You are on page 1of 19

The Relative Permeability of Class Boundaries to Cross-Class Friendships: A Comparative Study

of the United States, Canada, Sweden, and Norway


Author(s): Erik Olin Wright and Donmoon Cho
Source: American Sociological Review, Vol. 57, No. 1 (Feb., 1992), pp. 85-102
Published by: American Sociological Association
Stable URL: http://www.jstor.org/stable/2096146
Accessed: 04-06-2015 20:18 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.

American Sociological Association is collaborating with JSTOR to digitize, preserve and extend access to American Sociological
Review.

http://www.jstor.org

This content downloaded from 157.92.4.71 on Thu, 04 Jun 2015 20:18:58 UTC
All use subject to JSTOR Terms and Conditions
THE RELATIVE PERMEABILITY OF
CLASS BOUNDARIES TO CROSS-CLASS FRIENDSHIPS:
A COMPARATIVE STUDY OF THE
UNITED STATES, CANADA, SWEDEN, AND NORWAY*

ERIKOLINWRIGHT DONMOON CHO


Universityof Wisconsin Universityof Wisconsin

The structuralanalysis of classes can be divided into the analysis of class locations and the
analysis of permeabilityof boundariesseparating those locations. Marxist analysis of class
structurehas been primarily concerned with thefirst of these while Weberianclass analysis
hasfocused on the second. We attemptto combine a Marxist structuralclass concept, which
views class locations in capitalist societies as structuredby exploitation based on property
relations, authorityrelations and expertise, with the Weberianconcern with the ways lives of
individualstraverse the boundaries of that structure.We examinepatterns offriendship ties
across class boundaries infour contemporarycapitalist societies. the United States, Cana-
da, Sweden, and Norway. Three empirical conclusions stand out. (1) The property-based
class boundaryis the least permeable of the three exploitationdimensions; (2) the authority-
based class boundary is significantly more permeable than the expertise-based boundary;
and (3) patterns of inter-classfriendships are largely invariant across thesefour countries.

T he structural analysis of classes can be di- withinclass structures.Theirlives criss-crossthe


vided into two distinctbut interconnected classstructurein a varietyof differentways.Class
tasks:specifyinglocationswithinclass relations, structuresmay differ,not only in the distribution
and analyzingthe degree of permeabilityof the of peopleacrossthevariouslocationsin thatstruc-
boundariesthatdifferentiatethese locations. ture, but also in the extent to which people's
Specifyinglocationswithinthe class structure lives are bounded by specific class locations.
constitutesthe core of whatis generallyreferred Analyzing the degree of openness and closure
to as the analysisof class structure.At a theoret- (Parkin1974, 1979)of class boundaries- what
ical level, thisinvolveselaboratingthe conceptu- we call "permeability" - is the second task of
al foundationsfor the criteriawhichdifferentiate structuralclass analysis.
the various locations within a class structure. A familiarexampleof the analysisof the per-
Empirically,this involves studyingthe distribu- meabilityof class boundariesis the studyof so-
tion of populationsacross the locationsand in- cial mobility,both inter-and intra-generational.
vestigatinghow such distributionsvary across Class mobility, however, is not the only issue
time andplace. involved in understandingthe permeabilityof
People do not, however, simply fill locations class boundaries.Patternsof intimatesocial in-
teractionamong people such as marriagesand
* Direct all correspondenceto Erik Olin Wright, friendshipsare also relevantaspectsof the per-
Departmentof Sociology, Universityof Wisconsin- meability of such boundaries.The analysis of
Madison,Madison,WI, 53706. We expressourdeep family formationwithinclass structureshas re-
gratitudeto RobertHauserwho patientlyexplained centlyreceivedconsiderableattentionin debates
the ins and outs of log-linearmobility table analysis, over the properway of analyzingthe class loca-
andto RobertMarewho suggesteda numberof broad tion of marriedwomen in the laborforce (Gold-
strategiesfor analysingthe data. Marc Westernwas thorpe1983,1984;Erikson1984;Wright1989a).
an early collaboratoron this projectwho contributed Less attentionhas beengiven to thepatterningof
much to the preparationof the data and the initial friendshipties across structurallydefined class
forays into the analysis.ASR'scopy editorprovided locations.Analyzingsuch patternsin four con-
extremelyhelpful editorialsuggestionsin the prepa- temporarycapitalistsocieties is the centralob-
rationof the final manuscript. jective of this paper.

AmericanSociological Review, 1992, Vol. 57 (February:85-102) 85

This content downloaded from 157.92.4.71 on Thu, 04 Jun 2015 20:18:58 UTC
All use subject to JSTOR Terms and Conditions
86 AMERICANSOCIOLOGICALREVIEW

THEORETICALISSUES TheConceptof Class Structure


Marxistand WeberianTraditionsof
The specificconceptof class structurewe use has
Class Analysis
been elaboratedin some detail in earlierworks
The two primarysociologicaltraditionsof class by Wright(1985, 1990). In this conceptualiza-
analysis - Marxist and Weberian - have given tion, the class structureof capitalistsocieties is
differentprioritiesto class structureand bound- viewed as being defined by the intersectionof
ary permeability.In a varietyof ways, Marxists threedimensions:a propertydimension(owner-
(e.g., Poulantzas1975;Carchedi1977;Szyman- ship of the means of production),an authority
ski 1983;ResnickandWolfe 1988;Wright1978, dimension(controlover organizationin produc-
1985, 1990) have generallyput the analysis of tion), anda skill or expertisedimension(owner-
class structure(or a closely relatedconceptlike shipof scarceskills).Eachof thesedimensionsis
"modeof production")at centerstage and have viewed as generatinga particularformof exploi-
paidrelativelylittleattentionto the permeability tationin capitalistsociety, whereexploitationis
of class boundaries(foranexceptionsee Bertaux defined as the ability to appropriatepartof the
1977).Incontrast,thepermeability of classbound- socially producedsurplusproduct.Differentlo-
ariesloomslargeintheWeberiantradition,wheth- cations in the class structureare determinedby
er termed"classstructuration" (Giddens1973)or their relationshipsto differentcombinationsof
"closure"(Parkin1974, 1979).This is especially theseformsof exploitation.Thus,some locations
clear in the analysisof social mobility,which is in the class structurearesimultaneouslyexploit-
largelyinspired(if in a somewhatdiffuseway)by ing and exploitedthroughdifferentmechanisms.
Weberianconceptionsof class. Weberianstend Experts,for example,may be capitalisticallyex-
to devotemuchless attentionto therigorouselab- ploited while being skill exploiters.Such loca-
orationof theconceptof class structure.As Burr- tions, which Wright (1985, 1990) has termed
is (1987) and Wright(1990, pp. 313-23) have "contradictory locationswithinclass relations,"
argued,sociologistsworkingin theWeberiantra- correspondclosely to whatin everydaylanguage
ditiontypicallytreatlocationswithinclass struc- is calledthe"middleclass."A classstructure, then,
turesas soft categoriesrequiringonly loose defi- consistsof a set of basic class locationsandcon-
nitionsandrelativelycasualtheoreticaldefense. tradictorylocationswithinclass relations.
Ouranalysiscombinesthe conceptualappara- Throughoutthis discussionwe referto friend-
tus of the Marxisttraditionwith the substantive shipsbetweenexploiterandexploitedon each of
focus of the Weberiantraditionon the intersec- these threedimensionsas crossinga class or ex-
tion of people's lives with the class structure. ploitation boundary.A friendshipbetween an
This marriageof Marxistcategorieswith Webe- unskilledworkerand an expertmanager,there-
rianquestionsis motivatedby a desireto deepen fore,crossestwoboundaries,theexpertisebound-
the micro-analysisof class within the Marxist ary and the authorityboundary,while a friend-
tradition.Our assumptionis that the class pat- shipbetweenanunskilledworkerandanunskilled
terningof friendshipinteractionsis one of the petty bourgeois crosses only one, the property
microprocessesthatshapethewaysin whichpeo- boundary.Expressionssuchas "authority bound-
ple experienceclass structures.Forexample,po- ary"are thereforea shorthandfor "classbound-
litical coalitions across specific class bound- aryon the authoritydimensionof exploitation."
ariesshouldbe facilitatedto theextentthatfriend-
shipties crosstheseboundaries.All thingsbeing OrientingHypotheses
equal,higherlevels of classconsciousnesswould
be expectedin societies in which friendshipties Ourresearchwas not initiallylaunchedto test a
were overwhelminglywith others in the same series of a priorihypotheses.Becausethe litera-
class ratherthandiffusedacrossa varietyof class turein structuralclass analysishas not systemat-
locations. icallyexaminedfriendshipties (andcertainlynot
Our focus here is on the class patterningof in a comparativeframework)thereare no well-
friendshipties ratherthan on their possible ef- establishedhypotheses to test. Our objectives,
fects. We examine the differentialprobabilities therefore,were largelyinductive- to discover
of friendshipscrossing differentkinds of class coherentpatternsandto reportourconfidencein
boundariesin fourdevelopedindustrialcapitalist the observations.
countries:the United States, Canada,Sweden, Nevertheless,a numberof orientinghypothe-
andNorway. ses can be derived from class analysis and the

