Sei sulla pagina 1di 28

Temporality and Class Analysis: A Comparative Study of the Effects of Class Trajectory and

Class Structure on Class Consciousness in Sweden and the United States


Author(s): Erik Olin Wright and Kwang-Yeong Shin
Source: Sociological Theory, Vol. 6, No. 1 (Spring, 1988), pp. 58-84
Published by: American Sociological Association
Stable URL: http://www.jstor.org/stable/201914
Accessed: 25/10/2008 00:52

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at
http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at
http://www.jstor.org/action/showPublisher?publisherCode=asa.

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.

JSTOR is a not-for-profit organization founded in 1995 to build trusted digital archives for scholarship. We work with the
scholarly community to preserve their work and the materials they rely upon, and to build a common research platform that
promotes the discovery and use of these resources. For more information about JSTOR, please contact support@jstor.org.

American Sociological Association and American Sociological Association are collaborating with JSTOR to
digitize, preserve and extend access to Sociological Theory.

http://www.jstor.org
TEMPORALITY AND CLASS ANALYSIS:
A COMPARATIVE STUDY OF THE EFFECTS OF
CLASS TRAJECTORY AND CLASS STRUCTURE ON CLASS
CONSCIOUSNESS IN SWEDEN AND THE UNITED STATES1
ERIK OLIN WRIGHT AND KWANG-YEONG SHIN
Universityof Wisconsin,Madison, Wisconsin

Some of the importantconceptual debates between different approaches to class analysis


can be interpretedas reflecting different ways of linking temporalityto class structure. In
particular, processualconcepts of class can be viewed as linking class to the past whereas
structuralconcepts link class to thefuture. This contrast in the temporalityof class concepts
in turn is groundedin distinctintuitionsabout why class is explanatoryof social conflict and
social change. Processural approaches to class see its explanatorypower as derivingfrom
the way meanings and identities are linked to class via a history of experiences; structural
approaches, in contrast, emphasizethe linkage betweenclass andperceived interestsvia the
objectivepossibilitiesfacing people in differentclass locations. Thispaper tries to integrate
these two temporalities by exploring the ways in which trajectories of class experience
intersect structures of objective possibility in shaping different dimensions of class
consciousness.

STRUCTURALAND PROCESSUAL Structuralapproachestreatclass as a matrix


CONCEPTSOF CLASS of relationallydefined empty places filled by
individuals. Positions and people are concep-
As has often been noted, the concept of class tually distinguished: individuals are con-
is one of the most contested in sociology. Not ceived as moving into and out of these places,
only are there contending definitions of but the places themselves are treated as
specific classes, but there are wildly varying definable independentlyof such movements
general approachesto class analysis. In some of people. Wright's (1985) elaborationof an
instances these differences simply reflect exploitation-centeredconcept of class rela-
entirely distinct theoretical projects, where tions would be an example of such a
the same word is being used in completely structural approach.2 Processual views, in
different conceptual spaces. But frequently contrast, see classes as constituted by the
there is a broadly common theoretical lived experiences of people. Classes exist
project-explaining social conflict and social only insofar as the biographiesof individuals
change, for example-but differing views are organizedin such a way that they share a
about what kind of class concept is needed in set of experiences over time which define
order for class to have explanatorypower. their lives in class terms. E.P. Thompson's
In this paper, we will investigate two (1968:9) famous definition of class reflects
sharply different general approachesto class such a perspective: "class happens when
analysis-which we will call structuraland some men, as a result of common experiences
processualviews of class-both of which are (inherited or shared), feel and articulatethe
centrally concerned with explaining social identity of their interests as between them-
conflict and social change. While not all class selves and as againstthose whose interestsare
analysts explicitly identify with one or the different from (and usually opposed to)
other of these approaches, nevertheless we theirs". Individuallocation within production
believe that the distinctionbetween these two relations may matter as one of the mecha-
types of class analysis pervades current nisms that generate such a trajectory of
conceptualdebates. experiences, but it is not intrinsically more
importantthan forms of community, family
structure,or culture in constitutingindividual
1 We would like to thank Jerry Marwell and Robert biographiesas classes.
Hauser for helpful comments on the data analysis in an
earlier incarnationof this paper, and Michael Burawoy
2 Other
for convincing us that the earlier version was completely examples would include the work of Pou-
wrongheaded. lantzas (1975), Carchedi(1977), Cohen (1978).

