You are on page 1of 15

Sage Publications, Inc.

American Sociological Association

Stress, Social Status, and Psychological Distress


Author(s): Ronald C. Kessler
Source: Journal of Health and Social Behavior, Vol. 20, No. 3 (Sep., 1979), pp. 259-272
Published by: American Sociological Association
Stable URL: http://www.jstor.org/stable/2136450
Accessed: 27-10-2015 01:36 UTC

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://www.jstor.org/page/
info/about/policies/terms.jsp

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.

Sage Publications, Inc. and American Sociological Association are collaborating with JSTOR to digitize, preserve and extend
access to Journal of Health and Social Behavior.

http://www.jstor.org

This content downloaded from 128.238.66.100 on Tue, 27 Oct 2015 01:36:57 UTC
All use subject to JSTOR Terms and Conditions
Stress,Social Status,and Psychological
Distress
RONALD C. KESSLER
Universityof Wisconsin-Madison

Journalof Health and Social Behavior 1979, Vol. 20 (September):259-272

It has long been knownthatpersons in socially disadvantaged statuses are particularlylikelyto


exhibitsymptomsofpsychological distress. Whythisis so remainsthe subject of controversy.
This paper sheds some lighton the controversyby analyzing survey data on stress, social
status, and psychological distress. Two components of the status-distressrelationshipare
isolated empirically:(1) a differential
exposureto stresscomponent,interpretable as one sortof
social causation influence;and (2) a componentdue to the differential impact of comparable
stresseson people in different categories of variousstatus dimensions.The analysis shows that
differentialimpactis themoreimportantdeterminantin relationshipsbetweensocial class, sex,
and maritalstatus and self-reported distress. Onlyin the comparisonof whitesto nonwhitesis
differential exposure the key determinant.The implicationsof these findingsfor our under-
standingof the social status-psychological distressrelationshipare discussed.

A centralproblemto emergefromthe study place over a successionof generationsand


of sociodemographiccorrelatesof mentaldis- beingcharacterized byan inabilityto moveup,
order has been to explain why persons in cer- as wellas bya tendency todriftdown,theclass
tain disadvantaged social statuses have the hierarchy. This explanationhas been associ-
highrates of emotional distress they do. The ated withthe"drift"hypothesis in thelitera-
earliestwork on this problemconcerned itself tureon socialclassandmentalillness,andwith
with social class (Faris and Dunham, 1939; the"social selection"hypothesis in thelitera-
Hollingsheadand Redlich, 1958; Langner and tureon maritalstatusandmentalillness.How-
Michael, 1963). But over thelast two decades a ever, it is betterdescribedgenericallyas a
great deal of research has also been directed "constitutional frailty"explanation,2because
towardthe overrepresentation of certainemo- itemphasizesintrapsychic propensities to dis-
tional problems among the unmarried(Bach- tressratherthanenvironmental influences. (2)
rach, 1975), women (Weissman and Klerman, A second set of explanations,on the other
1977), and nonwhites (Dohrenwend and hand,has emphasizedthe importance of life
Dohrenwend, 1970).1 conditions as determinantsofemotional disor-
Two generalsortsof explanationshave been der. Althoughnot limitedto one hypothesis,
offeredforthe relationshipbetween disadvan- theseexplanations havebeendominated bythe
taged social status and emotional distress. (1) view that exposureto stressfulexperiences
The firstsays that persons who either have leads to emotionaldistressand thatthedistri-
chronic emotional problems, or are prone to butionof stressful experiencesin societyac-
develop such problems,are channeled intothe countsfordifferential ratesofemotional prob-
ranks of the unmarried and into the lower lems.Together thesetwoexplanations suggest
class, the latter channeling possibly taking that sociallydisadvantagedpersonswill be
bothmorehighlyexposedto stressful experi-
Workon thispaperwas partiallysupportedby ences and also more highlyinfluencedby
NIMH GrantsNo. 5 T32 MH14641-02and 1 R03 stressfulexperiencesthansociallyadvantaged
MH32490-01. I wishto thankJeromeK. Myersand persons.
hisassociatesforgenerously metouse their
allowing
datain thispaper.Theirresearchwas supportedby Documentation of theinfluence of differen-
PHS ContractNo. 43-67-743and ResearchGrant tialexposure,ifit couldbe done,woulddem-
No. MH15522 fromthe National Instituteof Mental onstrate thatthegreater psychological distress
Health,Department
of Health,Educationand Wel- of disadvantaged personsis due, at least in
fare. I wish to thank Cris Laskowski for providing
computerassistanceand Ron Diamondfor dis- part,to greaterexposureto distress-provoking
cussions.The comments
oftheanonymous
review- environmental experiences.In contrast, there
ers wereextremely
helpful. area number ofwaysto interpret a differential

259

This content downloaded from 128.238.66.100 on Tue, 27 Oct 2015 01:36:57 UTC
All use subject to JSTOR Terms and Conditions
260 JOURNAL OF HEALTH AND SOCIAL BEHAVIOR
impactacrosstwosetsofindividuals. One is in disadvantaged social statusesand an indicator
termsof psychicfrailty. That is, because of ofemotional functioning.As shallbe seen,this
theirrelativeinabilityto manage stressful analysissheds lighton why personsin less
situations-their lack of social competence- advantagedsocial statusesare in relatively
some people mightexhibita greatdeal more poor emotionalhealth.
emotionaldistressthanotherswho are more
robust.The sociocultural environment can be METHOD
the directsource of differential impactin a
numberofways.One oftheseis bydetermin-Data
ing the subjective meaningsof objective The data used herecomefromtheMyerset
events. Life situationsare not inherently al. (1972,1974,1975)New Haven survey.The
stressful,
ofcourse.Theymustbe interpreted.sample consistsof 720 personsinterviewed
These interpretations, in turn,are partlyde- bothin 1967and 1969froma largerrandom
pendenton the contextof one's life,the rel- sampleof adultsin the New Haven mental
evancies and circumstancesto which one healthcatchment area.
selectivelyattends.To the extentthatsocial Theinterviews includeda seriesofquestions
status shapes interpretations of life experi- aboutexposureto potentially stressfullifeex-
ences, the subjectivestressfulness, and thus periencesthatmightlead to psychological dis-
theimpact,oftheseexperiences willvarywith tress.I have focusedon fivesummary mea-
status. suresforpurposesof thisanalysis.Threeof
Therearealso a number ofwaysinwhichthe theseare life-event scales,createdfroma list
socialenvironment can modify theimpactsof of56eventsintheinterview. Allofthesescales
stresseventsas theyare subjectively experi- pertainto lifeeventsoccurring duringthetwo
enced.One oftheseis byinfluencing thecop- yearsbetweenthetwodatapointsofthepanel.
ing strategiesused by personssubjectedto The othertwo stressmeasurespertainto on-
stress.Justas objectiveoccurrences are inter- goinglifesituations thatcan be stressful.3The
pretedthrough subjectivefilters, so one's op- firstis a measureoffinancial status,definedas
tionsare determined by subjectiveappraisals a ratioofincometo number offamily members
of availability
and efficacy.One's positionin supportedby this income.The second is a
thesocialstructure also definesone's access to measure of physical illness, definedas a
variousresourcesthatcan be usefulin dealing weightedsumof self-reported symptoms.4
withstress.The stress-buffering effects of so- Therehas been considerabledebatein the
cial supportsystemshave been widelydis- life-events literature about the best way to
cussed (Cobb, 1976; Dean and Lin, 1977; conceptualizeglobal event indexes (B. P.
Kaplanand Cassel, 1977;Mechanic,1974),as Dohrenwend,1974;Dohrenwend et al., 1978;
have themoreobviousadvantagesofpolitical Vinokur andSelzer,1975).To provideas broad
and financialpower. an analysisas possible,resultsare shownhere
Empirically itis verydifficultto disentangle foreach of two separateeventconceptualiza-
thesevariousmodifying influences.However, tions.The first uses rawcountsofundesirable,
a researcher can estimatethe relativeimpor- desirable,and totallifeeventsas thethreein-
tance of differentialimpactin explaining the dexes of exposure.5This unweighted coding
relativefunctioning oftwoormoresetsofindi- schemeis one thathas been used in muchof
vidualswithoutteasingoutall thevariousin- theworkreportedin thelife-events literature.
fluencesthatgo to make up thisdifferentialThe secondemploysweighted countsofdesir-
impact.In theanalysisreported below,thisis able andundesirable eventsanda dummy vari-
exactlywhatis done.A setof stressmeasures able to describethosewho have experienced
are used to predictself-reported emotional no eventsat all overthecourseof thepanel.
distressamong a sample of persons inter- The weightsused hereare thosepresentedin
viewedin a majorepidemiological survey.I Holmesand Masuda(1974)forthelife-change
examinetherelativeimportance ofdifferentialunits (LCU) scale. The weightedcoding
exposureto thesestressesand of differentialschemeis a modification of the Myerset al.
impactof comparablestressesin accounting (1974) Desirability-Change Life Events
for observedrelationships betweenvarious Index.6