This content downloaded from 157.92.4.71 on Thu, 04 Jun 2015 20:18:58 UTC
All use subject to JSTOR Terms and Conditions
PERMEABILITYOF CLASS BOUNDARIESTO CROSS-CLASSFRIENDSHIPS 87
Table 1. Rank Orderingsof Relative Impermeabilityof Althoughthese antagonisticinterestsmight not
Class Boundariesto FriendshipTies in Different
TheoreticalPerspectives
inherentlyblock friendshipformationbetween
workersand capitalistswho were not in a direct
Theoretical Ranking
employmentrelationto each other,Marxisttra-
Perspective Property Authority Expertise dition suggests thatsuch friendshipsare unlike-
ly. To the extentthatclass interestsshapevalues
Class interests 1 2 or 3 3 or 2 and ideologies, which in turn affect the likeli-
(Marxianvariant)a
hood and durability of friendships, Marxists
Class interests 2 1 3 wouldpredictthatfriendshipscrossingthe prop-
(Dahrendorfvariant)
erty boundaryare particularlyunlikely, i.e. the
Class habitus 2 3 propertyboundaryshouldbe the most imperme-
(Bourdieu)
able.
Class as opportunity 1 or 2 3 2 or 1 Because there is considerabledisagreement
structureb
amongMarxistsaboutthe importanceof author-
a
The Marxianvariantof the class interestperspective ity and expertiseas bases of class interestsand
predictsthe propertyboundaryto be the most impermeable, class conflicts,it is unclearhow Marxistswould
but providesno clear basis for rankorderingthe authority rank the relative permeabilityof the authority
and expertiseboundaries.
I The "opportunitystructure"perspective predicts the andexpertiseboundaries.Insofaras Marxistsre-
authorityboundaryto be the most permeable,but provides
gard the interestsof managersas closely inte-
no clear basis for rank orderingthe propertyor expertise gratedwiththeinterestsof capitalists,theywould
boundaries. rankthe authorityboundaryas more imperme-
Note: Rankingsare from 1 = most impermeableto 3 = able thanthe expertiseboundary,but this judg-
most permeable. ment would be temperedby the realizationthat
segmentationof labormarketsby credentialsis a
deep sourceof conflict in contemporarycapital-
sociology of friendshipsto guideourexploration ist societies.
of thesepatterns.Thesehypothesesareorganized Class interests(Dahrendorfvariant).Dahren-
aroundthe rankingsof the three exploitation- dorf (1959) arguedthatin contemporarysociet-
based class boundariesby degree of imperme- ies authorityis the fundamentalbasis of class
abilityas shown in Table 1. The predictedrank antagonism.In earlyperiodsof capitalistdevel-
orderdependsupon theoreticalexpectationsof opment,his argumentgoes, authorityand prop-
how property,authorityand expertisegenerate ertycoincided,andthussocialtheoristslike Marx
various obstacles to the formationof intimate mistakenlyidentifiedpropertyas the fundamen-
interpersonal relations.We explorethreeways in tal axis of class conflict.In the twentiethcentury,
whichtheseexploitationmechanismsmightgen- however, the deepening separationof formal
eratesuch obstaclesand therebyshape patterns ownershipof propertyfromsubstantivecommand
of friendshipformation:(1) by structuringthe meansthatpropertyownershiphas declinedas a
interestsof actors,(2) by shapingactors' lifes- basis for class relations.This perspectivewould
tyles;and (3) by creatingdifferentialopportuni- therefore predict that the authorityboundary
ties for informalinterpersonalcontact.Each of shouldbe the most impermeable.To the extent
these causal processes suggests differentrank- that propertyownershipstill confers some au-
ings of the three kinds of class boundariesby thority,the propertyboundarywouldbe expect-
relativepermeability. ed to have intermediatepermeability.Since ex-
Classinterests(Marxianvariant).Marxismhas pertisewithoutorganizationalauthorityconfers
relatively little to say about interpersonalrela- littlecapacityto command,the expertisebound-
tions.Nevertheless,theMarxistapproachto class ary shouldbe the most permeableof the three.
analysis would generallypredictthat the more Class habitus(Bourdieu).As virtuallyall re-
antagonistictwo peoples' class interestsare,the searchon friendshipformationhas argued,one
less likely it is thatfriendshipswill formbetween of the primarymechanismsshapingfriendship
them,i.e., the moreimpermeablewill be the cor- patternsis commonvaluesandlifestyles.All other
respondingclass boundaries.This is clearly the thingsbeingequal,individualswho sharesalient
case for friendshipsbetween workersand their aspects of lifestyle are more likely to become
employers,since for these actorsthereis a direct friendsthanindividualswithhighlydisparatelif-
correspondencebetweentheirconcreteinterper- estyles. As Bourdieu(1984, 1985, 1987) has ar-
sonal relationsand theirgeneralclass interests. gued in his analysis of class habitus,a pivotal

This content downloaded from 157.92.4.71 on Thu, 04 Jun 2015 20:18:58 UTC
All use subject to JSTOR Terms and Conditions
88 AMERICANSOCIOLOGICALREVIEW

determinantof lifestyle is culturalcapital.This We offer no hypothesesbased explicitly on


suggeststhatthe odds of friendshipties between predictionsfromthe Weberiantraditionbecause
expertsandnonexpertsshouldbe particularly low the Weberiantraditionof class analysis is con-
sincepeopleon eithersideof theexpertisebound- sistentwithvirtuallyanyranking.Weberianslike
aryarelikelyto differsharplyin termsof cultural Giddens (1973), who speak with a relatively
capital.Thus, class habitustheorieswould pre- Marxianvoice, wouldlikely supportthe Marxist
dict the expertisedimensionto be the most im- predictions.Others,like Parkin(1979), who em-
permeable.Furthermore, becausewealthandin- phasize the role of social closure in labormar-
come are generallyviewed as crucial bases of kets, wouldbe likely to predicta rankinglike the
lifestyle (althoughless importantthan cultural classhabitusperspective.Weberianswho arepre-
capital),thepropertyboundarywouldbe expect- occupiedwithstatusgroupsandthe formationof
ed to be more impermeablethan the authority symboliccommunitieswouldalso sharethe pre-
boundary. dictions of class habitustheory. Dahrendorfis
Interactionopportunity.Sociologicalanalyses oftenconsidereda theoristin the Weberiantradi-
of friendshipsdecomposethe friendshipforma- tion becauseof his emphasison the importance
tionprocessintotwoconsecutiveprocesses,meet- of bureaucraticorganization,so his variantof
ing andmating.Meetingis the processof strang- classinteresttheorycouldalsobe consideredcon-
ers being convertedinto acquaintances;mating sistentwitha Weberian-inspired approach.Inany
is the conversionof acquaintancesinto friends case, the theoreticaltraditionsportrayedin Table
(Verbrugge1977, pp. 576-77). Althoughmeet- 1 arenot meantto be tightdeductionsfrom first
ing can simplybe a matterof "chance,"typically principles,but rathergeneralorientinghypothe-
it is the resultof people being in situationsthat ses that facilitatethe interpretationof our find-
systematicallyfacilitatefriendshipformation.In ings.
part,this is a questionof spatialproximity,as in
the importanceof neighborhoodof residenceas Relationto PreviousResearch
a factorinfluencingfriendshipformation- peo-
ple oftenmakefriendswithneighborsof dissim- Our researchdiffers from previousresearchon
ilar social position (Huckfeldt1983; Nahemow friendshipsandsocialstratification (e.g.,Holling-
and Lawton 1975). More significantthansheer shead 1949;Curtis1963;Laumann1966, 1973;
proximityforourpresentpurposes,certain"foci" Verbrugge1977, 1979; Jackson 1977) in three
of social interaction,to use Feld's (1981) expres- respects. First, previous researchhas revolved
sion, generatesustainedjoint activities among aroundstatusdefinitionsof socioeconomicstrat-
peopleandthusenhancetheprobabilitiesof peo- ification, whetheror not the term "class" was
ple gettingto knoweachotherin ways thatcould used, whereasour researchis organizedaround
lead to friendship. an economic structuraldefinitionof class. Sec-
Worksitesare an importantinstanceof such ond,within-status friendshipshaveoccupiedcen-
interactionalfoci. Furthermore, many worksites ter stage in these earlierstudies. These studies
involve joint activityamongpeople in different focusedprimarilyon the extentto which within-
class locations, thus creatingopportunitiesfor statusfriendshipsare more likely to occur than
cross-classfriendships,particularly betweenman- cross-statusfriendshipsand on the variabilityof
agersandnonmanagers.These opportunitiesare within-status friendshipsacrosssocialstrata.Less
furtherenhancedby the fact thatmany supervi- attentionhas been given to the focus of our re-
sorsandmanagers,spendsignificantpartsof their search- the extent to which friendshipscross
careersas nonmanagerialemployees.Careerau- status boundaries. Finally, when cross-status
thoritymobilityis undoubtedlymuchhigherthan friendshipshave been studied,systematicatten-
intracareer mobilityacrosseitherthe propertyor tion has not been given to the patterningof such
expertiseboundaries.To the extent that friend- friendshipsacross qualitativelydifferentstatus
ships survivepromotions,then, this would also boundaries.Previousresearchhasbeenconcerned
enhancethepermeabilityof the authoritybound- primarilywith the relationshipbetween the de-
ary.The opportunitystructurearguments,there- gree of statusdifferencesand the likelihoodof
fore,wouldsuggestthattheauthorityclassbound- friendshipties and with status-asymmetriesin
ary shouldbe the most permeable.The opportu- cross-statusfriendshipchoices, i.e., the tenden-
nity structureperspective,however, makes no cy to identifypeople of slightlyhigherstatusas
clearpredictionaboutrelativeimpermeabilityof one's friends (Jackson 1977; Hauser 1981). In
the propertyor expertiseboundaries. contrast,ourstudyof differentialpermeabilityof

This content downloaded from 157.92.4.71 on Thu, 04 Jun 2015 20:18:58 UTC
All use subject to JSTOR Terms and Conditions
PERMEABILITYOF CLASS BOUNDARIESTO CROSS-CLASSFRIENDSHIPS 89

RELATIONSHIP TO MEANS OF PRODUCTION

RELATIONSHIP TO
Owner
OwnerWith Without CREDENTIALED
Employees Employees Nonowner of Means of Production EXPERTISE