58 Sociological Theory, 1988, Vol. 6 (Spring:58-84)


TEMPORALITYAND CLASS ANALYSIS 59
The differencesin these two approachesare tives facing actors within a class structure
grounded, we believe, in two different defines a range of possible future states that
intuitions about what it is that makes class are available to them. History may matterin
explanatory of social conflict and social so far as it helps to explain why the class
change. Processual approaches see class as structure is the way it is, but it is not
explanatorybecause of the way it determines conceptually constitutive of the concept of
the subjectiveconditionsof conflict, above all class itself. The temporalityof class analysis
the identities and meanings of the various in structuralapproaches,therefore, is primar-
actors engaged in conflict. Structural ap- ily future-looking:class is an embodimentof
proaches, in contrast, see the central explan- possible futures in the present.
atory power of class coming from the way it Given these temporalitiesimplicit in class
determines the objective alternatives facing analysis, it is easy to see why the concept of
different actors, both by determining the class consciousness differs between process-
materialinterestsof individualactors, and by ual and structural approaches. Processual
shaping the various resources actors can approachesare preoccupied, above all, with
collectively deploy in the pursuit of these the problemof identity, structuralapproaches
material interests. To use the language of with the problem of interests. Identity is a
rational actor models, structural theories dimension of consciousness rooted in the
centeron explainingthe feasible set of actions past. An identity is a package of meanings,
that exist independentlyof the motivationsof symbols, and sentimentstied to one's individ-
the actors, processual theories on the prefer- ual biographyand the way that biographyis
ence orderings of actors themselves. In this tied to the history of a particularcommunity
sense, processual theories are rooted in what or social group. Interests,in contrast,refer to
can be called agent-centered theories of anticipations about the future.5 To say that
action, whereas structuralconcepts of class workershave interestsin strugglingfor higher
deploy relations-centeredapproaches. wages is to say that their welfare in the future
Another way of characterizingthis distinc- would be improved if they engaged in such
tion is in terms of temporality in class struggles in the present. To speak about
analysis. Processual approaches to class interestsis always to make a claim about the
revolve aroundthe problemof learning, with relationshipbetween futureand presentstates.
how people come to learn to be class- If certain workers anticipatebeing capitalists
members with the accompanying identities, in the future, then it makes little sense to say
worldviews, life-styles, meanings. At any that they have interestsin higher wages; their
given moment, what we have learned is interests would be linked to their future
explainedby the trajectoryof experiencesthat position in the class structureratherthan their
have accumulatedin our lives.3 This trajec- presentone.
tory is both a questionof individualbiography These various aspects of processual and
and collective history. The institutional and structural approaches to class analysis are
cultural legacies of victories and defeats in summarized in Table 1. Each of these
past class strugglesmatterjust as much as the approachestends to marginalizeor ignore the
accumulationof lived events in an individual central preoccupationsof the other. Thus, as
biography. The temporalityof class analysis Bowles and Gintis (1986:145-150) have
in processualapproaches,therefore,is primar- forcefully pointed out, traditional Marxist
ily backward-looking:class is an embodiment structuralconcepts of class tend to treat the
of the past in the present. process of how class members acquire the
Structuralapproachesto class, in contrast, cultural attributesof class, particularlyclass
revolve around the objective choices facing identities, as relatively unproblematic.Indi-
actors.4 The feasible set of objective alterna- viduals are bearersof roles and more or less
automatically express the central subjective
propertiescorrespondingto those roles (or in
3 By "experience" in this context, we do not mean functionalist accounts, required by those-
"raw experience" unmediatedby existing interpretations
and meanings, but experiencesas they become integrated
into subjectivity. 5
Subjective interests refer to the goals of intentional
4 The distinction between
analysis based on learning action. By "intentional action" we mean, following
and choosing comes from the insightful analysis of Elster (1985:8ff), action in which anticipated future
Bowles and Gintis (1986). states provide partof the explanationof the action itself.
60 SOCIOLOGICALTHEORY
Table 1. Alternative General Approaches to Class processual approaches to class together to
Analysis understandhow the trajectoryof experiences
Processual Structural in the past intersectsthe objective possibilities
for the future in explaining the practices and
Conditionsof conflict
explained by class subjective objective struggles in the present.6
Theoryof action agent-centered relation-centered In this paper we will attempt to link the
Centraldimension of temporalities of processual and structural
class consciousness identity interests approachesto class in an empiricalanalysis of
Temporalityof class past future class consciousness. The model will be built
aroundfour basic propositions:
roles). The objective conditions tied to Proposition 1: Basic Structural Proposi-
structuralpositions are seen as so powerful tion. The subjective awareness of class
and pervasive that they virtually dictate a set interests will be more systematically shaped
of interests to actors in those positions. At by a person's class location than by their
most, therefore,the problemof consciousness- class trajectory.This follows from the central
formationbecomes one of asking why actors argumentsof structuralapproaches to class.
often fail to understandthose interests, rather Class locations determine the objective class
than asking how the finely texturedpatternof interests of actors. All things being equal,
class subjectivitiesare formed. therefore, there will be a tendency for
Processual views tend to marginalize the subjective perceptions of interests to be
problem of objective relations and con- closely tied to incumbency in a class
straints. The subjective interests held by location.7A person's class biography-where
actors are treatedas almost entirely a cultural they came from, what kinds of class experi-
construct. Interests lack any "objective"
basis: the actions of individuals-in-classesare
6
to be explainedabove all by theirtrajectoryof This paperdoes not representthe first attemptto join
these two aspects of class analysis. In a quite different
experiences in the past, not by the objective idiom, Pierre Bourdieu's attemptat developing a theory
possibilities facing them for the future. of class position and class habitus and can be viewed
trying to join structuraland processual aspects of class
(see Bourdieu, 1985). The approachwe are pursuinghere
A MODEL OF TRAJECTORY, differs from Bourdieu's in two principle respects: first,
STRUCTUREAND CONSCIOUSNESS Bourdieuplaces explanatoryemphasis on the concept of
class habituswhereas, if anything, we assign explanatory
One can imagine special cases in which a primacy to structural concepts of class. Secondly,
purely structuralaccount of class or a purely Bourdieu's concept of class is framed in terms of the
social constructionof interest-basedgroups in general;
processualaccount would be adequate.If, for there is no necessarycontent to those interestsat all, and
example, the structuralconstraintson action certainly no restriction of "class" interests to material
were so narrowthat the feasible set of choices exploitation. Our usage of class will follow more
was reduced to a single option, then it really traditional Marxian lines of identifying the concept
would not mattervery much how classes were specifically with exploitationand materialinterests.For a
discussion of Bourdieu's work in these terms, see
culturally constituted, what identities and Brubaker(1985).
preferences characterized actors in class 7 The claim that the
objective interests determinedby
positions. Alternatively, if a class structure class structuresare tied to class locations of incumbents
was characterized by such high levels of is contingent upon the time horizons of movements of
individuals into and out of class locations. At one
mobility and openness that every individual extreme is a class structurein which the incumbentsof
could effectively choose their own class and
given class locations and their childrenstay in those class
thus class membershipwas almost entirely a locations permanently;at the other extreme is a structure
matter of the formation of preferences for in which incumbents have a very high probability of
particular ways of living, then structural being able to move into different locations at will. The
objective interest tied to locations are different in these
approaches to class would be of little two structures even if the description of the "empty
explanatorypower. places" themselves is identical. Interests are always
Real class societies, however, are far from claims about desired states in the future (since they take
either of these special cases, and thus class time to accomplish), and therefore if one's future class
location is likely to be different from one's present
analysis should attempt to combine the location, one's present interests are affected. At a
analytical strategies of processual and struc- minimum, therefore, high rates of individual mobility
tural approaches. In particular, we need to generate some uncertainty or indeterminacy in the
bring the two temporalitiesof structuraland relationshipbetween locations and interests.
TEMPORALITYAND CLASS ANALYSIS 61
ences they have had in the past-should emerge within individuals. Such tendencies,
matterfor individual's subjective understand- however, are mediated by the macro-
ing of class interestsonly to the extent that it historical context within which these micro-
gives incumbentsof class positions a clearer processes take place. Under historical condi-
perceptionof their class interests. tions of class formationin which classes are
Proposition 2: Basic Processual Proposi- highly organized economically and politi-
tion. The subjectiveclass identityof individu- cally, it would be expected that these micro
als will be more systematically shaped by tendencies would be much stronger than
their class trajectory than by their class under conditions in which class formations
location. Identitiesare formedover time; they were highly atomized and fragmented. The
are learned. Whethera given individualhas a "efficiency" with which a particular class
workingclass identity will depend more upon location at the micro-level generates subjec-
their entire biography of class-based experi- tive perceptionsof correspondingclass inter-
ences than simply upon their currentlocation ests, or a given class trajectory generates
within the class structure. particular class identities, therefore, will
Proposition 3: Interaction of processual increase as the collective institutionsof class
and structuraldimensions of class. The class formationare stronger.If we were to compare
identityof people will have an effect on their a country with relatively weak and disorga-
perception of their class interests net of the nized forms of working class formation
effect of their currentlocation within the class (such as the United States) with a country
structure. Class structures define a set of with relatively strong forms of working class
objective interests of incumbents in class formation (such as Sweden), we would
locations and this, in and of itself, creates a therefore expect the correlations posited in
tendency for subjective perceptions of inter- propositions 1-3 to be weaker in the former
ests to vary by class location (Proposition1). than in the latter.
Subjective perceptionsof interests, however, Taking these four propositionstogether we
are never a simple, unmediatedreflection of get the simple causal model presented in
"raw sense data", and, if processualtheorists Figure 1. This model will form the basis of
of class are correct, one of the central the empirical investigation in this paper of
determinantsof how people understandtheir class location, class trajectory and class
interestsis their identity. It is thereforeto be consciousness in Sweden and the United
expected that identity will shape perceptions States.
of interests even after controlling for the
individual's class location.
Proposition 4: Class Formation Proposi- ELABORATIONOF BASIC CONCEPTS
tion. The strength of the relationships at the AND OPERATIONALIZATIONS
micro-levelbetweenclass location and subjec-
tive class interests and between class trajec- Class Structure
tory and class identity will vary with the Different theoretical agendas yield different
degree of the collective formation of classes. kinds of structuralmaps of classes. Over-
Proposition 1-3 are all micro propositionsin whelmingly, sociologists who study class
that each is a claim about how some attribute have taken "occupation" as the central
or state of individuals generates tendencies category for this structuralmap of positions.
for particular forms of consciousness to Of course, there are thousands of distinct

CLASS TRAJECTORY > SUBJECTIVE CLASS IDENTITY

I
CLASS FORMATION >

CLASS LOCATION
J V
> SUBJECTIVE CLASS INTERESTS
Figure 1. Basic Causal Model
62 SOCIOLOGICALTHEORY
occupationsin any society, and thus the initial The basic class typology we will use has
task is one of aggregatingthese occupations four categories:Capitalists,petty bourgeoisie,
into some manageableand theoreticallymean- "middle class" wage earners and workers.
ingful set of researchable categories. Typi- Since a detailed defense of these categories is
cally this is done more or less througha logic available elsewhere (Wright, 1985; chapter
of the relative "status" of the occupations, 3), we will only provide a brief rationale
generatingthe traditionaloccupationalclassi- here.
fication scheme: lower manual, upper man- The most problematic category in this
ual, lower white collar, upper white collar. typology is the third, "middle class wage-
Needless to say, there are many variations earners."Indeed, as has often been noted, the
and nuances on this basic theme, some more problem of the middle class has always been
complex than others, but most empirical rather troublesome within Marxist theory.
studies of class operate along this kind of How can a conceptual framework built
logic. around a polarized concept of relationally
In this paper we will adopt a different defined classes accommodatecategories that
underlyinglogic for the structuralcategories, are in the "middle"?What does it mean to be
one rooted in Marxianconcepts of class rather in the middle of a relation?
than status. The basic distinctionbetween the Many solutions to this problem have been
two conceptual stances is this: status-based proposed in recent years. We will adopt a
concepts distinguish among positions on the solution based on the idea that certain kinds
basis of the culturalevaluationsof the relative of locations can be considered in the
desirabilityof positions, where "desirability" "middle"in so far as they share the relational
is usually an amalgam of a variety of valued properties of more than one class. They
attributes;class-based concepts, on the other occupy what Wrighthas called "contradictory
hand, attemptto distinguish among positions locations within class relations." More specif-
on the basis of the fundamental material ically, in capitalist societies it is possible to
interestsobjectively tied to positions. The two define a range of positions which are
approaches, of course, are not empirically simultaneously exploited by capitalists and
unrelated, since one of the values which yet are exploiters along some secondary
makes positions desirable is precisely the dimension of class relations.8Such positions,
income and power typically associated with for convenience, we shall call the "middle
them. There will therefore be a reasonable class".9
correlation between status typologies of There are two primary mechanisms for
positions and class typologies. Nevertheless, such secondary exploitation: control over
the underlying criteria and rationales are what Wright has called organizationalassets
distinct. and control over credentials (or "inalienable
Depending upon one's general theory of assets"). The first of these defines the
society and the empirical objectives of distinctive form of exploitation linked to
research, status-basedor class-based concepts managerial positions within the social rela-
of structural positions may be the more tions of production,the second to the form of
appropriate.If one is ultimately interestedin exploitationtied to professionaland technical
understanding social conflict and social
change, and one believes that both conflict 8
and change are fundamentallystructuredby The concept of "exploitation" is a complex and
contested concept in contemporaryMarxism. The basic
materialinterests, then some variantof a class idea is that occupying certain kinds of positions within
typology is probablythe most useful. If one is the social relationsof productionenables people to make
primarily interesting in understandingsuch claims on the social surplus though various kinds of
things as self-esteem, and one believes that mechanisms. Because there are multiple mechanisms of
esteem is fundamentally structured by the exploitation tied to different aspects of production
relations, it is possible for people to be exploited by one
normative evaluations of others, than status- mechanismand yet be exploiters by another. See Wright
based typologies of positions may be more (1985:64-92).
relevant. In any case, throughoutthis paper 9 Strictly speaking these positions should be referred
we will adopt a variantof the Marxistconcept by the more cumbersome expression "contradictory
of class for our structuralmap of positions. exploitationlocations within class relations".The use of
the common language term to designate such positions
The specific concept we will use is based should not be interpretedas suggesting a conceptualshift
on the recent work of Wright (1985). towards a gradationalperspectiveon class structure.
TEMPORALITYAND CLASS ANALYSIS 63