This content downloaded from 128.238.66.100 on Tue, 27 Oct 2015 01:36:57 UTC
All use subject to JSTOR Terms and Conditions
STRESS, SOCIAL STATUS, AND PSYCHOLOGICAL DISTRESS 261

The measureofemotional distressusedhere distressacross these subsamplesin termsof


is Macmillan's(1957)symptom checklistscale theestimation equationsobtainedin the sub-
as modifiedbyGurinet al. (1960).The scale is samples.
madeup of20 statements aboutbodilyfeelings If we takeas our model
thatindicatemoodsofdepressionandanxiety.
Overthecourseofthetwoyears,thescale has P= IV(S) + a, (1)
coefficient
a stability of from .50 (females)to
.56(married) inthesubsamples examinedhere, thenwe can systematically compareaverages
whichcan be takenas a lowerboundon the in thedistressscorebetweentwo subsamples
of thescale.
reliability (g,h)in termsof thefollowing decomposition
Althoughthe symptomsincludedin this (Winsborough and Dickinson,1971).
scale are taken fromthose presentedby
patientsin treatment, it shouldbe recognized Pg Ph = ISh(Vg Vh) + 1Vh(Sg - Sh)
thattheydo nottap a person'sneedfortreat- + I(Vg - Vh)(Sg - Sh)
ment.Nor do theyallowone, in a community + (ag - ah). (2)
sample,to definepsychiatric "cases." How-
ever,thescale does permitrespondents to be If we take as our model
orderedon a continuum of reporteddistress.
Scores on the scale rangefrom20 (all symp- ln [pW/(l-p)] = V(S) + a, (3)
tomsare reported as occurring "often")to 80
(all symptomsare reportedas occurring thenwe can systematically compareodds of
"never"). The sample-wide mean was 72.8. extreme distress(treated as a dummydepen-
Only16%ofthesamplereported extremedis- dentvariable,wherepi is thepredictedprob-
tress,defined, following Myerset al. (1974),as abilitythatpersoni is extremely and
distressed
a scoreof 66 or lower. p = E pi/n)in termsof thefollowing decom-
position(Kessler,1979):
Analysis Strategy ln[pI(l1-P)]g = e - Vh)]
[YSh(Vg
I have recentlypresentedan analysisstrat- ln[gI(l1-P)]h x e [Vh(sg- Sh)]
egy for studying the relativeimportanceof x e [l;(Vg Vh) (Sg - Sh)]
differentialexposureto stressand differential x e [ag - ah] (4)
impactof stress(Kessler,1979).This strategy
is used below to comparemenwithwomen, Thesedecompositions perfectly divideaver-
whiteswithnonwhites, the marriedwiththe agedifferences indistressintocomponents due
unmarried, andpersonsinthelowerclass with uniquelyto each of fourconceptually distinct
personsin themiddleand upperclasses. influences.8 In bothequations(2) and (4), the
The strategyis based on a model of the firstcomponentis a differential exposureef-
causal nexus betweenstressand emotional fect;theexpectedresultofgivingmembers of
functioning thatassumesemotional distress(P) statuscategory g theexposuresfoundto char-
is influencedby exposureto stressors(S) act- acterizemembersof statuscategoryh. The
ingon personswhovaryintheextentto which secondcomponent ineachoftheseequationsis
stressinfluencestheiremotionalfunctioninga differential impacteffect; theexpectedresult
(V). I assumethatthisimpactvariesdepending ofgivingmembers of statuscategory g theim-
on the intrapsychic defensesand social re- pactcoefficients of personsin statuscategory
sourcesusedbytheindividual and h. The thirdcomponentis an empiricalesti-
to interpret
managethestressesto whichhe or she is ex- mateoftheinteraction betweentheimpactand
posed. Thismodelcan be expressedas a pre- exposurecomponents. Thelastcomponent is a
dictionequationfora particular setofstressor residual,representing thedifference indistress
measures,whereV is expressedas a metric betweenpersonsinthetwocontrast categories
regression The comparisonstrat- standardized
coefficient. on theinfluences modeledin the
egyconsistsof,first,estimating modelsofthis prediction equations.
sortin subsamplesand, next,systematically The exposureandimpactcomponents areof
comparing observeddifferences in emotional the most substantive interest, for they tellus

This content downloaded from 128.238.66.100 on Tue, 27 Oct 2015 01:36:57 UTC
All use subject to JSTOR Terms and Conditions
262 JOURNAL OF HEALTH AND SOCIAL BEHAVIOR
therelativeimportance stresses by meansof statistical
of differential controls.Therefore, in
and differential influence ofcomparable stress- each of the decompositions reportedbelow I
es on symptomdifferences across the two sequentially includeand deletetheinitialdis-
statussubsamples.The interaction termrepre- tressscale as a wayofassessingtheinfluence
sents the extentto whichthe symptom dif- of selectionon myresults.
ferencesexplainedbytheprediction equations Decompositionof mean differences in the
cannotbe attributed uniquelyeitherto sub- distressscoresoftwocontrast subsampleslets
sample differencesin exposure or to dif- us glimpsetheforcesthatmake"typical"per-
ferencesin impact.Whentheinteraction com- sons in thesubsamplesdiffer fromeach other
ponentis largerelativeto the exposureand in emotionalfunctioning. Decompositions of
impactcomponents, thenit becomesimprac- differences inthelikelihood ofextreme distress
ticalto place a greatdeal of emphasison the give us a somewhatdifferent view, one of
relativemagnitudes of the lattertwo compo- "extremes"ratherthantypicalpersons.It is
nents. notalwaysthecase thatstressinfluences typi-
In theanalysesbelow,equationsoftheform cal and extremepatternsin thesameway. In
(1), (3) are estimatedseparately fora seriesof severalcases below we findthatthe typical
statuscategorysubsamples.These equations personinsubsampleA is less highly influenced
are thendecomposedby meansof (2), (4) to by a stressthanthetypicalpersonin subsam-
evaluatetherelativeimportance ofdifferentialple B, but thatthis stressleads to a higher
exposureand differential impacton the ob- proportion ofextremedistressin subsampleA
servedvariationsin self-reported distressbe- thanin B. The twopatterns together describe
tweenthecontrasted subsamples. one subsample(A) in whichmostpeople are
unaffected bystressbutinwhichthosewhoare
influenced respondin an extreme fashion,and
RESULTS
anothersubsample(B) in whichmostpeople
A numberof criticsof the life-events ap- areaffected bystressbutinwhichfewofthese
proachto studying psychological distresshave peoplerespondin an extremefashion.
notedthatdistressedpersonsfrequently ex- Tables 1 and2 presentR2coefficients forthe
pose themselvesto stress (Brown, 1974; 36 OLS regression and36 individual-level logit
Dohrenwend et al., 1978;Rabkinand Struen- equationsusedtoestimate (1), (3) respectively,
ing,1976).Therefore itcannotbe automaticallyforthedecompositions to be discussedbelow.
assumedthatrelationships observedbetween Comparison ofthecoefficients inTables 1 and
exposureandemotional disorderareduesolely 2 shows thatthe estimation equationsmore
to thecausal impactof stresson distress.The completely predictthedistressscale thanthe
causal influencemightwork the otherway "extreme"distressdummyvariable.This re-
around.One wayto guardagainstthissortof sultis to be expected,sincethesizableskewin
confounding is to includein one's eventin- theextremedistressdichotomy has theeffect
dexesonlythoseeventsthatare relatively un- of attenuating correlations betweenpredicted
affected by theprioremotionalfunctioning of and observedscores.
the focal respondents(Brown, 1974). How- The firsttwo coefficient columnsin each
ever,thisapproachforcestheanalystto dis- tableare takenfromequationsestimated with
card the life stressesthatdiscriminate most the unweightedpredictors.The thirdand
clearlybetweenpersonsinsociallyadvantaged fourthcoefficientcolumnsare taken from
and disadvantagedstatus categories-such weighted-predictor equations.Comparisonof
eventsas job layoff, arrest,evictionfromone's the coefficients in thesecolumnsshowsthat
home,andmarital dissolution. Fortunately,the theunweighted predictor set has a higherre-
factthatthisanalysisis concernedwithpanel lationshipto the distress scale than the
data makesit possibleto adjustforthe con- weightedset in 35 of the 36 pairsof coeffi-
founding ofcausaldirection inanother way,by cients.I do not wantto enterintothe con-
controlling Time 1 distressin thedecomposi- troversyabout the "best" way to concep-
tions. In this.way I can includeeventsthat tualize life-eventscales here. However, in
might be partially confounded by selectionef- examiningthe resultsof the decompositions
fectsintheindexesandpartialouttheseeffects reportedbelow,one shouldbear in mindthat