Manager- ? Professionals Professionals


Experts

. - - - - - - - - - - - - - - - - - - - -

Semi-
l| Professionas Semi-Professionals

Employers | PBourgeois Managers Supervisors Workers Nonexperts

Managers- Non- Managers Supervisors Non-


Supervisors managers managers

RELATIONSHIP TO AUTHORITY RELATIONSHIP TO AUTHORITY


FOR OWNERS FOR EMPLOYEES

Figure 1. Basic Class StructureTypology


Note: Dashed lines indicate distinctionsthat we cannot make in the analysis because of constraintsof sample size.
Among employersand petty bourgeoiswe cannotmake distinctionson the expertisedimension.We thus treatall employ-
ers andpettybourgeoisas if they were nonexperts.Among managerand supervisorexpertswe cannotmakethe distinction
betweenprofessionalsandsemi-professionalsor the distinctionbetweenmanagersandsupervisors.We treatthis combined
categoryas if it were entirelymanager-professionals.

class boundariesconcernsthe patternsof friend- sionals (occupationsgenerallyrequiringan ad-


ships acrossthe threequalitativelydistinctclass vancedacademicdegree),semi-professionals (oc-
boundaries. cupationsgenerallyrequiringa college or techni-
cal degree), and nonexperts.The specific mea-
suresusedto operationalizethesedimensionsare
STRATEGYOF ANALYSIS discussedin the Appendix.
Class locations within the class structureare
The Typologyof Class Structure
determinedby variouscombinationsof catego-
Forthepurposesof thisanalysis,we trichotomize ries on the three exploitationdimensions.Our
each of the threedimensionsof class structure. limited sample permitsus to define eight class
The propertydimensioncomprisesemployers', locations:employers,pettybourgeoisie,manag-
petty bourgeois (self-employedpeople without er-experts,managers,supervisors,professionals,
employees),andemployees;theauthoritydimen- semi-professionals, andworkers.Thedelineation
sion comprisesmanagers(directinvolvementin of these class locationsin termsof the threedi-
organizational decisionmaking), supervisors(con- mensionsis shownin Figure1. Theseclass loca-
trolover subordinatesbut no involvementin or- tions shouldnot strictlybe thoughtof as distinct
ganizationaldecisionmaking),andnonmanagers; "classes,"but as distinct "locations"in a class
and the expertisedimensioncomprisesprofes- structure,since some of thesecategoriesarecon-
tradictorylocationswithinclass relations.
'We will use theterm"employers"ratherthan"cap- Ourinterest,then,is in class boundaryperme-
italists"throughoutthis analysis since most of our abilityto close friendshipsbetweenpeoplehold-
employersown businessesthatare quite small. ingjobs in the laborforce.Throughoutthis anal-

This content downloaded from 157.92.4.71 on Thu, 04 Jun 2015 20:18:58 UTC
All use subject to JSTOR Terms and Conditions
90 AMERICANSOCIOLOGICALREVIEW

ysis we treatindividualsratherthanfamilies as the same criteriaused for respondents.Friends


the incumbentsof locations within class struc- who were not in the laborforce were excluded
tures (see Wright 1989a, 1990). This treatment fromthe analysis.
of thedatahastwo practicalimplications.First,it After all of the friendshad been allocatedto
meansthatwe ignorethe complex problemsin- class locations,we then transformedthe dataso
troducedby two-earnerfamilies. A friendship thatfriendshipties wouldbe the unitof analysis.
betweentwo women workers,one marriedto a In the U.S. sample,for example, 1,363 respon-
managerandthe otherto a worker,is considered dents provideddata on at least one friendfor a
an exampleof a WORKERIWORKER friendshiptie. totalof 3,114 friends.Thuswe hada U.S. sample
Second,we excludefriendshipsinvolvinghouse- by
of 3,114 friendshipties, eachtie characterized
wives fromthe analysisbecausetheirclass loca- two values- the class of the respondentandthe
tion derivesfrom theirhusband'sclass location class of the respondent'sfriend.2
ratherthandirectlyfromtheirown jobs.
Objectionsto the Structureof the Data
TheFriendship-TiesMatrix
Severalpossibleobjectionscould be madeto the
On the basisof thecategoriesdefinedin Figure1 structureof thedata.First,thedecisionto exclude
we can constructan 8 x 8 matrixof friendship friendswho arenot in the laborforce could bias
ties. This matrixis analogousto a mobilityma- our resultsif our primaryinterestwere to do a
trixin whichone axisrepresentsclass originsand comprehensivesocialstructural analysisof friend-
the second axis representsclass destinations.In shippatterns.However,becauseourinterestis in
the presentcase, one axis representsthe class lo- certainpropertiesof the class structure- name-
cationsof respondentsandthe secondaxis repre- ly, the differentialpermeabilityof class bound-
sents the class locationof respondents'friends. aries- the relevantdataarefriendshipties that
The cells in the matrixdesignatetypesof friend- linkpeopleoccupyinglocationsin thatclassstruc-
ship ties, not individuals.The diagonalof this ture.In anyevent,ourrespondents'dataarefrom
matrixrepresentsfriendshipties withina single a laborforce samplein most countriesandintro-
class location; the off-diagonalcells represent ducing friends from outside the labor force
friendshipties thatcross class locations. (housewives,unemployed,retirees)intotheanal-
Our analyticaltask is similarto that encoun- ysis wouldcreatean asymmetryin the matrices.
teredin the analysisof mobilitymatrices,i.e., we Second,restrictingtheanalysisto threefriends
wantto analyzethe relativelikelihoodof differ- ratherthanto a comprehensivelist of friendsmay
ent cross-classfriendships.If the likelihoodof introducedistortionsin the results.Hollandand
friendshipties linkingan employerwith an em- Leinhardt(1973) have arguedthat this kind of
ployeeis muchlowerthanthelikelihoodof friend- fixed-choice procedurefor studyingfriendship
shipties linkinga professionalwith a nonexpert, patternsmay introducesevere measurementer-
then we will say that the propertyboundaryis ror in sociometrybecause it forces people with
less permeablethanthe expertiseboundary.The morefriendsthanthe fixed numberto arbitrarily
statisticalstrategyfor modelingdifferentialrela- omit some friendsandforces people with fewer
tive odds of friendshipties is standardlog-linear
mobilitytableanalysis. 2 We ignore the issue of asymmetriesin friend-
The dataarefromthe ComparativeProjecton ships, i.e., a friendshipbetween a respondentworker
ClassStructureandClassConsciousness(Wright and a managerfriendis treatedthe same as a friend-
1989b). While the scope of the samples vary ship between a respondentmanager and a worker
slightlyacrossthecountries,all arebasicallyran- friend. To test whetherthe data roughly fit this as-
dom samplesof adultsin the laborforceranging sumption,we conducteda test of global asymmetries
in size from 1,145in Swedento 2,577 in Canada. aroundthe diagonalof the friendshipmatrixby com-
Respondentswere asked to thinkof the three paringthe goodness-of-fitforthequasi-independence
model (l's on the off-diagonalcells and fully differ-
people - friendsor relativesbut not immediate entiatedcells on the diagonal)with a model thatdif-
familymembers- to whomtheyfelt "personal- ferentiatescells above the diagonalfrom cells below
ly closest."(We referto thesepeopleas "friends" the diagonal.The model that differentiatedthe off-
even though some of them are relatives.) Re- diagonal cells in this way did not significantly im-
spondentswere then askeda series of questions prove the fit and thus the hypothesis that the data
about the jobs of each of these friends. The were roughlysymmetricalabove andbelow the diag-
friends'class locationswere coded using nearly onal could not be rejected.

This content downloaded from 157.92.4.71 on Thu, 04 Jun 2015 20:18:58 UTC
All use subject to JSTOR Terms and Conditions
PERMEABILITYOF CLASS BOUNDARIESTO CROSS-CLASSFRIENDSHIPS 91

thanthefixed numberto arbitrarily includesome models.The most conservativechoice wouldbe