positions.10The "middle class" then, will be much of the theoreticalthrustof the analysis
defined as all wage earnerswho eitheroccupy revolves around comparing the effects of
managerial positions within the authority class trajectory and class location on class
structure of their place of work or have consciousness, we decided that the most
professional, technical or managerial oc- appropriate strategy was to measure both
cupations.1 All other wage earners will be variables at comparablelevels of conceptual
considered working class. The working class precision. It is for this reason that authority
thus includes nonsupervisory manual wage position is being measuredby formalposition
earners and routinized white collar and ratherthan substantivepowers.13
service workers. Second, the credential/skillasset dimension
of the class concept is being measured by
Class Structure:operationalization occupation, rather than directly by educa-
tional credential.The rationalefor this is that
The formaloperationalizationof these catego- a person's actual class position should be
ries is presentedin Table 2. The details of the defined by assets that are actively deployed in
operationalizationsused for this analysis are production, not latent assets. Capitalists do
given in Wright (1985: 226, Table 6.13).12 not simply own a pot of money; they own
Several comments on this operationalization capital deployed in capitalist production. A
are needed. First, the criterion we use to person with a PhD driving a taxicab,
define position within authorityhierarchiesis therefore,will not be consideredthe owner of
whether or not the respondent's job is an exploitative asset, even though potentially
formally included in the managerial/supervi- such a person could become an exploiter by
sory hierarchy. In other research by Wright obtaininga "middle class" job.14
on class structure, position within authority Finally, we are aggregating all of the
hierarchieswas measuredby personalinvolve- diverse "contradictorylocations" that would
ment in decisionmaking responsibilities and be formed by combinations of authority
direct control over the activities of others. position and credentialled occupation into a
These more subtle measures, however, were single, ratherheterogeneous "middle class".
not available for the class trajectoryvariable We do this for two reasons. First, because of
in this analysis. We therefore had to choose limitations of sample size it would be
between having parallel constructionsfor the impossible to explore the various types of
class location and class trajectory variables detailedtrajectoriesamongcontradictoryloca-
(but sacrificingsome measurementquality for tions. And second, our central theoretical
the class location variable), c. _,ving asym- interest is with the working class-with the
metricalmeasuresfor the two variables. Since effects of working class locations and work-
ing class trajectorieson class consciousness-
O1It should be noted that this formulationincorporates not with the various distinctions within the
into a Marxistclass concept one of the durableinsights of nonworking class locations in the class
the Weberian tradition of class analysis-that the structure.
ownership of skills can, under appropriateconditions,
constitutethe basis for a class division between workers
and the "middle class." For a discussion of the
13 None of the substantive
relationshipbetween the traditionalWeberianformulation conclusions we make are
of this issue and the one adopted here, see Wright affected by the use of this somewhatweakercriterionfor
(1985:106-108). authority.
11It is
important to note the distinction between 14 We are
opting, in effect, for what could be termeda
occupying a managerial position in the authority behavioral criterion for class membership-what a
hierarchy and having a managerial occupation. While person actually does-rather than a modal definition-
most people in managerial occupations also occupy what a person could do if they optimized their material
managerial positions (about 85% of occupationally interests. Jon Elster (1985:319ff) has argued in favor of
defined managers are also positionally defined as modal definitions, saying that it makes no sense to treata
managers in the United States), the reverse is not true: Rockefeller who freely chooses to work in a factory as
most people in managerialpositions are not in occupa- thereby in the working class. Class membership is
tions that are called "manager". In any case, in the defined by constraints,not acts of will (unless, of course,
present analysis since both of these categories are being a particular act of will is irreversible and therefore
fused into a general "middle class" category, the imposes appropriateconstraints). The operationalprob-
distinctionwill not matterfor our empiricalinvestigation. lems in modal definitionsof class, however, make it very
12 The variable we will use collapses categories 3, 4 difficult to use in empirical research. In any case, for
and 5 from table 6.13 in Wright (1985:226) into the practicalpurposes, the behavioralclass and modal class
middle class. of most people are probablythe same.
64 SOCIOLOGICALTHEORY
Table 2. OperationalCriteriafor Classes

Occupies a managerialor supervisory


authorityposition and/or
occupationis classified as manager,
Self- Has professionalor technician
CLASS CATEGORY employed employees (1970 Census categories)a
CAPITALISTS YES YES
PETTY BOURGEOISIE YES NO
"MIDDLE"CLASS NO NO YES
WORKINGCLASS NO NO NO
a
Occupying a managerial or supervisory authority position is operationalizedby responses to the following
question: "Which of the following best describes the position which you hold within your business or organization?
Would it be a managerialposition, a supervisoryposition or a nonmanagementposition?".

Class Trajectory ture and experience as unproblematic, it


could be argued, is precisely what processual
Ideally, to measurefully class trajectorieswe theorists object to in structuralapproaches.
would need complete class biographies of
Second, mobility trajectorytaps a relatively
individuals, biographieswhich would include limited aspect of overall class trajectory-the
data on the class characterof such things as:
the individual's family of origin, the various link between class origins and destinations-
communities the individual has lived in, the rather than the broad inventory of class
class characterof the individual's schooling constituting processes discussed by process-
experiences, all previous jobs and the length ual theorists. Even aside from the problemof
of time within each and a range of other the relationshipbetween a structuralbiogra-
class-pertinent experiences (unemployment, phy and a trajectory of experiences, a
strikes, etc.). This complex, multidimen- mobility trajectorymay be too thin a basis for
sional space of class experiences over time tappingprocessualapproachesto class analy-
could then be collapsed into a smallernumber sis.
of theoreticallycoherent types of lives (lives These are pertinent objections and they
completely embedded in the working class; should be kept in mind as we explore the
lives which move back and forth between the results of the empirical investigation. Never-
working class and the petty bourgeoisie;lives theless, we feel that class mobility trajectories
with increasing distance from the working
are a sufficiently salient aspect of overall
class; etc.).
We do not have this kind of comprehensive class trajectoriesthat they are an appropriate
data on the class dimensions of biographies preliminaryway of empirically mapping this
and experiences. What we do have are data conceptual space. And we believe that the
on class origins and currentclass locations. experienceslinked to mobility are sufficiently
These data enable us to create a simple class consistent that it is reasonableto interpretthe
mobility matrix. Our strategy,then, will be to empirical effects of mobility patterns as
use the cells in this origins-destinationsclass reflecting the kinds of experiential mecha-
mobility matrix as a first approximationmap nisms postulated within processual views of
of possible class trajectories. class. In any case, with the presentdata these
Advocates of processual theories of class are the best measures of class trajectorywe
may object thatthis is an inappropriateway of have.
specifying a processual view of trajectories Using a mobility matrix to measure class
for two reasons. First it does not directly
measure "lived experiences" as such, but trajectoryin this way, it should be noted, is
quite different from the use of such matrices
simply the structural context within which in the conventional study of "mobility
such experiences occur. In order to treat a
class mobility trajectoryas a variablereflect- effects". In mobility effects research the
ing a processual approachto class analysis, typical analytical strategy is to begin by
we have to assume that particularbiographi- examining the additive model in which social
cal movements throughthe class structureare origins and destinationsare used to predictthe
closely tied to a trajectory of actual class outcome in question. Only after this model
experiences. Treatingthe link between struc- has been specified are the effects of "mo-
TEMPORALITYAND CLASS ANALYSIS 65