This content downloaded from 128.238.66.100 on Tue, 27 Oct 2015 01:36:57 UTC
All use subject to JSTOR Terms and Conditions
TABLE 1. Coefficients
of MultipleDeterminationin the Subsample OLS Regression

R2
Unweighted Weighted
(N)* Unweighted with G1 Weighted with G,
Men (318) .214 .417 .196 .402
Women (402) .222 .379 .127 .327
Classes I-III (173) .101 .368 .069 .331
Classes IV-V (547) .238 .400 .168 .361
Class V (223) .334 .454 .192 .378
White (627) .185 .386 .136 .355
Not white (90) .387 .473 .238 .349
Married (541) .149 .371 .097 .340
Sep/div (42) .474 .579 .332 .440
Note: All coefficientsare significantat p < .05. Both unweightedand weightedpredictorsets include
incomeper familymemberand the scale ofphysicalhealthcomplaintsas well as the threelife-eventsindexes.
G1 refersto the Time 1 Gurin symptomscale.
*Missing values on the predictorswere assigned mean scores. Complete data are available for the
criterion.

thedistressscoresmeasuredin thisparticular been unableto move beyondthe positionof


sampleare betterpredicted bytheunweighted theirbirth,are sociallyincompetent andhighly
predictor set.9 influenced by stress.Lower class people ap-
pear to be at a disadvantagewithrespectto
Social Class
learnedcoping responses.In theirChicago
study,Pearlinand Schooler(1978)foundthat
The literature on socialclass and mentalill- incomeand educationwere both positively
ness has emphasizedthe importanceof the correlated andto
totheuse ofcopingstrategies
distributionof stressacrossclasses as thekey the possession of intrapsychiccoping re-
determinant of class differencesin emotional sources foundto be effectivein managing
disorder(MarkushandFavero,1974;Myerset stress.Thereis also considerable indirectevi-
al., 1974).However,thereis good theoretical dence (Dohrenwendand Dohrenwend,1970)
reasontobelievethatthereshouldalso be class thatlowerclass peoplehavelow access to so-
differencesintheimpactsof stress.The social cial supports.These,inturn,are nowbelieved
selectionhypothesis, forone,arguesthatlower to be extremely importantresourcesinhelping
class personswho have drifteddown from a personcope withlifeproblems(Liem and
higherstatusfamiliesof origin,or who have Liem, 1978:148-51).

TABLE 2. Coefficients
of MultipleDeterminationin the Subsample Logits

R2
Unweighted Weighted
(N)* Unweighted with G1 Weighted with G,
Men (280) .195 .272 .175 .257
Women (344) .146 .297 .099 .265
Classes I-III (148) .082 .292 .084 .280
Classes IV-V (476) .176 .299 .133 .265
Class V (189) .240 .325 .184 .285
White (541) .120 .241 .083 .214
Not white (81) .298 .450 .329 .435
Married (478) .088 .269 .055 .245
Sep/div (39) .654 .725 .379 .284
Note: All coefficientsare significantat p < .05.
*Listwise deletion of missingvalues was used.

This content downloaded from 128.238.66.100 on Tue, 27 Oct 2015 01:36:57 UTC
All use subject to JSTOR Terms and Conditions
264 JOURNAL OF HEALTH AND SOCIAL BEHAVIOR
Table 3 presentsthe resultsof 16 decom- in thelowerclass subsample.The interaction
positionsin whichdifferential impactand ex- and residualcomponentsare comparatively
posure are compared across social class muchless important, contributinga combined
categories.10The firstpanel (A) compares difference of only .35 points.Withonlyone
peoplein themiddleand upperclasses (I-III) exception,the eightcontrastsof mean dif-
withthose in the lower and lowest classes ferencesshow thatthe differential impactof
(IV-V). The secondpanel(B) comparesthose comparablestressorsis moreimportant than
in classes I-ITTwiththeverylowestclass (V). differential exposureto stress.The exception
Withonlyone exceptionthe resultsin these is theweightedprediction of classes I-ITTvs.
twopanelsare consistent witheach other. classesIV-V inwhichtheTime1distressscale
In termsofmeandifferences in self-reportedwas includedas a predictor. Heretheexposure
distress,the data are quite clear in showing component(-.83) is sizablylargerthanthe
thatthelowerclassesaredisadvantaged intwo impact component(-.39).11
respects:Theyare exposedto morestressful These patternsare consistent withthedata
experiencesthan upper class persons,and reportedby Dohrenwend(1973), who found
comparableeventsimpacton theiremotional thatthepoorinherurbansamplewereexposed
functioning more severelythanon the func- to morestressesthanpersonsin the middle
tioningof upperclass persons.For instance, class and also thatthe relationship between
line6b showsthatthe2.08-point difference in exposureand distresswas higherin thelower
mean symptomsbetweenpersonsin classes thanin themiddleclass. However,myresults
I-III andV is madeup largelyof(1) a 1.00-point are at odds withthefindings of Myerset al.
difference due to thegreaterslopesofdistress (1974) who, analyzingthe New Haven data,
on thestressmeasuresintheregression forthe reported that"therelationship betweensocial
lowerclass subsample,and (2) a .68-point dif- class and psychiatricsymptoms is a reflection
ferencedue to thehighermeanlevelsof stress ofsocialclass differences inthedistribution of