nonfriends.We believe this problemis unlikely to use the numberof originalrespondents.This,
to seriouslydistortourresults.Unlike the socio- however,treatsthe dataas if we hada sampleof
metricworkreportedby HollandandLeinhardt, only one friendperrespondentandwouldthere-
we arenotstudyingwithin-group sociogramssuch fore understateour confidencein whateverpat-
as thepatternsof friendshipsamongchildrenin a ternswe observe.The least conservativechoice
classroom.In examiningsocial ties acrossclass wouldbe to treatthe numberof friendshipties as
boundarieswe are not interestedin the patterns the samplesize. However,becausethese friend-
of reciprocity,closure,asymmetry,etc., amonga ship ties arenot independentof each other,they
particulargroupof individuals,butratherthelike- arenot a randomsampleof ties. We do notknow
lihood of close friendshipsamong individuals of any definitivestatisticalrationalefor deciding
withparticularclass attributes.It is not clearthat wherein thisrangewe shouldset oursamplesize.
the kinds of distortionsHolland and Leinhardt We thereforereportresultsusingbothsamplesiz-
reportedin the shapeof sociogramsapplyto our es. Because thereare slightlydifferentnumbers
analyticalcontext.3 of friendsper respondentin differentcountries,
Finally,it could be objectedthatby lookingat the actualcoefficientsvaryslightlyin the analy-
"close" friends we are ignoring the important ses using the two differentsamplesize figures.
problemof the"strengthof weakties"(Granovet-
ter 1973). It is possible thatcertainclass bound- AlternativeApproachesto Conceptualizing
ariesaremoreeasily permeatedby casualfriend- Permeability4
shipsthanby intimateinterpersonal relations.By
restrictingour attentionto close friends,we may Therearetwo waysto conceptualizetheproblem
understatetheoverallpermeabilityof boundaries of "boundarypermeability" in theclass structure.
to interpersonalrelations.However,oursis not a Thefirststrategysees thisstructureas an arrayof
comprehensiveanalysisof all dimensionsof the theeightnominallocationsdefinedin Figure1.A
permeabilityof boundaries.Such an analysis friendshiptiebetweenanytwodifferentlocations
wouldincludemarriagepatterns,mobility,neigh- crosses a specific boundary.Thus, for workers,
borhoodinteractions,andparticipationin volun- therearesevenpossibleboundary-crossing friend-
tary associationsas well as casual friendships. ships:WorkerlEmployer, WorkerlPettyBourgo-
Ouranalysisis a reasonablefirststepdesignedto is, WorkerlManager-Expert, WorkerlProfessional,
examineone particulartype of social interaction WorkerlSemi-Professional, WorkerlManager and
thatcrossesclass boundaries- intimatefriend- WorkerlSupervisor. Forsupervisorstherearesix
ships. For this examinationit is appropriateto additional boundary-crossingfriendships (the
restrictthe datato the "threepeople with whom WorkerlSupervisor boundaryhas alreadybeen
one is personallyclosest." counted).Among the eight class locations,there
arethus 28 boundariesacrosswhich friendships
can be formed.One approachto measuringrela-
TheProblemof SampleSize
tivepermeabilityof boundaries,then,wouldbe to
Obviously,thenumberof friendshiptiesobserved measuretherelativepermeabilityof eachof these
is considerablylargerthanthenumberof respon- 28 boundariesandrankorderthemfromhighest
dents in the originalsample- the totalnumber to lowest degreeof permeability.
of respondentsforwhomwe hadappropriate data The secondstrategyanalyzesdirectlythethree
in the four countriesis 4,896, whereasthe num- underlyingexploitationmechanismsthatgener-
ber of friendshipties is 11,782. It is not obvious ate the locationsin the class structure:property,
whatsamplesize shouldbe used to calculatethe authority,andexpertise.Thesemechanismsmight
standarderrorsfor the variouscoefficientsin our be thoughtof as more fundamentalthan class
locationbecausethe conceptof class structureis
I Furthermore, even in theirown analyticalcontext constructedby combiningthese mechanismsin
Holland and Leinhardt(1973) stated that "a gross differentways. Dataanalysiswouldtheninvolve
structuralfeaturelikesex-cleavagein children'sgroups
I In the originalversion of this paper,the first of
will not be affectedby fixed-choiceerrorsif thereare
morechildrenof bothsexes thanallowedchoices . . ." these strategieswas adopted.One of theASRreview-
(p. 103). Ouranalysisof class boundarypermeability ers objected to this approachand suggested some-
resemblesan analysisof a gross structuralfeaturelike thing along the lines of the second. In the end, the
sex-cleavagemore thanit resemblesthe fine-grained same substantiveconclusionsare supportedby either
analysisof sociometricpatterns. strategy.

This content downloaded from 157.92.4.71 on Thu, 04 Jun 2015 20:18:58 UTC
All use subject to JSTOR Terms and Conditions
92 AMERICANSOCIOLOGICALREVIEW

assessingthe relativedensitiesof friendshipsin- betweenprofessionalsand nonexpertsare treat-


volving exploiterand exploited within each of ed as crossing the expertiseboundary,whereas
these threeunderlyingmechanisms. friendshipsbetween semi-professionalsand ei-
We employ both strategies,althoughour em- therprofessionalsor nonexpertsare not. In each
phasisis on the second.The substantiveconclu- case, a boundary-crossing friendshipis thus de-
sions are supportedby both strategies.The bulk fined as a friendshipthatcrosses two "levels"of
of the analysisrevolves arounda series of log- the relevantdimension.
linearmodelsof the relativelikelihoodof friend- Because of limits in samplesize, a numberof
ships across the property,authority,and exper- decisionshad to be madein operationalizingthe
tise boundaries.We thenuse the first strategyto threeboundary-crossing variables.First,thebroad
analyzethe boundarypermeabilitybetween the "managerexpert"classlocation(whichcombines
workingclass andotherclass locations. manager-professionals, supervisor-professionals,
manager-semiprofessionals, andsupervisor-semi-
professionals)posed a problemin constructing
ANALYSISAND RESULTS
the EXPERTISE and AUTHORITY variables.We de-
The dataanalysistests threemodels: cided treatthis categoryas if it were composed
(1) The relativepermeabilityof the threeunder- entirelyof manager-professionals. Thus, friend-
lyingexploitationboundariesintheclassstruc- ships betweenworkersandmanager-expertsare
ture:property,authority,andexpertise. treatedas crossing both the expertiseboundary
andthe authorityboundary.
(2) The relativepermeabilityof boundariesbe- Second,in the variableAUTHORITY, employers
tweenthe workingclass andotherclass loca- aretakenas havingfull managerialauthorityand
tions. aretreatedas managers,whereaspettybourgeois
(3) Thepresenceof countrydifferencesin theper- aretreatedas havingno authorityandthusclassi-
meabilityof thethreeexploitationboundaries. fied as nonmanagers.It could be arguedthatthe
petty bourgeoisie makes organizationalpolicy
TheRelativePermeabilityof the Three decisions and should thus be treatedas manag-
ExploitationBoundaries ers, but to maintaina sharpconceptualdistinc-
tion betweenthe propertyboundaryand the au-
We constructthreevariablesto test the relative thorityboundary,we treatthe managervalue on
permeabilityof theboundariesconstitutedby the the authorityvariableas requiringauthorityover
threeexploitationmechanismsin the class struc- people, notjust things.
ture:PROPERTY (a measureof friendshipsacross Third,the EXPERTISE variableposes problems
the propertyboundary),AUTHORITY (a measure for the two self-employedcategoriesbecausewe
of friendshipsacrossthe authorityboundary,and distinguishexpertsfromnonexpertsonly among
EXPERTISE (a measureof friendshipsacross the employees.Althoughemployersandperhapsthe
expertiseboundary).Thesevariablesarepresent- pettybourgeoisiecould be considered"experts,"
ed in Figure2. In operationalizingthe conceptof we classifiedall employersandpettybourgeois,
class structurewe trichotomizedeachof thethree regardlessof credentialsor occupationalactivi-
underlyingdimensionsof classrelations.A friend- ties, as nonexperts.In the exploitation-mecha-
ship tie is treatedas crossing a boundaryon a nism approachto class analysis, the ownership
particulardimensionof exploitationif it linksthe of scarceskills is a basis of exploitationinsofar
extremecategoriesin thesetrichotomies.Forex- as it enablesactorswho sell theirlaborpoweron
ample,the propertydimensionis trichotomized a labor marketto obtaina "rent"componentin
intoemployers,pettybourgeois,andemployees. theirwages. Becauseemployersandpettybour-
Onlyfriendshipsbetweenemployeesandemploy- geois do not sell theirlaborpowerand,withrare
ers aretreatedas crossingthepropertyboundary. exceptions,credentialsarenot requiredfor their
Friendshipsinvolvingthe pettybourgeoisiewith positions,expertisein the sense we are using it
eitheremployersor employees are thus treated becomeslargelyirrelevantfor shapingtheirclass
as not crossing the propertyboundary.On the interests.'Nevertheless,it would have been de-
Au-moRfry variable,friendshipsbetweenmanag- sirable to distinguishexperts from nonexperts
ers and nonmanagersare treatedas crossingthe
authorityboundary,whereasfriendshipsbetween I In oursample,no morethan4 percentof employ-
or
supervisorsandeithermanagers nonmanagers ers andpettybourgeoiswere credentialedprofession-
are not. On the EXPERTISEvariable,friendships als (doctors,dentists,lawyers,accountants).

This content downloaded from 157.92.4.71 on Thu, 04 Jun 2015 20:18:58 UTC
All use subject to JSTOR Terms and Conditions
PERMEABILITYOF CLASS BOUNDARIESTO CROSS-CLASSFRIENDSHIPS 93

0 0
0
Ho.
cju
.......... 0~~~~~~~~~
00 > > 0 o

0 0
U~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~....... - >)W

z -~~~~~~~~~~~~~~~~~~~~~~...
- o -.....

0- .-.0
0
E

.00.~~~~~~~~~~~~~~~~~~~~~~~~......

.0~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~........
a 32 E44 0 E

c ) 0
c)~c) c) c)c i5$
D

S~~~~~~~; Y
0.0
........... . ..... .....

C OOO O O O O0 0
C,,~~~~~~~~~~~~~~~~~~~~~~1
0c O OIt 0O O 00
0
O O 4.E
0

D~~~~~~~~~
Ct
CL m X o E -

o - o o o o
0a --
o - -

This content downloaded from 157.92.4.71 on Thu, 04 Jun 2015 20:18:58 UTC
All use subject to JSTOR Terms and Conditions
94 AMERICANSOCIOLOGICALREVIEW

amongemployersandpettybourgeois,the issue Table 2. Coefficients for Permeabilityof the Three Ex-


is not only how expertise shapes exploitation- ploitationBoundaries:Adultsin the LaborForce
in FourCountries,1980-1982
based interests, but also how it shapes life style
andinformalinteractionopportunities.Oursam- Sample Size Defined as
ple size, however, makes this impossible.The
Numberof Numberof
issue, then,is whetherthe self-employedshould FriendshipTies Respondents
be globallytreatedas "professionals,""semipro-
fessionals,"or "nonexperts"in constructingthe Variable CoefficientAntilog CoefficientAntilog
EXPERTISE variable.On the basis of the occupa- BoundaryPermeabilityCoefficients
tional distinction that we use to define nonex-
PROPERTY -.991* .37 -.980* .38
perts for employees, the large majorityof both (.087) (.135)
employersandpettybourgeoisfall into the non- AUTHORITY -.319* .73 -.296* .74
expertcategory(69.8 percentof employersand (.034) (.052)
81.4 percentof pettybourgeois).We thusdecid- EXPERTISE -.789* .45 -.765* .47
ed to treatall employersand petty bourgeoisas (.044) (.067)
nonexpertsthroughoutthis analysis.6
One othervariable,the standard"quasi-inde- Scaled deviance 370.1 139.0
pendence"matrix(QI) used in mobilitystudies, Degrees of freedom 185 185
is includedin all of the models we explore.This
Differencesin BoundaryPermeabilityCoefficients
variabletreatsall off-diagonalcells (i.e., ties in-
PROPERTY- -.672* -.684*
volving friendshipsthat cross any class bound- (.084)
AUTHORITY (.144)
ary) as a single categorybut distinguishesbe-
EXPERTISE- -.470* -.469*
tween each cell on the diagonal (i.e., the eight AUTHORITY (.043) (.086)
ties representingwithin-classfriendships).While -.202
PROPERTY- -.215
our primaryconcernis not within-classfriend- EXPERTISE (.097) (.151)
ships,properspecificationof theboundary-cross-
ing models requiresthat we include a variable *p < .05 *p <.01 **p <.001

capturingwithin-classfriendships.If all class lo- Note: Numbersin parenthesesare standarderrors.