bility" as such, understoodas the interactions Table 3. Possible Biographical Trajectories in Class
of origins and destinations, included in the Mobility Matrix
analysis.15 In the conventional wisdom, CLASS DESTINATION
origins and destinationsare treatedas having CLASS Petty Middle Working
a logical priority over mobility patterns(i.e. ORIGINS CapitalistsBourgeois Class Class
the patterns which link origins and destina-
Capitalists 1 2 3 4
tions) in the analysis. Indeed, mobility effects Petty Bourgeoisie 5 6 7 8
are treatedas a kind of residual-that part of Middle Class 9 10 11 12
the consequences of origins and destinations WorkingClass 13 14 15 16
that cannot be attributedeither to origins or
destinations as such. For the purposes of
defining class trajectories in our analysis, numbers has no inherent significance--they
however, origins and destinationsare simply designate qualitatively distinct types, not an
our best available way of distinguishing ordinal ranking. Where all the cells of the
different types of lives in class terms; the table are assigned only two numbers as in
additive model of origins and destinationsis models C and D, this implies that the original
not analytically privileged over an "inter- 16 value typological variable is being col-
active" model. lapsed into a single dichotomousvariable.
To illustratethis logic, consider the simple Models A and B in Table 4 representthe
mobility matrix in Table 3. Sixteen different simple class origins model and currentclass
kinds of biographicaltrajectoriesare defined location (or "destinations")model. By col-
in this matrix. Rather that treating these as lapsing the full typology along the rows in
interactions between two variables-origins Model A we say, in effect, that in class
and destinations-the cells in this table can be biographiesall that mattersis origins; for any
viewed as values in a single 16-category given origin, all destinations are equivalent.
typological variableof class trajectories. Model B, on the other hand, collapses the
If we had sufficient data we could, of table along the columns, implyingthat all that
course, deploy this entire 16-categorytypol- mattersis where you end up.
ogy in the form of 15 dummy variablesin an Models C through G represent different
analysis of trajectoryeffects. This would not, kinds of class trajectorymodels. C and D are
however, in general be the most interesting the simplest, each distinguishing only two
strategy theoretically, since not all of the kinds of lives. In Model C all people who
distinctions among these cells are of equal have lived their entire lives in what could be
theoreticalinterest. An alternative,therefore, termed the "privileged classes"-capitalists
is to collapse the cells of this typology in a among owning classes and the "middleclass"
variety of different ways depending upon among wage-earning classes-are distin-
one's theoretical and empirical objectives. guished from everyone else; in Model D,
Some of the possibilities are illustrated in people who have lived theirentire lives in the
Table 4. "popularclasses"-the petty bourgeoisie and
These models should be read as follows. the working class-are distinguished from
Cells which are assigned the same numberare people whose lives have touched privilege at
treated as equivalent. The ordering of the some time. Models E and F, then, combine
Models C and D in different ways: Model E
15 Good distinguishes a life of privilege (1), a life
examples of this kind of logic would include
Duncan (1966), Knocke (1973), Jackman(1972). Much entirelyin popularclasses (3) and a mixed life
researchon the effects of mobility, particularlyresearch (2); Model F differentiatesthis mixed cate-
before the late 1960s, did not consistently distinguish gory into an "upward"(popularto privileged)
between the additive effects of origins and destinations and "downward" (privileged to popular)
on an outcome of interestand interactiveeffects as such.
Thus, for example, the observationthat mobile individu- component. Finally, Model G collapses the
als have ideologies intermediarybetween the ideologies various "immobile"categories-the diagonal
of immobile individualsof lower and higher statuses has of the original mobility matrix-into a single
been taken to indicate that mobility matters (e.g., category. The result is a conventional up-
Thompson, 1971), whereas in fact this is the result one ward, downward, stable classification of
would expect if origins and destinations simply affect
ideology in an additive manner.See Knocke (1973) for a mobility experiences. Each of these models,
clear exposition of the differentlogics of the additiveand therefore, can be viewed as a theoretical
interactivemodels. hypothesis about what kinds of biographies
66 SOCIOLOGICALTHEORY
Table 4. Examples Aggregationsof Full BiographicalTrajectoryTypology

C = Capitalists;P = Petty Bourgeoisie; M = Middle Class wage earners;W = Workingclass


A. Origins Model B. Currentlocation model
DESTINATIONS DESTINATIONS
C P M W C P M W
C 1 1 1 1 C 1 2 3 4
ORIGINS P 2 2 2 2 ORIGINS P 1 2 3 4
M 3 3 3 3 M 1 2 3 4
W 4 4 4 4 W 1 2 3 4

C. Privilege Model D. UnderprivilegedModel


DESTINATIONS DESTINATIONS
C P M W C P M W
C 1 2 1 2 C 1 1 1 1
ORIGINS P 2 2 2 2 ORIGINS P 1 2 1 2
M 1 2 1 2 M 1 1 1 1
W 2 2 2 2 W 1 2 1 2
1 = entire life privileged owner (C) or 1 = trajectorytouches privileged
privileged wage-earner(M) (C or M)
2 = trajectorytouches popularclasses (P, W) 2 = entire life i n popularclasses (P or W)

E. Privilege, Underprivilege,Mover Model F. Privilege, Underprivilege,Upward, DownwardModel


DESTINATIONS DESTINATIONS
C P M W C P M W
C 1 2 1 2 C 1 3 1 3
ORIGINS P 2 3 2 3 ORIGINS P 2 4 2 4
M 2 1 2 M 1 3 1 3
W 2 3 2 3 W 2 4 2 4
1 = always in privileged classes 1 = always in privileged classes
2 = mixed trajectory 2 = mobile into privilege
3 = always in popularclasses 3 = mobile into popularclasses
4 = always in popularclasses

G. Upward, Downward Model


DESTINATIONS
C P M W
C 1 3 3 3
ORIGINS P 2 1 2 3 I = immobile
M 2 3 1 3 2 = upward
W 2 2 2 1 3 = downward
Note: these models assume that movement between middle class wage earners and the petty bourgeoisie is
downwardmobility.

should be treatedas essentially equivalentand both trajectoryand location simultaneouslyin


what kinds should be distinguished. a multivariateequation.16 When we compare
It is importantto note that the concepts of the explanatory power of trajectory and
class trajectoryand currentclass location are location in propositions1 and 2, therefore,we
not logically independent: class location will not examine their relative explanatory
provides one of the pieces of data used to power within a multivariateequationcontain-
construct the class trajectory variable. The
contrast between "Trajectory" and "Lo-
cation" is therefore not equivalent to the 16
This is not a problem of multicollinearity, but of
distinction between the "past" and the logical construction. Imagine that you created a four
category race by sex typology. It would be mathemati-
"present"; trajectory is meant to capture a cally senseless to include this typology in a regression(in
persons entire biography, which includes the the form of three dummy variables) along with sex or
present location. In the data analysis, there- race. The categories in the trajectorymodels are like the
fore, it does not make much sense to include categories in the race-sex typology.
TEMPORALITYAND CLASS ANALYSIS 67