TABLE 3. Decompositionof Class Differencesin Self-ReportedDistress

Predictors Impact Exposure Interaction Residual Total


A. Classes I-III versus Classes IV-V
Decomposition of Observed Mean Differences(I-III-IV-V)
la Unweighted -1.87 -1.44 .47 1.44 -1.37
lb Unweightedwith G1 -1.11 - .90 .09 .52 -1.37
2a Weighted -3.53 .01 -.74 2.89 -1.37
2b Weightedwith G1 - .39 - .83 .24 - .37 -1.37
Decomposition of Predicted Relative Odds* (I-IIIIIV-V)
3a Unweighted .52 -.66 .35 - .38 .84
3b Unweightedwith G1 .72 -.46 .12 -1.05 .52
4a Weighted 1.47 -.68 .62 -1.68 .77
4b Weightedwith G1 .76 -.52 .64 -1.53 .52
B. Classes I-III versus Class V
Decomposition of Observed Mean Differences(I-III-V)
5a Unweighted -2.97 -2.53 1.21 2.21 -2.08
Sb Unweightedwith G1 -2.37 -1.17 .11 1.34 -2.08
6a Weighted -2.54 -1.99 1.91 .59 -2.08
6b Weightedwith G1 -1.00 - .68 -.19 - .16 -2.08
Decomposition of Predicted Relative Odds* (I-IIIIV)
7a Unweighted .55 -1.36 .91 - .55 .63
7b Unweightedwith G1 .65 - .82 .32 -1.24 .32
8a Weighted 1.53 - .81 .73 -2.23 .46
8b Weightedwith G1 1.79 -1.17 1.17 -2.66 .41
*Logit decompositionsare expressed in additiveform.The "Total" column is the ratio of mean relative
odds forthe contrastcategories. For instance,in row 3a, the predictedprobabilitiesof extremedistressfor
personsin classes I-III and IV-V at the means of the independentvariableare .113 and .131 respectively.This
yields a relative odds [.113/(1-.113)]I[.131/(1-.131)] of .84. This total is equal to the product of the
exponentiationsof the fourcomponentsof the decomposition.

This content downloaded from 128.238.66.100 on Tue, 27 Oct 2015 01:36:57 UTC
All use subject to JSTOR Terms and Conditions
STRESS, SOCIAL STATUS, AND PSYCHOLOGICAL DISTRESS 265

Desirability-Change Life Eventsin the com- mightnothave the same subjectivemeanings


munity"(Myerset al., 1974:199-200).Their forlowerand middleclass people.Anotheris
analysisfailedto considerimpactdifferentials,thatlowerclass people, even thoughexperi-
thusmasking thegreaterimportance ofimpact encingthe same feelingsof helplessnessand
thanof exposurein thedata. hopelessnessas middleclass people, have
The secondsetofdecompositions inTable 3 learnedto adopta moreflexiblestyleofemo-
focuses on extremedistress.With extreme tionality. We do not knowenoughaboutthe
distressdefinedas a scale scoreof66 orlower, sociologyofemotionsto assess thispossibility
13% of thosein classes I-III, 17% of thosein systematically. But it is generalclinicallore
class IV, and 21% of thosein class V report thatpeoplevarybetweenveryrigidand very
extremedistressat Time2 of thepanel (I-III flexiblestylesofemotionality. Flexiblepeople
vs. IV-V: t = 1.3, NS; I-III vs. V: t = 2.07, p < tendto reactquicklyandwithgreatemotionto
.05. The resultshereareunlikethoseformean stress,butseemto recoverquickly,too. Rigid
symptoms in one striking respect.Personsin people,on theotherhand,tendnotto bend,
classes IV and V are less likelythanthosein onlyto break.As a result,whentheydo be-
classes I-III to become extremely distressed come emotionaltheytendto do so uncontrol-
whenexposedto comparablestresses.Despite lably.It is quitepossiblethattheflexibleand
thisadvantage,thepoorreportmoreextreme rigidstylescharacterize lowerand upperclass
distress.The muchgreaterexposureof the personsrespectively; sucha possibility would
poorto stressful experiencescancelsouttheir accountforthe anomalouspatternsof stress
relative advantage in effectivelybufferingimpactobservedin thesedata.12
stressimpacts.
This resultis quiteconsistent, occurring in Race
all eightdecompositions oftheset.The impact
componentsare always positive,indicating Of all thecontrastsexaminedin thispaper,
lowerimpactsamongthelowerclasses. There thereis least supportin the data, and in the
is no systematic magnification or diminution of literatureat large(Dohrenwendand Dohren-
theeffectwhenTime 1 distressis controlled. wend,1969:31;Warheitet al., 1973,1975),for
The effectis more evident,though,when a relationship betweenrace and psychological
classes I-III are comparedonlywithclass V distress.With"nonwhites"definedas thosein
thanwhencomparedwithIV and V together. thesamplewho describethemselves as either
This intriguing patternwill not be pursued "Negro" or "Puerto Rican," no significant
very deeply here. As we shall see below, difference existsin theaveragedistressscores
though,it appearsagainin thecomparison of ofwhitesandnonwhites (73.0vs. 71.8,t = 1.4,
whitesand nonwhites.Whatwe see is that NS).I3 However,nonwhites are twiceas likely
typicallowerclass personsrespondmoreto as whitesto reportextremedistress(28% vs.
stressthando typicalupperclasspersons.This 14%, t = 2.72, p < .05).
is to be expected,giventhe lowerlevels of Table 4 represents the resultsof eightde-
social support,theless effective copingstrat- compositions betweenwhitesandnonwhites in
egies,thesecondarystressesset offforlower thesample.I willhavelittleto sayaboutthese
classpersonswhenseriouslifecrisescomeup, results,for they closely parallel those de-
and themyriadotherwaysin whichthepoor scribedforthesocialclass contrasts inTable 3
are disadvantaged relativeto therichin terms withrespectto the reversalof signsforthe
ofmanaging stress.However,we findpropor- impactcomponentsbetweenthe decomposi-
tionally feweramongthepoorwhorespondto tionof averagesand thedecomposition ofex-
stresswithreallyextremedistressthanamong tremescores.I believethesameconsiderations
thosewho are moreadvantagedeconomically applyhereas in theearlierdiscussionofthose
and socially. results.
One interpretation of thispattern worthy of Thereis one respect,though,in whichthe
futureworkis thatlowerclass personsare resultsforwhitesand nonwhites contrastwith
accustomedto stressful experiencesand,as a the resultsreportedabove for social class;
result,seldom have extremeemotionalre- namely,thatthe greaterdistressamongnon-
sponsesto stress.Thatis, objectivestressors whitesthanamongwhitesis farmorea func-

This content downloaded from 128.238.66.100 on Tue, 27 Oct 2015 01:36:57 UTC
All use subject to JSTOR Terms and Conditions
266 JOURNAL OF HEALTH AND SOCIAL BEHAVIOR
TABLE 4. Decompositionof White-Nonwhite
(Negroand Puerto Rican) Differencesin Self-ReportedDistress

Predictors Impact Exposure Interaction Residual Total


A. Decomposition of Observed Mean Differences(White-Nonwhite)
la Unweighted .15 -2.30 1.07 - .14 -1.22
lb Unweightedwith G1 -.10 -1.56 .67 - .23 -1.22
2a Weighted -.67 -2.11 1.07 .48 -1.22
2b Weightedwith G1 .86 -1.23 .47 -1.35 -1.22
B. Decomposition of Predicted Relative Odds* (White/Nonwhite)
3a Unweighted .71 -1.24 .94 -1.05 .53
3b Unweightedwith G1 .69 -1.19 1.00 -1.07 .57
4a Weighted .42 -1.16 .76 - .68 .51
4b Weightedwith G1 .36 -1.05 .25 - .60 .58
Note: Entrieshave been roundedto two decimal places fromcomputationsbased on fivedecimal places.
As a result,totals may not correspondexactlyto those predictedby addingor multiplying
the components.
*See note to Table 3.