cations behaved the same way, then we could
have createda simple variable(1's for all cells
on the diagonal,0's elsewhere).However,previ- (9 categories), PROPERTY
is the friendships that
ous researchsuggests this to be an implausible cross two levels of the propertyboundary(2 cat-
assumption(Hollingshead1949,Laumann1966, egories), AUTHORITY
is the friendships that cross
Verbrugge1977). two levels of the authorityboundary(2 catego-
Using conventionallog-lineartechniques,we ries), and EXPERTISE
is the friendships that cross
estimateModel 1 by: two levels of the expertboundary(2 categories).
We includecountryinteractionswiththerowand
R + F + N + R*N + F*N + QI + PROPERTY+ columnmarginals(R andF) so thatcountrydif-
AUTHORITY + EXPERTISE, ferences in marginaldistributionsdo not affect
ourresults.
whereR is therespondent'sclass location(8 cat- Ourcentralinterestin Model 1 is in the coeffi-
egories), F is the friend'sclass location(8 cate- cients for the threeboundary-crossing variables
gories),N is nation(U.S., Sweden,Canada,Nor- - the smallerthe coefficient,the less permeable
way),R*N is theinteractionof respondent'sclass the class boundary.Thus,we interpretdifferenc-
withnation,F*N is theinteraction of friend'sclass es between these coefficients as reflecting the
withnation,QI is the quasi-independence matrix differentialpermeabilityof the class boundaries.
6Toseeif thisdecisionthreatened therobustnessof
ouranalysis,we mergedthedatafor thefourcoun- Resultsfor Analysisof ThreeExploitation
triesintoa singlematrixandwerethusableto make Boundaries
thefull set of distinctions
in theexpertisedimension
amongemployersandpettybourgeois.Thedistinc- Table 2 presentsthe core substantiveresultsfor
tiondidnotimprovethegoodness-of-fit of themod- Model 1. All threeboundaries,regardlessof how
els andso we feel reasonably confidentthattreating samplesize is defined,have statisticallysignifi-
thesecategoriesas nonexperts didnot seriouslyun- cant negative coefficients indicatingthat these
dermineourresults. boundariesdo in fact constituteobstaclesto the

This content downloaded from 157.92.4.71 on Thu, 04 Jun 2015 20:18:58 UTC
All use subject to JSTOR Terms and Conditions
PERMEABILITYOF CLASS BOUNDARIES TO CROSS-CLASSFRIENDSHIPS 95

formationof friendships.The negative coeffi- various combinationsof the three exploitation


cientsfor the propertyandexpertboundariesare boundarieswe areconsidering.7
significantlygreaterthanthose for the authority A significantimprovementin fit of Model 2
boundary.The authorityboundaryis thus more over Model 1 impliesthatthe differentialperme-
permeable than the property or the expertise abilityof boundariesbetweenthe workingclass
boundary.Takingtheantilogof thesecoefficients, andthe otherclass locationsis not fully captured
the odds of a friendship across the authority by the additive effects of the three underlying
boundaryarenearly 100 percentgreaterthanthe exploitation variables, i.e., interactioneffects
odds of a friendshipacrossthe propertybound- across specific class locations are also present.
aryand60 percentgreaterthana friendshipacross The WORKERvariable also enables us to con-
the expertiseboundary. structan overall rankingof the permeabilityof
Thecomparisonof thepermeabilityof theprop- the boundariesbetween the working class and
erty and expertiseboundariesis less clear. Cal- the other class locations by adding for each
culatedusing the numberof respondentsas the WorkeriNonworker friendship-tiematrixthecor-
sample size, the propertyboundarycoefficient respondingcoefficients for the PROPERTY,Au-
does not differ significantlyfrom the expertise THORITY, EXPERTISE, andWORKERvariablesas in-
boundarycoefficient. When sample size is the dicatedin Table4. The individualcoefficientsin
full numberof friendshipties, however,the coef- the equationfor Model 2 arelogs of odds ratios.
ficient of the propertyboundaryis significantly The anti-logof the sum of the coefficientscorre-
differentfrom that of the expertise boundary. spondingto eachcell thusgives the relativeodds
Takingthe antilogof these coefficientsindicates of a friendshiptie occurringin thatcell, given the
thatthe odds of friendshipscrossing the exper- constraintsof Model 2.
tise boundaryare about20 percentgreaterthan
the odds of friendshipscrossing the property
Resultsfor Analysisof WorkingClass
boundary.
BoundaryPermeability

TheRelativePermeabilityof Boundaries Table 3 gives estimates of the coefficients for


BetweenWorkingClass and NonworkingClass Model 2 for the threeadditiveexploitationvari-
Locations ables and the seven cross-class friendshipties
involvingthe workingclass. Thegoodness-of-fit
Model 1 indicates the relative permeabilityof
boundarieswithinthe threeexploitationdimen- 'One technicalcomplicationwith theWORKERvari-
sions - property,authority,and expertise.An ableshouldbe noted.Becausethe quasi-independence
matrixperfectlyidentifiesevery cell on the diagonal
alternativeanalyticalstrategyis to examine di- of the friendshipties matrixwhen the quasi-indepen-
rectlythe relativepermeabilityof the boundaries dence variableis includedin Model 2, it is then im-
between class locations of the class structure. possibleto estimateall seven of the coefficientsin the
Ratherthanconsideringall 28 boundariesin the Workervariable.One will be mathematicallyredun-
friendship-tiesmatrix,we focus on rankingthe dant because it is a linear combinationof the other
permeabilityof class boundariesbetween the Workercoefficients and the workerdiagonalcell in
workingclass (the categorythatis exploitedon the quasi-independencematrix.Because we are not
all three dimensions)and other class locations, specifically interested in the coefficients for the
becausethese boundariesareparticularlysalient WorkerlWorker diagonal cell, and because we want
to have an explicit coefficient for every boundary
in Marxisttheory. betweenthe workingclass andotherclass locationsin
Estimatesof theoverallpermeabilityof bound- Model2, we useda modifiedquasi-independence vari-
aries between the working class and the other able in Model 2 in which the WorkerlWorker diago-
class locationsareestimatedfrom Model 2: nal cell (category 8 on the original QI variable in
Figure2) is assignedto level 0 and thus mergedwith
MODEL I+ WORKER,
all of theoff-diagonalcells. This modifiedQI variable
whereWORKERis a set of dummyvariablesiden- has no effect on the goodness-of-fitof Model 2 or on
the values of the additivetermsin thatmodel; it also
tifyingties betweenworkersandeach of the oth- has no effect on the relativemagnitudeof the coeffi-
er seven class locations(see Figure2). If none of cients for the Workervariableand thus had no effect
thecoefficientsfor thesedummyvariablesis sig- on the rankingof the permeabilityof boundariesbe-
nificant, the likelihood of friendshipsbetween tween the working class and other class locations.
workersandotherclasslocationsis simplya func- However, the modified QI variable does make the
tion of the overall additive effects of crossing discussionof the coefficientsmore straightforward.

This content downloaded from 157.92.4.71 on Thu, 04 Jun 2015 20:18:58 UTC
All use subject to JSTOR Terms and Conditions
96 AMERICANSOCIOLOGICALREVIEW

Table 3. Coefficients for WORKERINONWORKER Boundary boundaryis more permeablerelative to other


