triceand Roman1978).
predictionsmade by the social learning Results of the present study suggest
model, our findingssuggestthat the real- severaldirectionsforfutureresearch.First,
theneedto controlfor
worldimpactof workstressorson alcohol- ourresultsunderscore
to
in modelspurporting
relatedoutcomesis likelyto be small.Indeed, individualdifferences
even among individualsidentifiedas most examinethe impactof work stressorsand
outcomes.Zerovulnerable(e.g., those high in avoidance distresson alcohol-related
as reportedin Table 1,
of workstressors ordercorrelations,
effect
coping),theestimated
betweenwork
relationships
problemswas revealsignificant
on the experienceof drinking
small:foreach sd increasein workdistress, stressorsand distressand several alcohol
resources,expectancies,and distress(H2e).
effects
Finally,we obtainedcounterintuitive
for several indicatorsof active coping,
JHactivecoping,and
angerreflect,
including
activecognitivecoping.6
271
viewoftheetiologic
for sociodemo- ablymorecircumscribed
outcomes.Aftercontrolling
graphicand psychosocialresourcevariables, roleof workstressin alcoholpathology.
relationhowever,we observedno significant
ships between work stress and alcohol NOTES
variables.This outcomesuggeststhatsignificant zero-orderrelationshipsmay reflect 1. In the interestsof parsimonywe have made
in theproposed
severalsimplifying
assumptions
characdue to sociodemographic
confounding
model.First,drinking
to cope, alcoholuse, and
deficitsin personal
teristicsor to underlying
drinkingproblems are treated as separate
and social resources.
indicatorsof alcohol pathology,with the
Second, futureresearch using domainassumption
of similarconsequentstatusin our
specific measures of personal and social
model. Althougha causal order may exist
resourceswould providea morecompelling
amongthesedependent
measures(cf. Cooperet
of
al. 1988), we have chosen not to model
testof our model.Indeed,theassumption
relationships
amongthemin orderto maintain
consistencyimpliedby the
cross-situational
withotherpublishedresearchin
comparability
use ofgeneralmeasuresofpersonalandsocial
the workstress-alcohol
literature
wherethese
(see Swindle
resourcesmaynotbe warranted
outcomes
have
been
treated
universally
as
et al. 1988 for a recentdiscussionof this
of alcohol-related
discreteindicators
pathology.
issue). Consequentlythe failure to use
The net effectof any bias introduced
by this
domain-specificmoderatormeasures may
ofouroutcomemeasuresshouldbe to
treatment
moderatorefunderestimate
systematically
themagnitude
of the workstress
overestimate
fects.
effectson alcohol use and drinking
problems.
in viewoftheessentially
nullresults
Third, futureresearchusing prospective Therefore,
in thispaper,modelinga causal order
reported
causal
designswould enable moreconfident
inference.Yet insofaras workstress-related amongtheseoutcomeswouldnotalterourbasic
conclusions.
we specuimportant,
effectsare etiologically
characteristics
are
Second,sociodemographic
is likelyto be played
latethattheirinfluence
not modeled as potentialmoderatorsof the
out over the courseof hoursor days rather
workstress-distress
or distress-alcohol
relationday at work
thanyears.That is, a stressful
ships. Althoughit is possible that some
maybe followedbyincreasedalcoholuse that
subgroupsdefined,for example, by racial/
ethnic or socioeconomiccharacteristics
are
evening;a stressfulweek at work may be
more vulnerablethan othersto work stress
followedby heavieruse thatweekend.Hence
effectson alcohol-related
outcomes,such efwithina designsensitiveto suchshort-term,
not
relevant
to an adequate
fects
are
specifically
a
fluctuations
diary
study)
daily
(e.g.,
person
test of either tension reductionor social
may be needed to test work stress-related learningmodels. Such analysesare therefore
effectson alcoholuse moredefinitively.