tionofdifferential
exposureto stressthanwas male and femalementalillness(Gove, 1972;
foundto be thecase in thegreaterdistressof Gove and Tudor,1973;Pearlin,1975;Radloff,
lowerclass persons.As we see in panelA of 1975;Weissmanand Klerman,1977).The em-
the table, the average differencein self- phasisin thisliteraturehas beenon theunique
reporteddistressbetweenwhites and non- experiences of women in our society as
whitesis muchmorea function of the manyhousewivesand mothersand on the stresses
stressfulexperiencesto whichnonwhites areassociatedwiththeseexperiences.It has been
exposed than of any impactdifferential be-arguedthat exposureto these stressesac-
tweenwhitesandnonwhites. Indeed,intwoof countsforthepreponderance ofwomenamong
the fourcontraststhe impactcoefficient isthementally ill (Gove and Geerken,1977).
positive,whichmeansthatcomparablestress- Less has been written abouthow menand
es have more impacton whitesthan non- womenmightdifferin theirvulnerability to
whites.PanelB showsthatnonwhites havean stress. However,both biologicalarguments
impactadvantageoverwhiteswhenconsidered aboutfemaleconstitutional frailty
(Weissman
intermsofextreme distress.However,innone and Klerman, 1977) and sociocultural
of the fourdecompositions is thisadvantagearguments aboutlearnedhelplessness (Kaplan,
greatenoughtoovercomethegreater exposure1977) and ineffectivecoping (Pearlin and
of nonwhitesto stress. Schooler,1978)suggestthatstressmayhavea
moresevereimpacton womenthanon men.
Sex
Table 5 presentseightdecompositions of
male-femaledifferences indistress.As inother
Over thelast decade a greatdeal has been data reportedfromsamplesurveys(Dohren-
written
aboutsex rolesand theirmeanings for wend and Dohrenwend, 1976; Gove and

TABLE 5. Decompositionof Sex Differencesin Self-ReportedDistress

Predictors Impact Exposure Interaction Residual Total


A. Decomposition of Observed Mean Differences(Men-Women)
la Unweighted -1.55 -.31 .00 .24 -1.62
lb Unweightedwith G1 - .60 -.31 .13 -.84 -1.62
2a Weighted -3.19 -.19 .24 1.52 -1.62
2b Weightedwith G1 -2.58 -.11 .27 .78 -1.62
B. Decomposition of Predicted Relative Odds* (Men/Women)
3a Unweighted - .92 -.09 -.09 .21 .41
3b Unweightedwith G1 - .39 -.11 -.02 -.29 .44
4a Weighted -1.15 -.14 -.07 .40 .38
4b Weightedwith G1 -1.26 -.05 -.07 .53 .43
Note: Entrieshave been roundedto two decimal places fromcomputationsbased on fivedecimal places.
the components.
As a result,totals may not correspondexactlyto those predictedby addingor multiplying
*See note to Table 3.

This content downloaded from 128.238.66.100 on Tue, 27 Oct 2015 01:36:57 UTC
All use subject to JSTOR Terms and Conditions
STRESS, SOCIAL STATUS, AND PSYCHOLOGICAL DISTRESS 267

Tudor,1973),thedata in thisstudyshowthat counts,formen,sincetheymustbe assertive


womenare, on the average,moredistressed and sociallycompetentin orderto attracta
thanmen(72.1 vs. 73.7, t = 2.8, p < .05). In spouse. Coupledwiththe assumeddisadvan-
addition,womenare significantly morelikely tagesof theunmarried withrespectto coping
thanmento reportsymptoms of extremedis- resources,this hypothesissuggestsstrongly
tress(19.9%o vs. 11.6%,t = 3.1, p < .05). The thattheimpactofcomparablestressors willbe
decompositions in bothpanelsA and B show greateramongtheunmarried thanamongthe
thatdifferentialimpactis by farthedominant married.
influence on theseobserveddifferences.14 In the New Haven data, significant dif-
Whilethereis nothing in thedata to helpus ferencesin average and extremedistress
know why this impactdifferential exists,a scores exist betweenthe marriedand those
morecompleteanalysisofthesedatapresented who are separatedor divorced(averagesof
elsewhere15suggestsstronglythat the less 73.05 vs. 69.19, t = 2.55, p < .05; extreme
adequate copingstrategiesof women,com- scores of 15% vs. 38%, t = 2.99, p < .05), but
paredto thoseof men,and structural condi- notbetweenthemarriedand eithersinglesor
tionsthatmagnify thesubjectivestressfulnessthe widowed.Therefore, the decompositions
of objectiveproblemsforwomenaccountfor presented inTable 6 contrastthemarried only
thegreater impactofcomparable stressexperi- withthe separated/divorced. In theseresults
ences on womenthanon men. confirmation
we see consistent oftheviewthat
theunmarried are disadvantaged bothinbeing
more exposed to stressand in being more
Marital Status by stressthanthe married.
highlyinfluenced
One of the mostconsistently documented The relativeimportance of thesetwocompo-
findings intheepidemiological literatureis that nentsvariesfromone decomposition to an-
themarried areinbettermentalhealththanthe other,but the resultsare consistentin their
nonmarried (Bachrach,1975). A numberof signsin all eightdecompositions. On balance,
reasonsforthisadvantagehave been hypoth- it wouldseemthatdifferential impactis more
esized. First,the marriedare betterbuffered important than differentialexposurein the
psychologically thantheunmarried byvirtueof comparisonof averagesand thatthetwo are
thestrongemotionalattachments thatdevelop equally important in the comparisonof ex-
betweenthemand theirspousesand children. tremedistress.Thesefindings, thatbothcom-
Next,themarried are less exposedto stressful ponentshave a roleto playin thegreaterdis-
lifesituationsthanthe unmarried because of tressof theunmarried, consistent
are entirely
theirgreaterfinancialpower,politicalpower, with recentresultsof Pearlinand Johnson
and social contacts.Finally,it has been hy- (1977).
pothesized(Martin,1976)thatconstitutionally It is not clear fromthe data why the
fragilepersonsare selectedout of marriage. separated/divorced are morehighly influenced
This is especiallytrue,accordingto some ac- than the marriedby comparablestressors.

Differencesin Self-ReportedDistress
TABLE 6. Decompositionof Married-Separated/Divorced

Predictors Impact Exposure Interaction Residual Total


A. Decomposition of Observed Mean Differences(Married-Separated-Divorced)
la Unweighted -4.47 -3.17 .94 2.83 -3.87
lb Unweightedwith G1 -4.71 -2.33 .89 2.30 -3.87
2a Weighted -7.65 - .27 -1.51 5.58 -3.87
2b Weightedwith G1 -5.52 - .08 -1.02 2.75 -3.87
B. Decomposition of Predicted Relative Odds* (Married/Separated-Divorced)
3a Unweighted -2.56 -1.79 1.24 1.67 .24
3b Unweightedwith G1 -2.56 -3.48 3.18 1.56 .27
4a Weighted -1.09 - .80 .38 -.10 .19
4b Weightedwith G1 -3.33 -1.43 1.19 1.57 .14
Note: Entrieshave been roundedto two decimal places fromcomputationsbased on fivedecimal places.
the components.
As a result,totalsmay not correspondexactly to those predictedby addingor multiplying
*See note to Table 3.