Permeability: Adults in the Labor Force in Four boundariesthan it is for the class structureas a
Countries, 1980-1982
whole.
Sample Size Defined as The results in Table 3 point to an important
Number of Number of
conceptualimplicationfor class structureanaly-
Friendship Ties Respondents sis. The presenceof significanttie-specificinter-
actionssupportsthe view thatthe class structure
Variable Coefficient Coefficient
cannot be understoodsimply as a sum of the
PROPERTY -.954* -.960* threeunderlyingexploitationdimensions.If there
(.105) (.163) were no tie-specific interactions,then the con-
AUTHORITY -.399* -.341' ceptof "classstructure," formedthroughthecom-
(.057) (.087) binationof the three "primitiveterms,"would
EXPERTISE -.715* -.666* simplybe a heuristicconvenience.Nothingwould
(.063) (.097) be lost by simply talking serially about the ef-
Tiesa
Worker/Nonwor-ker fects of propertyownership,the effects of exper-
-.167
(1) WORKERIEMPLOYER -.202 tise, andthe effectsof authority,andignoringthe
(.077) (.118) effects of specificlocationsin the class structure.
(2) WORKERIPETTY -.301 -.261 "Location"gets its analyticalbite from the syn-
BOURGEOIS (.072) (.112) ergetic consequencesof the specific combina-
-. 182*
(3) WORKERIMANAGER- -.228' tions of dimensionsthat generatea given loca-
EXPERT (.071) (.109) tion. To use a cliche, "thewhole is greaterthan
(4) WORKERIMANAGER .058 -.061 the sum of the parts"8and the presence of tie-
(.082) (.126) specific interactionscapturesthis.
.060
(5) WORKERISUPERVISOR .067 The additiveandinteractivetermsin Model 2
(.056) (.086) generatethe rankingof the likelihoodof the var-
(6) WORKERI -.356" -0.381
ious WorkerlNonworker friendshipspresentedin
PROFESSIONAL (.106) (.164) Table 4. Of the seven boundaries, the Wor-
(7) WORKERISEMI- -.308" -0.332*
kerlEmployerboundaryis the least permeable,
PROFESSIONAL (.057) (.089) whiletheWorkerlSupervisor boundaryis themost
permeable:Workersare nearlyfive times more
Scaled deviance 337.2 126.11 likely to have a friendshipwith a supervisorthan
Degrees of freedom 179 179 with an employer.The secondandthirdranksin
impermeabilityare the WorkerlManager-Expert
Improvement of scaled 32.0'** 12.89'
andWorkerlProfessional boundaries.As defined
deviance over Model 1
6 6
in Model2, bothof theseties crosstwo "levels"of
Degrees of freedom
the expertisedimension.The next threebound-
*p <.05 **p <.01 *p <.001 aries- WorkerlManager, WorkerlSemi-Profes-
a
These coefficients indicate the difference between a sional, and WorkerlPettyBourgeois - are of
given category of WorkerlNonworker ties and all friend- roughly equal permeability.In the structureof
ships that do not cross the worker boundary (i.e., Model2, thelasttwo of thesecrossonly one "lev-
WorkerlWorker and NonworkerlNonworker ties)
el" of theirrespectiveexploitationdimension
Note: Numbers in parentheses are standard errors.
WorkerlSemi-Professional friendshipscross one
level on expertiseand no levels on propertyor
of Model 2 is a significantimprovementover authority;WorkerlPettyBourgeois crosses one
Model 1 indicatingsignificantfriendship-tie-spe- level on propertyand no levels on authorityor
cific interactionsbetweenthe workingclass and expertise. Thus, WorkerlManagerfriendships,
other class locations. With the exception of whichcrosstwolevels of theauthoritydimension
WorkerlManager andWorkerlSupervisor friend- (butcross no levels on the propertyor expertise
ships,friendshipsbetweenworkersandnonwork- dimensions),areaboutas likelyasfriendshipsthat
ers aresignificantlyless likelythanwouldbe pre- cross only one level of the expertiseor property
dicted by the additivemodel. Furthermore,the
coefficients for the WorkerlManager and IThesamepointcanbe madeabouttheclaimthat
WorkerlSupervisor asa wholehaseffectsthatcannot
boundariesare both signifi- the"classstructure"
cantly greaterthan those for any other catego- be reducedto theseparateadditiveeffectsof thedi-
ries, indicating that for workers the authority mensionsthatconstitute
it.

This content downloaded from 157.92.4.71 on Thu, 04 Jun 2015 20:18:58 UTC
All use subject to JSTOR Terms and Conditions
PERMEABILITYOF CLASS BOUNDARIESTO CROSS-CLASSFRIENDSHIPS 97

Table 4. Rank Order of Boundary Impermeability for Friendship Ties Between Workers and Other Class Locations

Odds of
Friend-
Antilog ship Ties
of Sum Relative
Coefficients Added to of Coeffi- to
FriendshipTie EstimateOverallPermeability Sum of Coefficients cients Employers

WORKERIEMPLOYER PROPERTY + AUTHORITY + WORKER (1) -.954 -.399 -.167 = -1.520 .219 1.00

WORKERIMANAGER- AUTHORITY + EXPERT + WORKER (3) -.399 -.715 -.182 = -1.296 .274 1.25
EXPERT

WORKERIPROFESSIONAL EXPERT + WORKER (6) -.715 -.356 = -1.071 .343 1.57

WORKERIMANAGER AUTHORITY + WORKER (3) -.399 + .058 = -.341 .711 3.24

WORKERISEMI- WORKER (7) -.308 = -.308 .735 3.36


PROFESSIONAL

WORKERIPETTY WORKER (2) -.301 = -.301 .740 3.38


BOURGEOIS

WORKERISUPERVISOR WORKER (4) .060 = +.060 1.062 4.85

Note: Sample size is number of friendship ties.

boundary.Thisaffirmstherelativelyhighperme- A statisticallysignificantimprovementof Model


abilityof the authorityboundary.The resultsfor 3b over Model3a suggeststherearecountrydif-
theWorkerlPetty Bourgeoisboundaryindicatethat ferencesin permeability.It is possible,however,
the salientissue for the propertyboundaryis not thatalthoughthereareno globalcountryinterac-
self-employmentas such, but capitalistproperty tions, there may be significantdifferences be-
relations.The odds of friendshipties between tween specificpairsof countriesfor specific co-
workersandpettybourgeoisareover threetimes efficients in Model 1. We thereforealso look at
greaterthanthosebetweenworkersandemploy- the differencesin the coefficientsfor eachpairof
ers.9 countries.
We areinterestedin countrydifferencesin per-
meabilityfor two reasons.First, in an explora-
CountryInteractions
tory analysisit is useful to know how generaliz-
Models 1 and 2 assumethatthe relativeperme- able the findings are. If the patternsare funda-
abilitiesof boundariesarethe samein eachof the mentally the same in various countries with
fourcountries.We test for the presenceof inter- roughlysimilarclass structures,thissuggeststhat
countrydifferences'0in thecoefficientsin Model the patternsaregenerateddirectlyby underlying
1 by comparingthe goodness-of-fitof the fol- class processesratherthanby contextuallyvari-
lowing two models: able historical,cultural,or politicalfactors.Sec-
ond, part of the motivationfor studying class
Model 3a = Model 1 + QI*N boundarypermeabilityis thatpermeabilitymay
Model 3b = Model 1 + QI*N + PROPERTY*N have a bearingon class formation,class allianc-
es, andclass struggle.Knowingwhetherperme-
+ AUTHORITY*N+ EXPERTISE*N
abilityvaries significantlyacrosscountrieswith
differentorganizedclass forces,can set the stage
9 Arguably, the relative permeability of the
for future explorationsof the possible linkage
WorkerlPetty Bourgeoisboundaryhas historicallyfa- between class boundarypermeabilityand class
cilitatedvariousformsof populism,which frequently
formation.
involve political and ideological alliances between
workersand small propertyowners, e.g., in the U.S.
in the late nineteenthcentury. in-class friendshippatterns.Nevertheless,the QI*N
10Because we are concernedwith class boundary interactionsmustbe includedin the equationstesting
permeability,we do not explicitlyexaminethe QI*N country differences in the three boundary-crossing
interactions,i.e., differencesacrosscountriesin with- coefficients.

This content downloaded from 157.92.4.71 on Thu, 04 Jun 2015 20:18:58 UTC
All use subject to JSTOR Terms and Conditions
98 AMERICANSOCIOLOGICALREVIEW

ResultsforCountryInteractions Table 5. Coefficient for Permeability of Exploitation


Boundaries,by Country:Adults in the Labor
Force, 1980-1982
Resultsin Table5 indicatethatthereis littleevi-
dencethatourcentralconclusionsfromModel 1 U.S. Sweden Norway Canada
varysignificantlyacrossthefourcountriesin our
sample.In the globaltestfor countryinteractions PROPERTY -1.11 -.67 -1.03 -1.06
in Model 3b, thereis no statisticallysignificant AUTHORITY -.28 -.30 -.35 -.33
improvementin fit over Model 3a, even when EXPERTISE -.72 -.81 -.80 -.89
we use the numberof friendshipties as the sam-
ple size when calculatingstandarderrors."Fur- Note: Samplesize is numberof friendshipties. Improve-
ment of scaled deviance in Model 3b over 3a is 4.8 with 9
thenrmore, a pairwiseexaminationof all possible d.f. (n.s.). No coefficients differ significantlyacrosscoun-
differencesacrosscountriesin boundarycoeffi- tries.
cients (a total of 18 pairwisetests), revealedno
significantdifferences.
However,in spiteof the absenceof statistical- are built exclusively aroundthe authorityaxis.
ly significantdifferences,there is an apparent Not only is authoritythe most permeableof the
differencein thepatternsof resultsfromthecoun- threeboundariesin relativeterms,it is also quite
try-levelanalysis.Thebasicrankorderof bound- permeablein absoluteterms.
ary impermeability- PROPERTY first, then Ex- Whataboutthethreeothertheoreticalperspec-
PERTISE and AUTHORITY - holds for the U.S., tives? Marxisttheorypredictsthat the property
Norway, and Canada.In Sweden, however,the boundaryshouldbe the most impermeable;this
expertise boundaryis less permeablethan the predictionis supportedby our analysis.The re-
propertyboundary(thecoefficientsare-.81 and- sults for the expertiseand authorityboundaries,
.67 respectively),althoughthedifferencebetween however, are not entirely what most Marxists
these coefficients is not statisticallysignificant. wouldexpect:Theexpertiseboundaryis less per-
Witha largersamplesize, the Swedishpatternof meableand the authorityboundaryis moreper-
cross-classfriendshipsmay be shown to be dif- meablethanwouldbe expectedon the basis of a
ferent from the other three countries.With the theory of exploitation and common interests.
presentdata, however, none of the differences While the Marxistconcernwithexploitationand
betweenSwedenandtheothercountriesis statis- class interestsmaybe consistentwiththefinding
tically significant. thattheexpertiseboundaryis less permeablethan
On the basis of the presentanalysis,therefore, the authorityboundary,Marxistclass analysis
we cannot reject the hypothesisthat, for these wouldnot expectthe relativemagnitudeof these
fourcountries,thepatternsof classboundaryper- two permeabilitycoefficients to be so sharply
meabilityto theformationof friendshipsarenear- different.
ly invariant. Thefindingsfortheexpertandauthoritybound-
aries, therefore,seem more consistentwith the
class habitusand opportunitystructureperspec-
CONCLUSION
tives. On theone hand,the relativelyhighimper-
Ouranalysissupportstwo mainconclusions.First, meabilityof the expertiseboundaryis consistent
resultsfor the relativeimpermeabilityof the ex- withtheoriesof culturalcapital,even if suchthe-
ploitationboundariesindicatethat the property oriestendto minimizethecontinuingimportance
boundaryis the most impermeable,followed by of propertyas a basis for structuringclass prac-
theexpertiseboundary,withtheauthoritybound- tices. On the otherhand,the high relativeperme-
arybeing the most permeable.This rankorderis abilityof the authorityboundaryis best predict-
mostsharplyinconsistentwithperspectivesbased ed by the opportunitystructureperspectiveon
on class interests,such as Dahrendorf's,which friendships.In many workplacesthere are di-
verse opportunitiesfor informalinteractionbe-
" There is also no improvementin the fit of the tweenworkersandsupervisors,andevenbetween
models if we dropthe QI*NtermsfromModel 3b and
workersand managers.This density of interac-
compareit directly with Model 1. However, Model
3a is a significantimprovementin fit over Model 1, tionalpossibilities,combinedwithrelativelyhigh
indicatingthattherearesignificantQI*N interactions. levels of careermobilityacrossauthoritybound-
While class boundarypermeabilitydoes not varysig- aries comparedto across propertyand expert
nificantlyacross countries,the likelihood of within- boundaries,may accountfor the relativelyhigh
class friendshipsdoes. permeabilityof the authorityboundary.