ing both variables, but their individual about working class consciousness is to make
explanatorypowers in separateequations. the claim that many of the ideas, beliefs,
values and perceptions that comprise con-
sciousness may have a distinctively working
Class Trajectory:operationalization class characterto them.18
The class origins variable in the models in We will examine two aspects of the class
Table 4 are operationalizedthrougha series of content of consciousness: subjective class
questions about the work of the person "who identity, and attitudestowardsthe interestsof
provided most of the financial support for workers and capitalists. As we have argued,
your family while you were growing up". these two dimensions of consciousness have
These questions were not pegged to a specific very differenttemporalstructuresto them and
age of the respondent,but to the typical work thus are likely to be relatedto the problemof
of the head of householdwhile the respondent biographicaltrajectoriesand class locations in
was growing up. The actual criteria are the quite differentways.
same as for the respondent's own class Class identity refers to the ways in which
location in Table 2. people consider themselves "members" of
differentclasses.19As such, it constitutesone
of many ways in which people define what is
Class Consciousness salient about their lives and what differenti-
Class consciousness is at least as contested a ates them from others. "Identity"has both a
concept as class structure.Some Marxists, in cognitive and affective component. Cogniti-
fact, have argued that the concept does not vely, identity simply defines the ways people
even refer to the empirical attitudesof actual place themselves into different systems of
workers,but ratherto the latentconsciousness formal classification. Affectively, identity
implicit in patterns of collective action.17If refers to the kinds of classifications that are
one adopts this stance towards the concept, subjectively salient in a person's system of
then it is hardly appropriateto describe the meanings. This affective aspect of "identity"
empirical objectives of this paper as the is formed throughthe individual's biography
investigation of the effects of class structure of experiences in communities and social
and class trajectoryon class consciousness. interactionswithin which particularidentities
We will adoptwhat can be termeda "thin" are culturally salient. It is because meanings
concept of class consciousness. We will not are not formed and reformedinstantaneously
make assumptions about what is "true" or as a person moves into new class locations
"false" consciousness, but ratheruse the term that class identity can be viewed as a
to refer to those aspects of a person's "backward-looking"concept: it is rooted in
consciousness which have an identifiable one's personalhistory.
class content. By "consciousness" we mean Consciousness structuredaround interests,
those aspects of the mental life of people on the other hand, can be thought of as
which are discursively accessible to the "forward-looking".When Marxists say that
individual'sown awareness. Consciousnessis workers have particular class interest, the
thus counterposed to "unconsciousness"-- presumptionis that the individuals involved
discursively inaccessible aspects of mental are likely to remain workers sufficiently long
life. The elements of consciousness-ideas, so that their individual interests can be
perceptions, beliefs, values-may not, of identified with their positions as workers. If a
course, be continually present in one's particular worker knows that he or she is
awareness, but they must be accessible to that likely to exit the working class in the near
awarenessto count as elements of conscious- future-for example, if they are the children
ness. of wealthy parents and know that they will
In these terms, then, we can talk about the
class characterof consciousness by examining
the class content of the various elements of
18
consciousness, and perhaps even the class See Wright(1985:242-250) for a furtherelaboration
character of gestalts of elements. To talk of this conceptualizationof class consciousness.
'9 Subjective class identity is perhaps the aspect of
class consciousness most studied by sociologists. For a
17
For a classic statementof this position, see Lukacs recent comprehensivediscussion of the problemof class
(1971 [1922]). For a more recent argumentalong similar identity, see Jackmanand Jackman(1983). The classic
lines, see Bertell Ollman (1986). early study on class identity is by Centers (1949).
68 SOCIOLOGICALTHEORY
inherit substantial funds, or if they are 2 = spontaneous workingclass identification
a
attending university and are confident that 1 = forcedchoiceworkingclassidentification
they will get "middle class" jobs upon 0 = no class identityeven when asked to
choose
graduation-then their individual interests -1 = forced choice nonworkingclass identifi-
will be only weakly tied to their currentclass
cation
position. Consciousness of class interests, -2 = spontaneous nonworking classidentifica-
therefore, is oriented towards the future tion
reflectingthe time horizons in terms of which
individuals understand their relationship to We will also analyze workingclass identityas
the class structure. a dichotomousvariable, referredto as simple
workingclass identity, in which 1 = working
class identification (values 1 and 2 in the
Subjectiveclass identity:operationalization working class identity index) and 0 = all
In principle, one would want to know two others.
things about an individual's identity with re-
spect to class: first, how importantor salient Class interest consciousness:
class identitywas to the personrelativeto other
operationalization.
possible identities-ethnic, national, reli-
gious, sexual, etc.; and second, with which A number of Likert-type attitude items on
class the individualidentifies. People may say the survey deal directly or indirectlywith the
that they are in the working class, but class problemof class interests:
identity may have no salience at all to them; 1. Corporations benefitownersat the expense
their identity as Catholics, Irish, football fans
of workersandconsumers(do you strongly
or men may matter more to them than their
agree,somewhatagree,somewhatdisagree,
identityas workers.A full analysisof identity- stronglydisagree).
formation, therefore, has to contend with the 2. During a strike, managementshould be
issue of salience as well as class content. prohibitedby lawfromhiringworkersto take
The data we will be using does not allow the placeof strikers.
for a nuanced treatment of the problem of 3. If given the chance, the nonmanagement
salience. No questions are asked which get at employeesat theplacewereyou workcould
the relative importanceof class compared to runthingseffectivelywithoutbosses.
other possible sources of identity. There is, 4. Big corporations havefartoo muchpowerin
American/Swedish Societytoday.
however, a weaker indicatorof salience in the 5. It is possiblefor a modemsociety to run
survey. Before asking respondents in which effectivelywithoutthe profitmotive.
social class they would place themselves,
they were asked whether or not they thought In addition there was one item which dealt
of themselves as belonging to a social class. with attitudestowardsoutcomes of strikes:
If they responded "yes," they were asked to 6. Imaginethatworkersin a majorindustryare
name the class. If they answered "no", they out on strikeover workingconditionsand
were asked the following: "Many people say wages. Which of the following outcomes
they belong to the working class, the middle wouldyou like to see occur:
class or the uppermiddle class. If you had to 1. The workerswin their most important
make a choice, which class would you say demands.
you belonged to?" We are thus able to 2. the workerswin some of theirdemands
distinguish between people who spontane- andmakesomeconcessions.
that are in the working class 3. the workerswin only a few of their
ously say they demands andmakemajorconcessions.
from people who only say that they are in the
into 4. the workersgo back to work without
working class after being prodded winninganyof theirdemands.
making a choice. If we assume that the
spontaneous response indicates greater sa- These questions were then used to construct
lience of class identitythan the forced choice, two working class interest scales based on
then this enables us to create a working class alternativestrategiesof aggregation:
identityvariablewhich taps both the cognitive 1. Additive scale: Of these six items,
and affective aspects of identity. We will numbers1, 2 and 6 are the most explicitly and
refer to this variable as working class salient narrowlyconcernedwith perceptionsof class
identity. It has the following values: interests. We therefore constructeda simple
TEMPORALITYAND CLASS ANALYSIS 69
additive scale restrictedto these items. Each reformulatedas specific hypotheses about the
item was coded 1 for a proworking class relationshipsamong the operationalizedvari-
interestresponse (strongly agree or somewhat ables we have just defined. These hypotheses
agree on the Likert items and response will guide the statisticalanalysis of this paper:
category 1 on the strike outcome question),
- 1 for a nonworkingclass interest response Hypothesis 1. Class location will explain more
of the variance in the values of
(somewhat or strongly disagree on the Likert the working class interest scales
items and response categories 3 and 4 on the than will class trajectory.
strike outcome question), and 0 for a Hypothesis 2. Class trajectory will explain
class-neutral response (don't know on the more of the variance in working
Likert items and response category 2 on the class identity than will class
strike outcome question). These items were location.
then summedto form a scale in which + 3 = Hypothesis 3. Controllingfor the direct effects
pure proworkingclass interest consciousness of class location, WorkingClass
and -3 = pure antiworking class interest Identity will still be significantly
related to the working class
consciousness.
interest scales.
2. Factor Scale: In addition to the simple
Hypothesis 4. All of the effects in Hypotheses
additive scale, we aggregated all six items 1-3 should be stronger in Swe-
into a scale on the basis of factor loadings on den than in the United States.
a principle components factor analysis. Each
Likert item was code 1-5, with 3 = don't
know. A factor analysis was then conducted DATA AND METHODS
on the six items, and they were aggregated
according to the factor weights on the Data
principle component. The factor loadings are The data we will use come from the Swedish
given in Table 5. and United States surveys of the Comparative
FORMALIZATIONOF HYPOTHESES Project on Class Structure and Class Con-
sciousness (see Wright, 1982). These are both
The four proposition laid out earlier can be nationaland randomsamples of adults in the
Table 5. PrincipalComponentsFactorAnalysis of Class labor force conducted in 1980. The total US
InterestItems
employed labor force sample is 1498 and the
Factor Swedish sample 1145. Details of the samples
Loadings are given in Wright (1985:159-161). The
(Varimax analysis in this paper will exclude all
QuestionnaireItems Rotation)a
nonemployed respondents (i.e. unemployed
1. Corporatebenefit owners at the people in the labor force and housewives).
expense of workers and consumers. .65124
2. Big corporationshas far too much
power in a society today. .56999 Statistical Procedures
3. It is possible for a modem society
to run effectively without the
The class trajectoriesin the various models in
profit motive. .47851
4. I given the chance, the nonmanagement Table 4 are operationalized as a series of
employees at the place where you work dummy variables, one for each of the groups
could run things effectively of cells distinguishedin the models. In those
without bosses. .46712
5. Workersin a major industryare out
analyses in which the consciousness variable
on strike over working conditions and
is continuous-the analysis of salient working
wages. Preferencesfor outcome. .45089 class identity and the working class interest
6. During a strike, managementshould be scales-standard OLS multiple regression
prohibitedby law from hiring workers will be used. For the analysis of simple
to take the place of strikers. .36873
working class identity, where the dependent
(N= 2181)b variableis a dichotomy, formal model testing
Eigen-Value 1.53572
Percent of VarianceExplained 25.6 will rely on logistic regressions.
a There are two principle differences be-
The class interest scale is constructed from the
loadings on this factor. tween logistic regressions and normal regres-
b
This factor analysis was done on the merged US and sions that should be noted for readers
Swedish sample. unfamiliarwith the technique. First, while as
70 SOCIOLOGICALTHEORY
in ordinarylinear regressions the coefficients however, is somewhat easier to run and thus
in logistic regressionscan be used to indicate in the data analyses which follow all of the
the strength of the relationship between results will be from this kind of logistic
independent and dependent variables, the regression.
technical interpretationof the coefficients in
logistic regressionsis different. Insteadof the Formatfor Presentationof Results
value of such coefficients indicting the
number of units change in the dependent Most of the independent variables in the
variable expected for a single unit change in equations we are calculating are structural
the independent variable, they indicate the typologies, generally with four categories
change in the log of the odds of scoring a 1 on (four class locations, four class origins, four
the dependentvariablefor a unit change in the class trajectories). The coefficients of the
independent variable. In more technical dummy variables representingthese typolo-
terms, the functional form of logistic regres- gies in these equations will be presented in
sion is: L = log(p/l-p) = a + biXi, where the form of a triangularmatrixin which each
the X, are the independentvariables,the bi are dummy variable is contrastedwith all of the
the regression coefficients of these variables, others. With four dummy variables, there are
p is the probabilityof identifying oneself in therefore six possible contrasts. These six
the working class and the expression p/(l-p) contrasts are estimated from three different
in the dependentvariable is the odds ratio of regressionequationswith different "left out"
identifying with the working class. The bi categories. The R2s for these three equations,
coefficients of the independent variables, of course, are identical. Presenting the
then, tell you how much log(p/l-p) increases dummy variable coefficients in this way
for each unit increase in the independent allows the reader to evaluate all of the
variables. possible differences among the structural
The second difference between logits and categories in the typology.
standardregressions is that in logits there is
no statisticexactly comparableto R2 to use as
a measure of goodness of fit for alternative Weights
models. Instead, the relative goodness of fit For reasons which are not entirely clear, the
of differentmodels is tested by examiningthe overall U.S. survey contains a disproportion-
improvementsin what is called the "scaled ate numberof respondentswith high levels of
deviance" of the model. Since the scaled education and high occupational status. To
deviance has an asymptotic chi-squaredistri- correct for this overrepresentationwe have
bution, it is straightforwardto test whether a applied a set of post hoc weights to the US
model is improved when additionalindepen- sample which have the effect of reproducing
dent variables are added by seeing if the the 1980 Census occupation-by-education
difference in the scaled deviances in the two distributionsin the sample. The weights are
models is significant given the differences in designed in such a way that they do not alter
the degrees of freedom of the two models. the number of cases in the analyses. These
One final technical note: when all of the weights will be used throughoutthis analysis.
independentvariables in a logistic regression In no instance do they change the results in
are categorical, then the regressions can be ways which would substantively alter the
run either as conventional individual-level interpretationof the findings. No weights
regressions or as what is called "grouped" were needed for the Swedish sample.
logit regressions. In a grouped logit, the
statistical analysis is performeddirectly on a RESULTS
frequency table rather than on individual
observations. The degrees of freedom in a Throughoutthe presentationof resultswe will
groupedlogit, therefore,reflect the numberof discuss the variouspropositionsseparatelyfor
cells in the table rather than the number of men and for women. We do this for two
cases in the sample. The substantiveresultsof reasons. First, there has been considerable
the two analyses are identical: in both cases discussion in recent years about whether the
there will be the same best fitting model and class location of marriedwomen in the labor
the patterns of coefficients in these models force should be defined by their own job class
will be the same. The grouped logit, location, the class location of their husband's
TEMPORALITYAND CLASS ANALYSIS 71
Table 6. Comparisonof R2 andF-testof the TrajectoryModelsPredictingClass InterestsConsciousness(OLSRegression)