This content downloaded from 128.238.66.100 on Tue, 27 Oct 2015 01:36:57 UTC
All use subject to JSTOR Terms and Conditions
268 JOURNAL OF HEALTH AND SOCIAL BEHAVIOR
Unfortunately, thesampleis too smallto com- ficultto disentanglethe effectsof constitu-
pare the impactcoefficients across pairwise tionalmakeup,social resources,and coping
contrastsofthemarried withtheseparated, the styleson the strength withwhichstressim-
divorced,the widowed,and the single.It is pactson emotionalfunctioning. Anyor all of
possiblethata detailedcomparison ofthissort these influencescould modifythe impactof
wouldbe able to shedfurther lighton thede- stress.However,the existenceof negligible
terminants of the impactdifferential, for,as impactcomponents orimpactcomponents that
Gove (1972) has pointedout, any effectthat show low-status persons to be advantaged
selects constitutionally fragilepeople out of allowus to makecertaininferences aboutthe
marriage willcharacterize singlesandpossibly forcesat workin certaincontrasts.
theseparated/divorced, butnotthewidowed. The mostobviouspointto be madeis that
However,thesortof socialsupportdisadvan- contrastsdominatedby differential exposure
tages thataccompanybeing singleprobably components can usuallybe assumedto be un-
existforthewidowedas wellas forsinglesand influencedby constitutional frailty.In the
the separated/divorced. Therefore, the social simplestsense thisis so because sucheffects
selectionargument wouldpredictthattheim- would appear in substantialimpactcompo-
pactdifferentials foundinTable6 wouldfailto nents.However,a morecomplexsortof rea-
appear in contrastsof the marriedwiththe soning,suggested byDohrenwend (1966),adds
widowed.But the argument thatdifferentialextraweightto thisclaim.Let us takewhites
social resourcesaccountforthe greaterdis- and nonwhites in Table 5. Here differences in
tressof theunmarried wouldpredictthatthe mean symptomscores are foundto be due
verysameimpactcomponent wouldappearin overwhelmingly to the greaterstresses to
the marriedvs. widowed contrastas was whichnonwhitesare exposed. There is no
found here for the marriedvs. separated/ meaningful impactcomponentin two of the
divorcedcontrast. decompositions, and the signsof thecompo-
nentsvaryintheremaining decompositions. It
DISCUSSION
might be, then, that whites and nonwhites
sharethe same sortsof copingstrategies and
The mostgeneralconclusionto be drawn resources,and also sharethe same aggregate
fromtheabove resultsis thatdifferential im- characteristicsin terms of constitutional
pact is an important component in mostcon- strengths andfrailties.However,to theextent
trastsbetweensociallyadvantagedand disad- thatdifferences do existit is almostcertainly
vantagedstatuscategories.Over the last de- the case that nonwhitesare disadvantaged
cade therehasbeena tendency forresearchers socioculturally relativetowhites(Dohrenwend
to focustoo narrowly on the hypothesis that and Dohrenwend,1969:49-56).Therefore, if
"personsin low social statusare dispropor- there are any variations in intrapsychic
tionatelyexposedto stressful lifeevents,and strengths, whitesmustbe disadvantaged rela-
that this exposure can explain the higher tiveto nonwhites. The argument hereis that
symptomlevels of those with low social the insubstantial impactcomponent foundin
status"(Markushand Favero,1974:184).This the white/nonwhite decompositioncan be
differentialexposurehypothesis is centralto plausiblyinterpreted as evidence eitherof
Gove's (1972)sex-roletheoryofmentalillness completelyabsent impacteffects-including
andtotheworkoftheDohrenwends (1969)and any due to constitutionalmakeup-or of
Myerset al. (1974)on socialclass and mental counterbalancing impacteffectsof morethan
illness,to namea fewof themostprominent one sort.In the latterinterpretation the only
examples.The resultsof my decomposition plausibleline of thoughtsuggeststhatnon-
analysisshowthatdifferential exposureis only whiteshave a stronger constitutional makeup
partofthestoryforthecontrasts examinedin thanwhites.
thisstudy.For sex differences, infact,thereis A similarsortofargument might be madefor
no evidencethatdifferential exposureplays therelationship betweensocialclass and psy-
any meaningful role in thegreaterdistressof chologicaldistress.Atleastthiswouldseemto
women. be the case forthe contrastsof extremedis-
As I mentioned above, it is extremely dif- tress,wherethe impactcoefficients all show

This content downloaded from 128.238.66.100 on Tue, 27 Oct 2015 01:36:57 UTC
All use subject to JSTOR Terms and Conditions
STRESS, SOCIAL STATUS, AND PSYCHOLOGICAL DISTRESS 269
lowerclasspersonstobe less highly influenced be alcoholics thanmen (Weissmanand Klerman,
by stressthantheirmoresociallyadvantaged 1977), and the poor are less likely to be
counterparts in themiddleand upperclasses. obsessive-compulsivethan the richor upwardly
mobile (Langner and Michael, 1963). There is
However,the impactcomponentsin the de- some clinical speculation, in fact, that certain
compositions of meansymptom scoresare all clustersof problemsare functionallyequivalent,
positive,whichmeansthatimpactis higherin but are structurallymore likelyto appear in cer-
the lowerclass. Much moreanalysisof this tainpersons thanin others.Hysteriain women is
thoughtto be an equivalent of sociopathy in
discrepancy betweenthetwotypesof impact men; depressionin womento be an equivalentof
components willbe necessarybeforeanything alcoholismin men. Were it possible to formalize
can be said abouttheimportance ofconstitu- and substantiatethese equivalences it mightwell
tionaldifferences in the relationship between be that we would conclude, not, as now, that
socialclass and distress. disadvantagedpersons are more distressed,but
only that they are distressed differentlythan
For the remaining two contrasts,between their more advantaged counterparts. At the
menand womenand betweenthemarried and presenttime,though,it is not clear how to make
the separated/divorced, importantimpact these equivalences formally.The best strategy
components appearin theresults.Indeed,for would seem to be to conduct separate analyses
bothcontrastsit wouldseemthatdifferential of multiple domains of distress, coupled with
systematicassessments of the extent to which
impactis moreimportant thandifferentialex- causal processes found to triggerone type of
posure.Here we cometo thegreyarea in our problem in one particularcategory of people
understanding of sociodemographic variations parallelthe triggering of othertypesof problems
in emotionaldistress.We knowthatwomen in other categories of people.
2. There are several different versions of the "so-
and the unmarried sufferbecause of relative cial selection" argument.One of these suggests
inabilitiesto buffer themselves fromtheemo- that early psychological disabilitydifferentially
tionaldistressaccompanying lifestresses.But predisposes an individualto fail in economic af-
why? fairsand in the marriagemarket.To the extent
that the disability is permanent, then, such
Without directmeasuresof someofthefac- people will tend to inflateaggregatemeasures of
torsthatmight plausiblylead to thedifferential psychological disability and distress in these
influence of stress,it is extremely difficult
to statuscategories. However, since thisargument
answerthis question.Althoughthereare a makes no referenceto impairedcoping abilities,
number of theories about the genetic, it is notobvious thatthe people who "drift" into
these disadvantagedsocial statuses will be more
neurochemical,and neurobiologicalfactors highlyinfluencedthan other people by stresses
thatmightgo intoconstitutional differences, that they come to experience as members of
thereis no obvious way to index such dif- these status categories.
ferences insurveydata.However,itis possible A broader selection argument,on the other
to measure coping styles (Pearlin and hand, suggests that predispositionstoward dis-
tress internalto the individual, determined,in
Schooler,1978),social support(Cobb, 1976), part, by the early childhood environment,and,
and manyothersociocultural factorsthatcan in part, by genetic, neurochemical, and
buffer theimpactofstressondistress.Byusing neurobiological factors unrelated to the adult
measuresofthissortto interpret impactcom- environment,are morecommonamongthe poor
thanthe well-to-do,the unmarriedthanthe mar-
ponents,we mightbe able to tease out the ried, women than men, and nonwhites than
contributions ofthevariousplausibledetermi- whites. For the poor and forthe unmarriedsuch
nantsof differential impactdescribedabove. disadvantagesin social competence mustbe due
I havenotbeenable to do thisinthepresent to some sort of social selection process-since
these are bothacquired statuses. For womenand
paper.However,I feelthatfutureworkmust nonwhites, though, arguments about compe-
movein thisdirection ifwe are to disentangle tence usuallyare based eitheron factorsintrinsic
thethorny problemsof causationthathave so to body chemistry(for women) or on interac-
long baffled those of us who study the tions between prenatal environmentalassault
sociodemographic correlates ofmentalillness. and genetic predispositions(for nonwhites).
3. At the same time, these life situations can
NOTES provideresourcesfordealingwithothertypesof
stress. For instance, the lack of financial re-
1. Some emotionalproblemsare less characteristic sources has been hypothesizedto be a central
of socially disadvantaged than of advantaged determinantof the coping disadvantages of
persons. For instance, women are less likelyto blacks and persons in the lower class (Dohren-