beyond the scope of the presentpaper. In
weak
Finally,in keepingwiththegenerally
thesetests,however,we do notmean
omitting
resultsreportedin this and several other
to imply that these effects may not be
recent studies (e.g., Mensch and Kandel
important.In particular,gender has been
identified
as a potentially
moderator
important
1988), we believe that increasedresearch
to HorwitzandWhite(1987),
effect.According
attention
shouldbe focusedon otherpossible
stylesof
influence
on drinking malesand femalesmayhavedifferent
sourcesof work-related
is
pathology;female deviance hypothetically
usefulareas forfuture
behavior.Potentially
characterizedby internalization
of distress,
featuresof the
researchinclude structural
whereas male deviance is more outwardly
workenvironment
thatpromotedrinkingor
directedand antisocial(see Robbins 1989 for
allow it to go undetected,as well as
relateddata). Hence femalesmaybe proneto
that
workplacenorms,cultures,or networks
internalize
stress-related
whereasmales
effects,
maybe morelikelytouse alcoholandtoexhibit
encouragedrinking.
overtlyor covertly
alcohol-related
problemsas a resultof workof this study,we
Despite the limitations
relatedstressorsand distress.The available
findlittlesupportforthe widelyheld belief
data, however, lend little supportto this
alcohol
thatworkstresspromotes
problematic
hypothesis.Data reportedby bothParkerand
use. Althougha smallnumberof individuals
Farmer(1988) and House et al. (1986) suggest
who are deficientin adaptiveresourcesor
littledifference
betweenmales and femalesin
who hold strongpositiveexpectanciesfor
the magnitudeof effectsfor work-related
alcohol's effectsmay be vulnerable,our
on alcoholuse.
stressors
findingssuggestthe adoptionof a consider- 2. Accordingto LazarusandFolkman(1984), low
272
reliabilitymay reflectan
internal-consistency
rather
property
of thecopingconstruct
inherent
perse. In particthaninadequatemeasurement
ular,theseauthorsarguethatthesuccessfuluse
withina givencopingdomain
of one strategy
may obviatethe need to use others.Thus the
low reliability
obtainedforseveralof
generally
the coping indices in the presentstudymay
reflectaccuratelythe underlyingnature of
coping processes. Nonetheless,we must acreliability
knowledgethat lower-than-optimal
variablesmay
amongsome of our moderator
effects 6.
attenuateestimatesof theirinteractive
(Baronand Kenny1986).
equationmodelingthat
3. The use of structural
takesintoaccountthe effectsof measurement
overstandard
errorhas severalclearadvantages
(Bollen 1989;Jameset
pathanalytictechniques
equationmodel- 7.
al. 1982). Even so, structural
ingalso has a numberof drawbacks,severalof
whichmilitateagainstits use in the present
context.First,a large numberof interaction
effectsare specifiedin the presentmodel;
fortesting
sucheffects
amonglatent
techniques
variablesare complexand carrya numberof
stringentassumptionsthat in practice are
to meet(Bollen 1989). Moreover,the
difficult
consequencesof violatingthese assumptions
(Kennyand
are notgenerallywell understood
Judd 1984). In addition,as Schmitt(1989)
pathanalyticmodelsmay
pointedoutrecently,
equation models
be preferableto structural
insofaras theyare morereadilyaccessibleto a
wider audience. Hence in at least some
themethodological
gains associated
situations,
equationmodeling
withthe use of structural
maybe morethanoffsetby theaccompanying
loss in comprehensibility.
and drinking
prob4. Both alcohol consumption
lems were highlyskewed(skewness = 4.64
To reduceskewness,we
and4.08 respectively).
both variables(skewness =
log-transformed
forconsumption
and
.01 and 1.77 respectively
verforproblems);using the log-transformed
sions, we reranall analysesinvolvingeither
dependentvariable.Resultsobtainedby using
variablesdidnotdiffer
substanthetransformed
tivelyfromthosebased on thenontransformed
variables.Thus onlyresultsusingtheuntransin thisstudy.
formedvariablesare reported
5. To assess the effectof deletinginfrequent
drinkers,we reestimatedthe path model,
whodrankless thanonce
including
respondents
a month.Resultswere essentiallyidenticalto
in Figure2, exceptthatthe
thosesummarized
path fromworkdistressto drinkingto cope
became significant
(p = .035). This change,
however,appearedto reflectan increasein
powerdue to thelargersamplesize (N = 711
vs. 574) ratherthanan appreciableincreasein
ofthepathcoefficient
themagnitude
(B = .088
273
274
AlcoholMisuse." HumanRelations40:833-52.
Matteson,Michael T. and JohnM. Ivancevich.