Degrees US Male US Female Swedish Male Swedish Female


of
Freedom Adj R2 F Adj R2 F Adj R2 F Adj R2 F
1. SIMPLEADDITIVE SCALE
Modelsa
A. Origins 3 .006 2.34 .000 .34 .026 5.98*** .046 7.20***
B. Locations 3 .076 18.88*** .007 2.39 .176 40.37*** .046 7.26***
(C. 1 .030 21.14*** .000 1.02 .041 24.43*** .020 8.82***
I. 1 .025 17.71*** .000 .68 .097 60.45*** .023 10.08***
E. 2 .039 14.27*** .000 .63 .102 32.32*** .028 6.61**
F. 3 .056 13.81*** .000 .44 .120 26.09*** .026 4.50**
(. 2 .015 5.80** .000 .40 .027 8.74*** .004 1.84
B + E, Fb 5 .075 11.59*** .004 1.49 .179 25.19*** .061 6.08***
A + BC 6 .075 9.72*** .003 1.25 .183 21.63*** .089 7.29***
2. SIX ITEM FACTOR SCALE
A. 3 .011 3.42* .001 .76 .035 7.79*** .041 6.57***
B. 3 .060 14.91*** .001 1.25 .182 42.14*** .054 8.45***
C. 1 .025 17.81*** .007 5.08* .063 38.17*** .020 8.93**
D. 1 .019 13.54*** .000 .58 .103 64.60*** .052 22.34***
E. 2 .031 11.51*** .006 2.54 .118 38.02*** .052 11.74***
F. 3 .046 11.35*** .004 .75 .134 29.46*** .050 7.87***
G. 2 .007 3.13*** .000 .07 .025 8.23*** .015 3.88*
B + E, F 5 .059 9.16*** .003 1.30 .189 26.88*** .070 6.88***
A + B 6 .062 8.13*** .000 .89 .195 23.37*** .088 7.21***
* p<.05 ** p<.01 *** p<.001.
a The letters for different models
correspondto the models displayed in Table 4:
A. Origin Models
B. CurrentLocation Model
C. Privilege Model
D. UnderprivilegedModel
E. Privilege, Underprivilege,Mixed Model
F. Privilege, Underprivilege,Upward, DownwardModel
G. Upward, Downward, Immobile Model
b The Model B + F uses up only 5 degrees of freedombecause the distinctionbetween
category2 and 3 within model
F disappearsin the B + F model (i.e. the model B + E and B + F are mathematicallyequivalent.)
c The model A + B is the conventionaladditive
origin + destinationsmodel.

job or some amalgam of both.20We do not Class Interest Consciousness


wish to engage this debate here and the
easiest way to bracketout the implicationsof Table 6 presents the R2's for the various
the debate is to analyze the results separately models predictingclass interestconsciousness
for men and for women. Secondly, as we using the restrictiveadditive scale and the six
shall see below, the empirical findings vary item factor scale. There is considerable
dramaticallyby gender. In particular,in none variation across categories of actors (the
of the analyses do any of the class location or columns in the table) in how well any of the
class trajectoryvariables have any effect on models fit. In general, the models fit much
class consciousness for American women. better for Swedish men than any other
The interpretabilityof the results, therefore, category, and roughly equally well for
is called into question when the equationsfor American men and Swedish women. Among
men and women are combined. American women none of the models fit the
data very well at all.
For men in both countriesand for Swedish
20 This debate has most women, the currentlocation model (Model B)
recently been sparked by a
contentiouspaperby John Goldthorpe(1983) in which it is a betterpredictorof class interestconscious-
was argued that the class of marriedwomen is strictly ness than is any of the trajectorymodels. For
derivedfrom the class of theirhusbands,regardlessof the
Swedish men, the best trajectory model-
employmentstatusof the woman. For a discussion of the
"cross-class family debate" using the same data as this Model F (the privilege, underprivilege, up-
paper, see Wright (1988). ward, downward model)-explains about 12%
72 SOCIOLOGICALTHEORY
of the variance in class interest conscious- testing the goodness of fit of alternative
ness, whereas the class locations model models from Table 4 for the simple class
(model B) explains abut 18%of the variance. identity dichotomy and for the salient class
While the explained variancesare smaller for identity variable.
American men and especially for Swedish As in the resultsfor class interestconscious-
women for both of these variables, the ness, for Americanwomen in the simple class
patternsare essentially the same. For Ameri- identitylogistic models in Table 8, none of the
can women, on the other hand, neither class models represents a significant reduction in
location for any of the trajectorymodels have scaled deviance over the model containing
any explanatorypower whatsoever. only a constant term (i.e. the model of no
From these results we can conclude that at effects). And in the salient class identity
least for men, currentlocation has a stronger models neitherclass origins, nor currentclass
impact on class interest consciousness than location, nor biographicaltrajectoryexplains
does biographicalclass trajectory.This sup- more than 1%of the variancein class identity
ports the claims of Hypothesis 1. of American women. In the more detailed
analyses of different models which follows,
therefore, the discussion will be restrictedto
Class Identity Americanmen and the Swedish sample.
Table 7 presents the distributionof respon- Among the biographicaltrajectorymodels,
dents with a subjective working class identity Models E and F are clearly the best for these
(simple identity variable) in the full sixteen data. Both of these models provide distinc-
cell class mobility matrixfor Sweden and the tions between three kinds of biographies:(1)
United States. Table 8 presents the results of people who have lived theirentirelives within
the "popular" classes (workers and petty
Table 7. Proportionof Working Class Identificationof
U.S. and Swedish Males and Females by Class
bourgeois); (2) people who have lived their
entire lives in the exploiting classes (middle
Origin and CurrentClass Location
class wage earners and capitalists); and (3)
CurrentClass Location people whose biographiesinvolve movement
(Respondent'sClass) between these two. Model F provides a
Capitalist Petty Middle Working further distinction within the mixed biogra-
Class Bourgeois Class Class phies between those who have moved from
Class popular classes to exploiting classes, and
Origin those who have moved from exploiting
(Parent's classes to popularclasses. Model F is clearly
Class) the best model for Swedish Men, whereas
US Males Model E is the best for Swedish women and
Capitalist .16a (19)b .10 (10) .12 (42) .42 (26) Americanmen.21
Petty B. .43 (21) .58 (12) .20 (30).53 (32)
Middle .21 (19) .50 (4) .20 (106) .37 (86)
How do these trajectorymodels compareto
Working .13 (8) .60 (10) .25 (106) .36 (118)
21
US Females Strictly speaking, the "goodness of fit" cannot be
Capitalist .17 (6) .38 (16) .26 (38).33 (36) directly comparedfor all of these models, since they are
Petty B. .33 (6) .50 (12) .25 (28).44 (57) not all hierarchicallynested. When we compare nonhier-
Middle .22 (9) .14 (14) .23 (61) .39 (66) archically related models, therefore, we are comparing
Working .25 (8) .40 (5) .33 (69).34 (116) their efficiency in reducing the scaled deviance of the
random effects model (the model with only a constant
SwedishMales term). If two models have the same degrees of freedom,
Capitalist .27 (11) .20 (5) .24 (25) .62 (13) and one reduces the scaled deviance more than the other,
Petty B. .38 (13) .58 (12) .49 (35) .71 (75) then we will describe it as a "better"model. In the case
Middle .17 (6) .50 (6) .20 (71) .70 (53) of the comparisonof Models E and F, we can make a
Working .33 (9) .67 (15) .48 (77) .73 (129) direct assessmentof relative goodness of fit since Model
F is nested in Model E. For Swedish men the scaled
Swedish Females
deviance is reducedby 5.98 when we move from Model
Capitalist .00 (1) .00 (0) .08 (13) .40 (20) E to Model F which is significantat the .05 level. There
Petty B. .50 (2) .20 (5) .32 (25) .68 (65) is no improvementin fit whatsoeverfor Swedish women
Middle .50 (2) .00 (2) .16 (56) .43 (49) in model F comparedto model E. For Americanmen the
Working .33 (3) .40 (5) .47 (46) .58 (133) scaled deviance is reduced by 3.56, which is not
a
Entriesare the percentof individualsin the cell who statistically significant at the .05 confidence level.
sayb that they are in the working class. Similar results occur when the R2 in the OLS models in
Numberof cases observed are given in parentheses. Table 8 are compared.
TEMPORALITYAND CLASS ANALYSIS 73
Table 8. Goodness of Fit of Models PredictingClass Identity