This content downloaded from 128.238.66.100 on Tue, 27 Oct 2015 01:36:57 UTC
All use subject to JSTOR Terms and Conditions
270 JOURNAL OF HEALTH AND SOCIAL BEHAVIOR
wend and Dohrenwend, 1970). Similarly,physi- theirposition on the index ratherthan the con-
cal health can be a resource on which one can strainedassignmentof themto thehighestscore.
draw in coping with certainevents that require Given the factthatmy threeindexes relax three
physical intervention.It is generallythe case, in assumptions of the Myers index, it is not sur-
fact, that ongoing life situations or char- prisingto find,as I did in preliminarycompari-
acteristicsthatcan be thoughtof as stressescan, sons, thatmy threeindexes always have higher
when considered fromanother perspective, be R2values in subsample regressionsthandoes the
considered resources. It is possible, then, that Myers Desirability-ChangeIndex.
complex interactions exist in the prediction 7. Since the stress measures correlatesignificantly
equations consideredin thispaper. For instance, withthe Time 2 scale, controllingforthe Time 1
physical health problems might influence the score, we know thatsome real change has taken
emotionalfunctioningof a person who lacks fi- place in symptomsover time.On the basis of the
nancial resources more severely than one who regressionresults,thischange componentcan be
possesses financial resources. Interactions of estimated to be a minimumof 2%o of the ob-
this type, though,will not be examined in the served Time 2 symptomscale variance. On the
analyses that follow. basis of thisinformation,we can adjust the esti-
4. Since the scale of emotionaldistressused in this mate of the scale's reliability upward to a
analysis consists largelyof psychophysiological minimumof .6.
symptoms,great care was taken in constructing 8. The decompositionin (2) perfectlydecomposes
the physical illness index. The interview observed mean differencesin distress. The de-
schedule contained a list of some 44 physical composition in (4) perfectlydecomposes the
health complaints. Some of these were clearly ratio of the meanpredicted log odds of extreme
psychosomatic.To create the scale of physical distress. Although the mean of the predicted
illness, I selected a subset of 16 complaintsfor probabilitiesdoes equal theobserved probability
which a psychosomaticcomponentis probably of extreme distress in the sample, the mean of
of only minorimportance.I arbitrarily gave dou- the predictedlog odds does not equal the log of
ble or triple weights to some of those that I the observed mean odds. Consequently, the
considered on commonsensical grounds to be "Total" columns in Tables 3-6 vary for pre-
the most serious. dicted relative odds, even thoughthey do not
5. Of the 56 eventsin theinterviewschedule, 28 are vary for observed mean differences.
definedas undesirable,14 as desirable,and 14 as 9. One possible reason forthisis thattheeventlists
ambiguous. The ambiguous-eventssubscale has are shorterin the weighted scales than in the
been omittedhere in order to avoid collinearity unweightedscales. FollowingMyers,only those
between the total-events scale and the events that were included in the Holmes and
undesirable-and desirable-eventssubscales. The Rahe (1967) events list were made part of the
overall results of the decomposition are unaf- weightedevent scales. In the unweightedscales,
fected by the decision to exclude the 28 undesirable, 14 desirable,and 56 total events
ambiguous-eventsscale ratherthan one of the were included. In the weightedscales, only 16
others. However, this decision would become undesirable and 10 desirable events were in-
more important if one were interested in cluded.
evaluatingthe significanceof specificslope dif- 10. The theory necessary to develop significance
ferences(see Note 10), in whichcase it would be testsfordecompositioncomponentshas only re-
preferableto include all the component scales centlybeen workedout by Michael Sobel at the
and exclude the scale of total events. Universityof Wisconsin. No machine-readable
6. The Myers Desirability-ChangeIndex is drawn procedures for making the tests yet exist.
up in two steps. First, weighteddesirable-and Therefore,no significancetestsare presentedfor
undesirable-eventindexes are created that are thedecompositionsin Tables 3-6. However, itis
identicalto those I use in thisanalysis. Next, the possible to evaluate the significanceof mean
two indexes are combined, the desirable-event differencesacross contrast subsamples with t-
index being given a positive score and the tests and to evaluate the significanceof slope
undesirable-eventindex a negative score. Re- differenceseitherwith t-testsor pooled regres-
spondents who report no events are given a sion models containing interactionterms. An
score higher than the highest empirical score example of this sortof analysisis givenin Kess-
obtained among those who reportat least one ler (1979:Note 4). However, since the purpose
event. hereis to describe the majorpatternsin the data,
The use of threeindexes in the procedurefol- ratherthan to make inferencesto some broader
lowed in this paper allows the weightingof the populations, this type of significance-testing
desirable and undesirable indexes (implicitly procedure has not been employed.
given equal weightin the Myers procedure) as 11. One reviewer suggestedthat this exception de-
well as theirsigns (constrainedto be +, - forthe serves attentionbecause it occurs in one of the
desirable and undesirableindexes in the Myers most realisticspecifications.However, it should
procedure) to be estimatedempirically.In addi- be rememberedthatthe weightedscales contain
tion,the use of a dummyvariable forthose who fewer items than the unweighted scales (see
reportno events allows empiricalestimationof Note 9), making inconsistentresults based on

This content downloaded from 128.238.66.100 on Tue, 27 Oct 2015 01:36:57 UTC
All use subject to JSTOR Terms and Conditions
STRESS, SOCIAL STATUS, AND PSYCHOLOGICAL DISTRESS 271
models usingthese scales somewhatless persua- Brown, G. W.
sive than they otherwisemightbe. In addition, 1974 "Meaning, measurement,and stress of life
substantiveinterpretation of the resultin row 2b events." Pp. 217-43 in Barbara S. Dohren-
would imply that the same reversal should be wend and Bruce P. Dohrenwend (eds.),
found in row 6b. That this reversal does not Stressful Life Events: Their Nature and
appear calls a substantiveinterpretation further Effects.New York: Wiley-Interscience.
into question. Cobb, S.
12. It is importantto distinguishthis clinical notion 1976 "Social support as a moderator of life
of emotional flexibilityfromKohn's (1969) no- stress." Psychosomatic Medicine 38:300-
tion of intellectualflexibility.A reviewerof an 314.
earlierdraftcommentedthatthereseemed to be Dean, A., and N. Lin
a contradictionbetween my speculation that 1977 "The stress-buffering role of social support:
lower class persons mightbe emotionallyflexi- Problems and prospects for systematicin-
ble and Kohn's findingthatlower class persons vestigation." Paper presentedat the annual
are intellectuallyinflexible.No such contradic- meeting of the American Sociological
tion exists, though.Kohn's findingrefersto the Association, Chicago.
idea thatpersons in the lower class "experience Dohrenwend, B. P.
conditionsof life that fosterconceptions of so- 1966 "Social status and psychological disorder:
cial realityso limitedand so rigidas to impair An issue of substance and an issue of
theirabilityto deal resourcefullywiththe prob- method." American Sociological Review
lematic and stressful" (Kohn, 1977:xliv). The 31:14-34.
clinical notion of emotional flexibility,on the 1974 "Problems in definingand sampling the
otherhand, refersto the idea thatpeople can be relevant population of stressful life
quick or slow to hold back emotional reactions events." Pp. 275-310 in Barbara S.
fromthemselvesand others. It is perfectlycon- Dohrenwend and Bruce P. Dohrenwend
ceivable thatlower class persons could be both (eds.), StressfulLife Events: Their Nature
intellectuallyinflexibleand emotionallyflexible. and Effects.New York: Wiley.
13. The extremely small numbers of respondents Dohrenwend,Bruce P., and Barbara S. Dohrenwend
who were assigned codes as "Japanese," 1969 Social Status and Psychological Disorder.
"Chinese," "Other," or "Unknown" were ex- New York: Wiley.
cluded fromthis contrast. Only 90 respondents 1976 " Sex differences and psychiatric disor-
were assigned codes as either "Negro" or ders." American Journal of Sociology
"Puerto Rican." These were combined, since 81:1447-54.
neitherof these two response categories con- Dohrenwend, B. S.
tainedsufficient respondentsto warrantseparate 1973 "Social status and stressfullife events."
analysis. It should be noted, though, that the Journalof Personalityand Social Psychol-
Dohrenwends (1969) have reportedcertain dif- ogy 9:203-14.
ferences between blacks and Puerto Ricans on Dohrenwend, B. S., and B. P. Dohrenwend
scales like the Gurinscale, some of whichmight 1970 "Class and race as status-relatedsources of
reflectdifferencesin response styles.The results stress." Pp. 111-40 in Sol Levine and
obtained for the combined subsample in this Norman A. Scotch (eds.), Social Stress.
paper, then,should be interpretedwithcaution. Chicago: Aldine.
14. The stresses that have been discussed in the Dohrenwend, B. S., L. Krasnoff,A. R. Askenasy,
sex-role literatureas key determinantsof dis- and B. P. Dohrenwend
tress among women are thoughtto be sex-role 1978 "Exemplification of a method for scaling
stresses, ones that are relatively unique to life events: The PERI Life Events Scale."
women. The impactof these stressesis captured Journal of Health and Social Behavior
in the residualcomponentsin Table 5. As we see 19:205-29.
there,though,women have positive residuals in Faris, Robert, and H. WarrenDunham
six of the eightdecompositions,which suggests 1939 Mental Disorders in Urban Areas. New
that they are less distressed than men net the York: Hafner.
influencesmodeled explicitlyin the prediction Gove, W. R.
equations. Were unique sex-role stresses actu- 1972 "The relationshipbetween sex roles, mari-
ally at work, we would expect to have found tal statusand mentalillness." Social Forces
negativeresiduals. 51:34-44.
15. Data available fromthe author on request. Gove, W. R., and M. R. Geerken
1977 "The effectof childrenand employmenton
the mental health of married men and
women." Social Forces 56:66-76.
REFERENCES
Gove, W. R., and J. F. Tudor
Bachrach, Leona L. 1973 "Adult sex roles and mental illness."
1975 Marital Status and Mental Disorder: An American Journalof Sociology 98:812-35.
AnalyticalReview. DHEW PublicationNo. Gurin, Gerald, Joseph Veroff,and Sheila Feld
(ADM) 75-217. Washington, D.C.: U.S. 1960 AmericansView Their MentalHealth. New
GovernmentPrintingOffice. York: Basic Books.