WorkStress.San Francisco:
1987. Controlling
Jossey-Bass.
Menaghan,ElizabethG. 1983. "IndividualCoping
between
Efforts:
Moderators
of theRelationship
Life Stressand MentalHealthOutcomes."Pp.
157-91inPsychosocialStress:TrendsinTheory
and Research, edited by H.B. Kaplan. New
York:AcademicPress.
Mensch,BarbaraS. and Denise B. Kandel. 1988.
"Do Job ConditionsInfluencethe Use of
Drugs?"JournalofHealthand Social Behavior
29:169-84.
Miller,Annetta,KarenSpringen,JeanneGordon,
AndrewMurr,Bob Cohn,Lisa Drew,andTodd
Barrett.1988. "Stresson theJob." Newsweek,
April25, pp. 40-45.
Moos, RudolfH., RuthC. Cronkite,AndrewG.
Billings,andJohnW. Finney.1986. Healthand
Daily LivingForm Manual. Revised ed. Palo
Alto: Social Ecology Laboratory,Veterans
Administration
and Stanford
Medical
University
Centers.
275
Moos, RudolfH., JohnW. Finney,andDarrowA.
Chan. 1981. "The Process of Recoveryfrom
Alcoholism.I. ComparingAlcoholic Patients
and MatchedCommunity
Controls."Journalof
Studieson Alcohol42:383-402.
Olkinuora,Martti.1984. "Alcoholismand Occupation." ScandinavianJournalon WorkEnvironment
and Health10:511-15.
Parker,Douglas A. and Gail C. Farmer.1988.
"The Epidemiologyof Alcohol Abuse among
EmployedMen and Women." Pp. 113-30 in
RecentDevelopments
in Alcoholism,
Volume6,
editedby M. Galanter.New York:Plenum.
Pearlin, Leonard I., Elizabeth G. Menaghan,
MortonA. Lieberman,and JosephT. Mullan.
1981. "The StressProcess." JournalofHealth
and Social Behavior22:337-56.
Pearlin,LeonardI. and Clarice W. Radabaugh.
1976. "Economic Strains and the Coping
Functionsof Alcohol." AmericanJournalof
Sociology82(3):652-63.
Pearlin,Leonard I. and Carmi Schooler. 1978.
"The Structure
of Coping." Journalof Health
and Social Behavior19:2-21.
Polich, J. Michael and Bruce R. Orvis. 1979.
AlcoholProblems:Patternsand Prevalencein
theU.S. AirForce. SantaMonica:Rand.
Robbins, Cynthia. 1989. "Sex Differencesin
Psychosocial Consequences of Alcohol and
Drug Abuse." Journalof Health and Social
Behavior30:117-30.
Robins, Lee N., John E. Helzer, Jack L.
Croughan, and KathrynS. Ratcliff. 1981.
of MentalHealthDiagnostic
"NationalInstitute
InterviewSchedule."Archivesof GeneralPsychiatry38:381-89.
Robins,Lee N., JohnE. Helzer,JackCroughan,
JanetB.W. Williams,and RobertL. Spitzer.
1981. NIMH Diagnostic InterviewSchedule:
VersionIII. Rockville,MD: NationalInstitute
of
MentalHealth.
Rogosa,David. 1979. "Causal Modelsin Longitudinal Research: Rationale, Formulation,and
Interpretation,"
Pp. 263-302 in Longitudinal
Researchin theStudyofBehaviorand Development,edited by J.R. Nesselroadeand P.B.
Baltes. New York:AcademicPress.
Rohsenow,Damaris J. 1983. "DrinkingHabits
and Expectanciesabout Alcohol's Effectsfor
Self versus Others." Journalof Clinical and
51:752-56.
Consulting
Psychology
Rosenberg,Morris.1965. Societyand theAdolescentSelf-Image.Princeton:
Princeton
University
Press.
Sadava, S.W., R. Thistle,and R. Forsyth.1978.
of Alcoholand
"Stress,Escapismand Patterns
Drug Use." Journal of Studies on Alcohol
39:725-36.
Schmitt,Neal. 1989. "Editorial." Journal of
AppliedPsychology
74(6):843-45.
Seeman,Melvinand CarolynS. Anderson.1983.
276