This content downloaded from 157.92.4.71 on Thu, 04 Jun 2015 20:18:58 UTC
All use subject to JSTOR Terms and Conditions
PERMEABILITYOF CLASS BOUNDARIESTO CROSS-CLASSFRIENDSHIPS 99

Theanalysisthussuggeststhatthecausalmech- the fourcountriesexamined.Thesecountriesare


anismsidentifiedby theoriesof class interests(at all highly developedcapitalistcountries,but in
least the Marxistvariant),class habitus,and op- otherrespectsthey arequitedifferent.The Unit-
portunitystructureprobablyall operateto create ed Statesis ethnicallyand raciallyrelativelydi-
obstaclesandopportunitiesfor friendshipforma- verse; the two Scandinaviancountriesare rela-
tion across class boundaries.The result of the tively homogeneous.Laborunionsare strongin
jointoperationof thesethreeclustersof causesis Scandinavia,and in Sweden even include most
thatthe boundaryMarxistswould predictto be managersand professionals;they are notorious-
the least permeableis indeed the least perme- ly weak (anddeclining)in theUnitedStates.The
able. This mightimply thatthe property-exploi- state is a majorcounterweightto the capitalist
tation-interestmechanism is a more powerful organizationof productionin Sweden, employ-
structuringmechanismthanarethe class habitus ing well over40 percentof the laborforce.In the
or opportunitymechanisms.Such a conclusion, U.S. the figureis less than20 percent,with Can-
however, is vulnerable to criticism on two adaandNorwayfallingin between.Sweden,and
counts.12First,claims aboutthe relativepotency to a lesserextentNorway,have long traditionsof
of causalprocessesarealwaysvulnerableto mea- powerfulsocialdemocraticpoliticsin whichclass
surementcriticisms.Our conclusion about the issues are often at the forefrontof politicalcon-
relatively high permeability of the authority sciousness,whereasin Canadaandespeciallythe
boundarymight change if we adopteda more U.S., radicalworking-classpolitics are at best a
restrictivedefinitionof authority,e.g., limiting marginaltradition.
"managers" to high-levelexecutives.Also, if we One mightexpectthis level of diversityin cul-
distinguishedamongprofessionalsbetweenhigh- turalandpoliticalpracticesto be associatedwith
ly credentialedprofessionalsand other profes- significant variationsin patternsof friendship
sionals (e.g., surgeonsvs. engineers),the exper- formationwithinclass structures.Thereis little
tise boundarymight become the "least"perme- evidence to supportsuch an expectation- the
able. While such conjecturedresultscould po- patternof permeabilityof threeboundariesis re-
tentiallybe counteredwitha comparablerespeci- markablysimilarin the fourcountries.This sug-
ficationof theproperty-boundary, thiswouldonly geststhatalthoughthemicrocultural factorstalked
reaffirmthe sensitivityof claims aboutrelative aboutin theoriesof class habitusmaybe relevant
causalpotencyto measurementchoices.13 for understandingfriendshipformation,cross-
Second,even if a morefine-grainedinspection nationalculturalvariationseems largely irrele-
revealedthatourcore resultswere robustacross vant.
alternative specificationsof theseboundaries,
there Initiallywe hoped to deepen the Marxistun-
is stilltheproblemof ascribingthisimpermeabil- derstandingof structurallydefined locations with-
ity to "exploitationinterests"ratherthan class in class structuresthroughan investigationof the
habitusor opportunitystructure.Employerscer- ways in whichpeople's lives areorganizedwith-
tainlylive differentlifestylesfrommostnonprop- in those structures.Classboundarypermeability
erty owners, and the physical opportunitiesfor to friendshipformationis one aspectof thisprob-
informalinteractionbetweenmostemployeesand lem. We assumedthatsuch processeswould be
employersarefew. Therefore,while ourdataare important forunderstanding variationsintheover-
consistentwith the claim thatproperty-basedin- all processby which structurallydefinedclasses
terestshavestrongereffectson friendshipforma- becomecollectivelyorganizedas politicalforces.
tionthaneitheropportunitystructureorclasshab- Ouranalysisundermines thisassumption.Because
itus,theycannoteffectivelyrefutecounterclaims. classboundarypermeability tofriendshipsis large-
The secondmajorconclusionfromour analy- ly invariantacrosscountriesthatdiffer substan-
sis is thatpatternsof friendshipformationacross tially in patternsof organizedclass formation,it
class boundariesare relativelyinvariantacross seems unlikely that this aspect of permeability
will contributemuch to explainingvariationsin
12 Fora generalphilosophicaldiscussionof theprob-
lem of assessingthe relativecausalpotencyof differ- macropolitical outcomes.Theseresultsforfriend-
ent causes in a multicausalprocess, see Wright,Le- shipsarequitesimilarto resultsforcross-national
vine, and Sober (1992). research on occupational mobility, which also do
13In a separateanalysis of employers with more not find dramaticdifferencesacrosscountries."4
than 10 employees (not reportedhere) we found that Of course, the common patterns of class bound-
the likelihoodof these employershaving friendships
withemployeeswas significantlylowerthanforsmall- 14Despite some challenges, the intercountrysimi-
er employers. larityin occupationalmobilityis broadlysupportedin

This content downloaded from 157.92.4.71 on Thu, 04 Jun 2015 20:18:58 UTC
All use subject to JSTOR Terms and Conditions
100 AMERICANSOCIOLOGICALREVIEW

arypermeabilityin developedcapitalistcountries pertisedimensionwas constructedusing only occupational


may still be importantin helping to understand codes. Occupationsin each nationalsample were initially
coded using that country's occupational-classification
the emergenceof commonkindsof class forma- scheme. These occupations were then aggregated into a
tions in thesecountries,such as the tendencyfor common 27-categoryoccupationvariable.For the present
populistmovementsto involvealliancesbetween analysis, "experts"were defined as anyone occupying a
the workingclass and the pettybourgeoisie.But professional,technical,or managerialoccupation.(A man-
agerialoccupationis not the same as a a managerialposition
differentialpermeabilityis unlikely to explain as definedbelow). Among experts,we thendistinguishpro-
variationsin such tendenciesacrosscountries. fessionals from semi-professionalsin a conventionalman-
ner by restrictingprofessionalsto those occupationswhich
ERIK OLIN WRIGHT is C. WrightMills Professor of requirepost-undergraduate educationalcredentials.Individ-
Sociology at the Universityof Wisconsin,Madison. uals in all otheroccupationswere defined as "nonexperts."
His research has focused primarilyon comparative 3. Authority.The authoritydimension of class relations
class analysisandproblemsin rethinkingthefounda- posed a problem because we had more refined data for
tions of contemporaryMarxisttheory.WithAndrew respondentsthanwe hadfor theirfriends.Respondentswere
LevineandElliottSoberhe has recentlypublishedRe- asked, "Whichof the following best describesthe position
constructingMarxism:Essayson Explanationandthe whichyou hold withinthe businessor organizaitonin which
Theoryof History(Verso, 1992). you work?Would it be a managerialposition,a supervisory
position, or a nonmanagementposition?"Those in a non-
DONMOONCHO is a graduatestudentat the University managementposition were classified as "nonmanagers."
of Wisconsin,Madison.He is completinga disserta- To distinguishmanagersfromsupervisors,we used detailed
tion on the labor movementand the state in South informationaboutparticipationin decision makingand su-
Korea and Mexico. pervisoryresponsibilitiesas well as formal position in the
managerialhierarchy.A respondentwas definedas a "man-
ager" if he/she participateddirectly in making workplace
Appendix. Definitions of Categoriesin the Class Structure
policy decisions and was either in a position that was for-
Typology
mally called a "managerial"position (as opposed to a su-
pervisoryposition)or hadthe powerto impose sanctionson
This analysis is based on a survey of the relationalaspects
subordinates.All others who said they were in managerial
of each respondent'sjob, including detailed questions on
or supervisorypositions were classified as "supervisors."
participationin workplacedecisions, formalposition in hi-
erarchies,work autonomy,and supervisorypower. There
were also detailedquestionson propertyownership,occu- The Class Str-uctur-e Typologyfor Friends
pation,andeducationalcredentials.Thus, it was possible to For friends, respondentswere asked, "Does [this friend]
construct a highly "nuanced"map of class locations for occupy a managementor supervisorposition at the place
respondentsin the sample (see Wright 1985, appendix2). where he/she works?"A "No" response was classified as
The datafor the respondents'friends,however, were much "nonmanager."Unfortunately,the data on friends lacked
thinner.To preservesymmetryin the analysisof friendship the informationneeded to distinguish managersfrom su-
patterns,therefore,the criteriaavailablefor the friendswere pervisors. Because the classification of individuals along
generallyadoptedfor respondentsas well. these three dimensions was used to create the eight class
locations for the subsequentfriendship-tiesmatrixand be-
The Class StiuctutreTypologyfor Respondents cause we wanteda symmetrical8 x 8 matrix,we needed a
We have constructedthe class structuretypology around strategyto estimate the numberof managersand supervi-
threedimensions: sors among friends. Since in the friendship-tiemodels we
did not differentiatemanagersfrom supervisorsamong ex-
1. Property (relationshipto the means of production). perts for either respondentsor friends (because of sample
Respondentswere asked whetherthey were self-employed. size constraints),we only had to make these estimates for
If self-employed, they were asked if they had any employ- nonexpertfriends.
ees. Self-employed respondentswith any employees were
The estimationprocedurewe adoptedassumedthat the
defined as employers; self-employed respondentswithout
friendship-tiematrixis roughlysymmetricalaroundthe di-
employees were defined as petty bourgeois; and respon-
agonal.Foreach class locationof friends,we knew the ratio
dents who were not self-employedwere definedas employ-
of nonexpert-managers to nonexpert-supervisors amongre-
ees. The questionsfor friendswere identical.
spondents. Each such ratio was then used for the corre-
2. Expertise.Because informationon the educationalcre- spondingrespondentclass location to divide the undiffer-
dentials of respondents'friends was not available,the ex- entiatednonexpertmanagersandsupervisorsamongfriends
into separatecomponents.That is, the ratio of nonexpert-
the literature.Overcomingthe methodologicalcrude- managers to nonexpert-supervisorsamong friends of re-
spondentworkers,for example,was assumedto be the same
ness of early advocates(Lipsetand Zetterberg1959; as the known ratio of nonexpert-managersto nonexpert-
Erickson, Goldthorpe,and Portocarero1979), con- supervisorsamong respondentswith workerfriends. The
temporaryproponents(Erickson 1982; Featherman, known ratioof managersto supervisorswithin the "Undif-
Jones, and Hauser 1975; Gruskyand Hauser 1984; ferentiated-ManagerlWorker" cell was then applied to the
HauserandGrusky1988)haverefinedanalyticalstrat- "WorkerlUndifferentiated-Manager" cell. In this way the
egies considerably and convincingly defended the full 8 x 8 friendship-tiematrixwas constructed.
propositionthat countriesdo not vary much in pat- The operationalizationof the class structurevariableis
ternsof structuraloccupationalmobility. summarizedin AppendixTable A.