This content downloaded from 128.238.66.100 on Tue, 27 Oct 2015 01:36:57 UTC
All use subject to JSTOR Terms and Conditions
272 JOURNAL OF HEALTH AND SOCIAL BEHAVIOR
Hollingshead, August B., and FredrickC. Redlich Mechanic, D.
1958 Social Class and Mental Illness. New York: 1974 "Social structureand personal adaptation:
Wiley. Some neglecteddimensions." Pp. 32-44 in
Holmes, T. H., and M. Masuda George V. Coelho, David A. Hamburg,and
1974 "Life change and illness susceptibility." JohnE. Adams (eds.), Coping and Adapta-
Pp. 45-72 in Barbara S. Dohrenwend and tion. New York: Basic Books.
Bruce P. Dohrenwend(eds.), StressfulLife Myers, J. K., J. J. Lindenthal,and M. P. Pepper
Events: Their Nature and Effects. New 1974 "Social class, life events, and psychiatric
York: Wiley. symptoms: A longitudinal study." Pp.
Holmes, T., and R. H. Rahe 191-206 in Barbara S. Dohrenwend and
1967 "The social readjustment rating scale." Bruce P. Dohrenwend(eds.), StressfulLife
Journal of Psychosomatic Research Events: Their Nature and Effects. New
11:213-18. York: Wiley.
Kaplan, B. H., and J. C. Cassel 1975 "Life events, social integrationand psychi-
1977 "Social supportand health." Medical Care atric symptomatology."Journalof Health
15:47-58. and Social Behavior 16:421-27.
Kaplan, H. B. Myers,J. K., J. J. Lindenthal,M. P. Pepper, and D.
1977 "Gender and depression: A sociological R. Ostrander
analysis of a conditionalrelationship."Pp. 1972 "Life events and mental status: A lon-
81-113 in WilliamE. Fann, Ismet Karacan, gitudinalstudy." Journalof Health and So-
Alex D. Pokorny,and Robert L. Williams cial Behavior 13:398-406.
(eds.), Phenomenology and Treatment of Pearlin, L. I.
Depression. New York: SpectrumPublica- 1975 "Sex roles and depression." Pp. 191-207 in
tions. Nancy Datan and Leon Ginsberg (eds.),
Kessler, R. C. Proceedings of Fourth Life-Span Devel-
1979 "A strategyfor studyingdifferentialvul- opmental Psychology Conference: Norma-
nerability to the psychological conse- tive Life Crises. New York: Academic.
quences of stress." Journalof Health and Pearlin, L. I., and J. S. Johnson
Social Behavior 20:100-108. 1977 "Marital status, life-strainsand depres-
Kohn, Melvin L. sion." American Sociological Review
1969 Class and Conformity:A Study in Values. 42:104-15.
Homewood, Ill.: Dorsey. Pearlin, L. I., and C. Schooler
1977 "Reassessment, 1977." Pp. xxv-lx in Class 1978 "The structure of coping." Journal of
and Conformity,2nd ed. Chicago: Univer- Health and Social Behavior 19:2-21.
sity of Chicago Press. Rabkin, J. G., and E. L. Struening
Langner, Thomas S., and Stanley T. Michael 1976 "Life events, stress and illness." Science
1963 Life Stress and Mental Health. New York: 194:1013-20.
Free Press. Radloff,Lenore
Liem, R., and J. Liem 1975 "Sex differencesin depression: The effects
1978 "Social class and mental illness reconsid- of occupation and marital status." Sex
ered: The role of economic stressand social Roles 1:249-65.
support." Journalof Health and Social Be- Vinokur,A., and M. L. Selzer
havior 19:139-56. 1975 "Desirable versus undesirable life events:
Their relationshipto stress and mentaldis-
Macmillan, A. M.
tress." Journal of Personality and Social
1957 "The Health Opinion Survey: Technique
Psychology32:329-37.
forestimatingprevalenceof psychoneurotic
Warheit,G. I., C. E. Holzer III, and S. A. Arey
and related types of disorder in com-
1975 "Race and mentalillness: An epidemiologic
munities." Psychological Reports
update." Journalof Health and Social Be-
3:325-29.
havior 16:243-56.
Markush, R. E., and R. V. Favero Warheit,G. I., C. E. Holzer III, and J. J. Schwab
1974 "Epidemiologic assessment of stressfullife 1973 "An analysis of class and racial differences
events, depressed mood, and in depressive symptomatology:A commu-
psychophysiological symptoms-A pre- nity study." Journalof Health and Social
liminaryreport." Pp. 171-90 in Barbara S. Behavior 14:291-99.
Dohrenwend and Bruce P. Dohrenwend Weissman, M. M., and G. L. Klerman
(eds.), StressfulLife Events: Their Nature 1977 "Sex differencesin the epidemiologyof de-
and Effects.New York: Wiley. pression." Archives of General Psychiatry
Martin,W. T. 34:98-111.
1976 "Status integration, social stress, and Winsborough,H. H., and P. Dickinson
mental illness: Accounting for marital 1971 "Components of Negro-Whiteincome dif-
status variations in mental hospitalization ferences." Proceedings of the Social
rates." Journalof Health and Social Be- Statistics Section. Washington: American
havior 17:280-94. StatisticalAssociation.

This content downloaded from 128.238.66.100 on Tue, 27 Oct 2015 01:36:57 UTC
All use subject to JSTOR Terms and Conditions

You might also like