1. Simple Class Identity (logistic regression)


Degrees US Sweden
of
of Male Female Male Female
free-
dom Scaled Deviancea Scaled Deviance Scaled Deviance Scaled Deviance
Modelb
Constant 15 46.26 15.05 93.26 60.07
A Origins 12 35.61 11.98 64.42 31.37
B Locations 12 19.58 8.67 23.02 27.28
C 14 26.22 9.18 30.15 24.06
D 14 29.77 10.91 45.53 28.99
E 13 19.60 7.67 14.22 12.68
F 12 15.94 7.19 8.24 12.66
G 13 43.98 13.59 86.65 58.62
B + E,FC 10 14.84 6.92 4.20 6.13
A + Bd 9 10.98 6.40 8.49 7.10

2. Salient Class Identity (OLS multiple regressions)


D.F. Adj R2 F Adj R2 F Adj R2 F Adj R2 F
A Origins 3 .011 3.41* .000 1.06 .045 9.74*** .056 8.64***
B Locations 3 .037 9.22*** .009 2.57 .123 26.81*** .078 11.98***
C 1 .026 18.24*** .009 5.94* .108 67.19*** .073 31.76***
D 1 .030 21.23*** .001 5.83* .084 51.69*** .074 31.85***
E 2 .040 14.49*** .012 4.30* .133 43.64*** .101 22.73***
F 3 .043 10.84*** .012 3.14* .143 31.93*** .099 11.92***
G 2 .002 1.73 .001 .76 .010 3.66* .000 1.07
B + E,FC 5 .041 6.54*** .008 1.88 .148 20.27*** .113 10.93***
A + Bd 6 .044 5.99*** .007 1.67 .141 16.11*** .112 9.16***
* p<.05 ** p<.01 *** p<.001
a The scaled deviance are for
groupedlogistic regressionspredictingthe dichotomoussimple class identityvariable.
The letters for different models correspondto the models displayed in Table 4.
A. Origins Models
B. CurrentLocation Model
C. Privilege Model
D. UnderprivilegeModel
E. Privileged, Underprivileged,Mixed Model
F. Privileged, Underprivileged,Upward, DownwardModel
G. Upward, Downward, Immobile Model
The two models B + E and B + F are mathematicallyequivalent,and thus have the same scaled deviancein all cases.
d
The Model A + B is the conventionalorigins + destinationsmodel.

the location model (Model B)? In Table 8, the again reflecting the somewhat greaterexplan-
currentlocation model and model F both use atory power of the trajectoryvariable com-
three degrees of freedom, but for Swedish paredto the currentlocation variable. For US
men and women the scaled deviance is nearly men the improvementof explainedvarianceis
15 less in the trajectorymodel. For American much smaller, from 3.7% to 4.3% for model
men, model E (using only two degrees of F, but in any case is in the predicted
freedom) has the same scaled deviance as the direction. Overall, then, the evidence sup-
current location model and thus can be ports Hypothesis 2: class trajectorieshave a
considered a better model on grounds of stronger effect on class identity than does
efficiency. When the continuous variable currentclass location.
"salient class identity" is the dependent
variable, the adjusted R2 for the current Class Identityeffects on Class Interests
locations model is 12.3% for Swedish men
and 7.8% for Swedish women. This increases Table 9 presents the results for the multivari-
to 14.3% for model F for Swedish men and ate regressions of Class location and class
10.1% for model E for Swedish women, identity on class interest consciousness. In
Table 9. The Effects of Class Locationand Class Identityon Class InterestConsciousness (unstandardizedregressioncoefficients