This content downloaded from 157.92.4.71 on Thu, 04 Jun 2015 20:18:58 UTC
All use subject to JSTOR Terms and Conditions
PERMEABILITYOF CLASS BOUNDARIESTO CROSS-CLASSFRIENDSHIPS 101

AppendixTable A. Operationalizationof Class Structure

PropertyDimension AuthorityDimension ExpertiseDimension


Has Managerial
Decision-Making
Responsibilities
Occupies and Is Eitherin a
Managerial FormalManagerial
or Positionor Has
Class Self- Has Supervisory StrongSanctioning
Location Employed Employees Position Powers Over Subordinates Occupation

Employers Yes Yes Any occupation


Petty bourgeois Yes No Any occupation
Manager-experts No Yes Yes Professional,managerial
technicalor semi-professional
Managers No Yes Yes Nonexpert
Supervisors No Yes No Nonexpert
Professionals No No No Professionaloccupations
requiringpost-undergraduate
degree
Semi-professionals No No No Technical,managerialor semi-
professionaloccupationsnot
requiringpost-undergraduate
degrees
Workers No No No Nonexpert

REFERENCES
Bertaux,Daniel. 1977. Destins Personnels et Struc- Journalof Sociology 33:1-34.
ture de Class (PersonalDestiny and Class Struc- . 1984. "Social Class of Men, Women and
ture).Paris:PressesUniversitairede France Families."Sociology 18:500-514
Bordieu,Pierre. 1984. Distinction.Cambridge:Har- Featherman,D. L., F. L. Jones, and R. M. Hauser.
vardUniversityPress. 1975. "Assumptionsof Social Mobility Research
_ . 1985. "The Social Space and the Genesis in the U.S.: The Case of OccupationalStatus."So-
of Groups."Theoryand Society 14:723-44. cial Science Research4:329-60.
_ . 1987. "What Makes a Social Class?: On Feld, Scott L. 1981. "'TheFocused Organizationof
theTheoreticalandPracticalExistenceof Groups." Social Ties." American Journal of Sociology
BerkeleyJournalof Sociology 32:1-17. 86:1015-35.
Burris, Val. 1987. "The Neo-Marxist Synthesis of Giddens,Anthony. 1973. The Class Structureof the
MarxandWeberon Class."Pp.67-90 in TheMarx- AdvancedSocieties. New York:HarperandRow.
WeberDebate, editedby N. Wiley. London:Sage Goldthorpe,John. 1983. "Womenand Class Analy-
Publications. sis: In Defense of the ConventionalView."Sociol-
Carchedi,G. 1977. The Economic Identificationof ogy 17:465-88.
Social Classes. London:Routledge& KeganPaul. . 1984. "Women and Class Analysis: A
Curtis,J. 1963."DifferentialAssociationandtheStrat- Reply to the Replies."Sociology 18:491-99.
ificationof the UrbanCommunity."Social Forces Granovetter,MarkS. 1973. "The Strengthof Weak
42:68-77. Ties." AmericanJournal of Sociology 78:1360-
Dahrendorf,Ralf. 1959. Class and Class Conflictin 80.
IndustrialSociety. London:Routledgeand Kegan Grusky,DavidB. andRobertM. Hauser.1984."Com-
Paul. parativeSocial MobilityRevisited:Modelsof Con-
Erickson,Robert,John H. Golthorpe,and L. Porto- vergenceandDivergencein 16 Countries."Ameri-
carero.1979. "Intergenerational Class Mobility in can Sociological Review49:19-38.
ThreeWesternEuropeanSocieties:England,France Hauser, Robert M. 1981. "The Structureof Social
andSweden."BritishJournalof Sociology30:415- Relationships:Cross-classificationsof Mobility,
441. Kinship,andFriendship."IHS-Journal5:1-5 1.
Erickson,Robert.1982. "SocialFluidityin Industrial Hauser,RobertM. andDavidB. Grusky.1988."Cross-
Nations: England, France and Sweden." British national Variationin OccupationalDistributions,

This content downloaded from 157.92.4.71 on Thu, 04 Jun 2015 20:18:58 UTC
All use subject to JSTOR Terms and Conditions
102 AMERICANSOCIOLOGICALREVIEW

Relative Mobility Chances,and Intergenerational nal of Personalityand Social Psychology 32:205-


Shiftsin OccupationalDistributions."American So- 213.
ciological Review53:723-41. Parkin,Frank.1974. "Strategiesof Social Closurein
Holland,Paul W. and SamuelLeinhardt.1973. "The Class Formation."Pp. 1-18 in TheSocial Analysis
StructuralImplicationsof MeasurementErrorin of the Class Structure,edited by F. Parkin.Lon-
Sociometry."Journal of MathematicalSociology don:Tavistock.
3:85-111. . 1979. Marxism and Class Theory: A Bour-
Hollingshead,August B. 1949. Elmstown's Youth. geoisie Critique.New York:ColombiaUniversity
New York:JohnWiley. Press
Huckfeldt,R. R. 1983. "SocialContexts,Social Net- Poulantzas,Nicos. 1975. Classes in Contemporary
works,and UrbanNeighborhoods:Environmental Capitalism.London:New Left Books.
Constraintson FriendshipChoice."AmericanJour- Resnick,S. andR. Wolfe. 1988.KnowledgeandClass.
nal of Sociology 89:651-69. Chicago:Universityof ChicagoPress.
Jackson,R. M. 1977. "SocialStructureandProcessin Szymanski,Albert. 1983. Class Structure:A Critical
FriendshipChoice." Pp. 59-78 in Networks and Perspective.New York:Praeger.
Places: Social Relationsin the UrbanSetting,edit- Verbrugge, L. M. 1977. "The Structureof Adult
ed by C. S. Fischer,R. M. Jackson,C. A. Stueve, FriendshipChoices."Social Forces 56:576-97.
K. Gerson,L. M. Jones, and M. Baldassare.New . 1979. "Multiplexity in Adult Friendships."
York:Free Press. Social Forces 57:1286-1309.
Laumann,Edward0. 1966.Prestige andAssociation Wright,ErikOlin. 1978. Class, Crisis and the State.
in an UrbanCommunity.Indianapolis:Bobbs-Mer- London:New Left Books.
rill Company. . 1985. Classes. London: Verso.
_ . 1973. Bonds of Pluralism. The Form and . 1989a. "Women in the Class Structure."
Substanceof UrbanSocial Networks.New York: Politics and Society 17:35-66.
JohnWiley and Sons. . 1989b. "The Comparative Project on Class
Lipset,S. M. andH. Zetterberg.1959. "SocialMobil- StructureandClassConsciousness:An Overview."
ity in IndustrialSocieties." Pp. 11-75 in Social Acta Sociologica 32:3-22.
Mobilityin IndustrialSociety,editedby S. M. Lip- . 1990. The Debate on Classes. Lon-
set andR. H. Bendix. Berkeley:Universityof Cal-don: Verso.
ifornia. Wright,ErikOlin, AndrewLevine, andElliottSober.
Nahemow, L. and M. P. Lawton. 1975. "Similarity 1992. ReconstructingMarxism.Essays on Expla-
and Propinquityin FriendshipFormation."Jour- nationand the Theoryof History.London:Verso.

This content downloaded from 157.92.4.71 on Thu, 04 Jun 2015 20:18:58 UTC
All use subject to JSTOR Terms and Conditions

You might also like