DependentVariable= simple additive interest scale Dependent Variabl


Eq.(l) Eq.(2) Eq.(3) Eq.(l)
1. US MALES (N=650)
Classa Cap PB Mid Cap PB Mid Cap PB
Petty Bourgeoisie .39 .34 .71
Middle Class .50* .11 .52* .18 .72* .01
WorkingClass 1.35*** .96** .85*** 1.31*** .97*** .79*** 1.79*** 1.09**
Class Id Dummy .46** .31*
AdjustedR2 .08 .01 .08 .06
2. US FEMALES(N = 549)
Classa Cap PB Mid Cap PB Mid Cap BP
Petty Bourgeoisie -.48 - .51 .08
Middle Class -. 15 .33 -.16 .35 .08 .00
WorkingClass .08 .56* .23 .04 .55* .20 .39 .31
Class Id Dummy .29* .28*
AdjustedR2 .01 .01 .01 .00
3. SWEDISHMALES (N = 555)
Classa Cap PB Mid Cap PB Mid Cap PB
Petty Bourgeoisie 1.14** .87** 1.54**
Middle Class 1.28*** .14 1.23*** .36 1.74*** .21
WorkingClass 2.48*** 1.34*** 1.20*** 2.04** 1.17** .81** 3.83*** 2.29***
Class Id Dummy 1.47** 1.12**
AdjustedR2 .18 .16 .26 .18
4. SWEDISHWOMEN(N = 389)
Classa Capb PB Mid Cap PB Mid Cap PB
Petty Bourgeoisie -.79 -.60 .21
Middle Class .46 1.25** .63 1.23** 1.15 .95
WorkingClass .92 1.71*** .47** .87 1.47*** .24 2.23* 2.03*
Class Id Dummy .87*** .78***
AdjustedR2 .05 .08 .10 .05
* p<.05 ** p<.01 *** p<.001
a The coefficientsfor the class location
dummy varaiblesrepresentthe differencesbetween the column category and the row categ
equationswith differentleft out categories. With four dummy variablesthere are thus six possible contrasts.
b
There are only five Swedish woman capitalistsin the sample, so the standarderrorsfor these contrastsare quite large.
TEMPORALITYAND CLASS ANALYSIS 75
each of the equations, class identity signifi- sample is the currentlocation model predict-
cantly predicts subjective class interests. The ing class interestconsciousness for American
results are particularlystriking for Swedish men compared to Swedish women, but the
men: Class location and working class difference in R2 is relatively small (6.0%
identity alone account for 18% and 16% of comparedto 5.4% for the factor scale).
the variance in working class interest con- Second, when we look at the coefficients in
sciousness respectively; when combined they the models we have been examining, the
account for 26% of the variance. For such a magnitude of these coefficients are consis-
simple multivariatemodel, this is a high R2 tently greaterin the Swedish equationsthan in
for a regression on attitudes. The explained the US equations, particularly when we
variances are lower for Swedish women, but contrast Swedish men with both men and
it is still the case that class identity predicts women in the United States. Table 10
class interests net of class location. For the presentsthe coefficients for the class interests
American equations the effects of class models and Table 11 for the class identity
identity on interests, while still statistically models. (Only the coefficients for Model F
significant for both men and women, are among the trajectorymodels are shown since
much weaker than in Sweden, adding at most this is the model of greatest theoretical
a marginalincrementin explained variance. interest.) Table 12 presents t-tests for signifi-
cance of differences in coefficients across the
ComparingSweden and the United States samples.22
In the class identity models, for Model F
Hypothesis 4 argues that the effects postu- the coefficients involvingcontrastswith trajec-
lated in Hypotheses 1, 2 and 3 should be tory (1) for both men and women in Sweden
stronger under macrosocial conditions in are generally roughly twice as large as those
which social and political institutions are for American men. In the current location
systematically organized along class lines. model (Model B) for the equationspredicting
Given the working class-centered ideology simple class identity, the coefficient for the
which continues to play a much more contrast between the working class location
importantrole in the Swedish social demo- and the capitalistlocation is 1.71 for Swedish
cratic party than in the US DemocraticParty, men but only .65 for Americanmen, and 1.18
and given the organizationalstrength of the and .44 respectively for the equationspredict-
labor movement in Sweden, this hypothesis ing salient class identity. Among Swedish
would lead us to expect considerablystronger women, the contrasts between workers and
patterns of results in Sweden than in the capitalists are not statistically significant in
United States. any of the identity equations (there are only
A numberof results are consistent with this eight Swedish women capitalists in the
expectation. First, as already noted on a sample which makes the standarderror for
numberof occasions, the overall explanatory this contrast quite large) but the contrast
power of the variousmodels tends to be much between workers and petty bourgeoisie is
greater in Sweden than in the United States. nearly twice as large for Swedish women as
For the class identity variablein table 8, all of for Americanmen.
the models have higher R2 or greater The cross-national contrasts are equally
improvements in scaled deviance for both strong for the class interest consciousness
Swedish men and women than for American equations. Not only does the currentlocation
men. The models have little or no explanatory equationfor Swedish men explain three times
power for American women. For the class as much variance as the equation for
interest consciousness variables, the models American men, but the coefficients in con-
for Swedish men all explain more variance trasts between workers and capitalists are
than the models for US men. The class roughly twice as large (see Table 11). The
location model (B) in Table 6 explains just coefficient in the current location model for
over 18% of the variance in class interest
consciousness (factor scale) among Swedish
22 We have
men, but only about 6% of the variance adopted a conservative test of the
difference in coefficients across samples: t = (coeffl
among American men. The only contrast in - coeff2)/V(sel2 + se22). This test will always give a
which a model for the Americansample has a smallert-ratiothan a test based directly on an interaction
greater R2 than a model for the Swedish term in an equationwhich combines the two samples.
Table 10. Estimatesof Coefficientsfor Models PredictingClass Identity (unstandardizedregressionscoefficients)

US
Modelsa Male Female Mal
1. SIMPLECLASS IDENTITY
(groupedlogistic regression)
MODELA (ORIGINMODEL)b
CAP PB M CAP PB M CAP PB
PM Origin .99** .40 1.13**
M Origin .38 -.61** -.04 -.44 .31 -.82**
W Origin .59* -.40 .21 .17 -.23 .21 1.21*** .08
MODELB (CURRENT
LOCATIONMODEL) CAP PB M CAP PB M CAP PB
PB Location .86* .48 1.02*
M Location -.26 - 1.12** .18 -.31 .22 -.81*
W Location .65* -.21 .90*** .64 .15 -.46** 1.71** .69*
MODELA + B (ORIGIN +
CURRENTLOCATIONMODEL) CAP PB M CAP PB M CAP PB
PB Origin .93** .34 .77*
M Origin .39 -.55* -.05 -.39 .12 -.65**
W Origin .51* -.42 .13 .12 -.23 .26 .85** .09
PB Location .85* .49 .93
M Location -.21 - 1.06** .19 -.29 .21 -.72*
WLocation .66* -.18 .88*** .62 .13 .43** 1.59*** .65
MODELF (BIOGRAPHIC
TRAJECTORYMODEL)C (1) (2) (3) (1) (2) (3) (1) (2)
(2) .49* .35 1.16***
(3) .98*** .48* .54* .19 1.87*** .71*
(4) 1.23*** .74** .25 .71** .35 .16 2.21*** 1.05*
[See end of Table 10 on next page for table footnotes]
Table 10. (Continued)

US
Models" Male Female Ma
2. SALIENTCLASS IDENTITY
(OLS regression)
MODELA (ORIGIN'MODEL)b
CAP PB M CAP PB M CAP PB
PB Origin .51** .30 .70**
M Origin .10 -.40** .08 -.22 .15 -.55*
W Origin .27 -.24 .17 .18 -.12 .10 .81*** .10
MODEL B. CURRENT
LOCATIONMODEL CAP PB M CAP PB M CAP PB
PB Location .48 .31 .70*
M Location -.08 -.56* .03 -.28 .16 -.54*
W Location .44** - .04 .52*** .34 .03 .31 1.18*** .49
MODELA + B (ORIGIN +
CURRENTLOCATIONMODEL) CAP PB M CAP PB M CAP PB
PB Origin .45* .26 .40
M Origin .08 -.37* .09 -.18 .01 -.39*
W Origin .20 -.26 .12 .16 -.10 .08 .50* .10
PB Location .45 .33 .62*
M Location -.04 -.49* .04 -.29 .16 -.47
W Location .46** .01 .50*** .33 -.00 .29* 1.08*** .46
MODELF (BIOGRAPHIC
TRAJECTORYMODEL)C (1) (2) (3) (1) (2) (3) (1) (2)
(2) .20 .16 .69***
(3) .47*** .27 .31 .15 1.21** .51
(4) .70*** .50*** .23 .45** .29 .13 1.41*** .72
* p<.05 ** p<.01 *** p<.001
a
See Table 4 for full definitionof models:
CAP= Capitalistclass; PB = Petty Bourgeoisie; M = Middle Class Wage Earners;W = WorkingClass Wage Earners.
b
The coefficients in the triangularmatrixrepresentthe differencebetween the column and the row category. The coefficients wer
column categoriesare successively the "left out categories".
c The numbersin this column refer to the levels in model F in Table 2:
Level (l)=a life entirely in the privileged classes.
Level (2)= upwardmobility from popularclasses to privileged classes.
Level (3)= downwardmobility from privileged classes to popularclasses.
Level (4)= a life entirely in the popularclasses.
78 SOCIOLOGICALTHEORY
Table 11. Estimatesof OLS Regression Coefficients for Models PredictingClass InterestConsciousness

US Sweden
Male Female Male Female
1. DEPENDENTVARIABLE= ADDITIVEWORKINGCLASS INTERESTSCALE
MODELA (ORIGINMODEL)b
CAP PB M CAP PB M CAP PB M CAP PB M
PB .23 .02 .45 1.03**
M .32 .09 .04 .03 .27 -.18 .93** -.10
W .52* .29 .20 .15 .14 .11 .90*** .45* .63** 1.34*** .31 .41*
MODELB (CURRENTLOCATIONMODEL)
CAP PB M CAP PB M CAP PB M CAP PB M
PB .39 -.48 1.14** -.79
M .50* .11 -.15 .33 1.28*** .14** .46 1.25**
W 1.35*** .96** .85*** .08 .56* .23 2.49*** 1.34*** 1.20*** .92 1.71*** .47**
MODELA + B (ORIGIN+ CURRENTLOCATIONMODEL)
CAP PB M CAP PB M CAP PB M CAP PB M
PB Origin .18 -.04 -.05 1.07***
M Origin .16 -.03 .00 .05 -.11 -.06 1.00** -.07
W Origin .26 .08 .11 .06 .10 .06 .32 .37* .43* 1.32*** .25 .32
PB Now .37 -.47 1.09** -.94
M Now .46* .09 -.16 .31 1.25*** .16 .44 1.39**
W Now 1.29*** .92** .83*** .07 .54* .23 2.41*** 1.32*** 1.16*** .83 1.77*** .39*
MODELF (BIOGRAPHIC TRAJECTORYMODEL)
(1) (2) (3) (1) (2) (3) (1) (2) (3) (1) (2) (3)
(2) .21 .10 .08 .48
(3) .90*** .69*** .15 .05 .93*** .85*** .33 -.15
(4) .94*** .73*** .04 .19 .10 .05 1.40*** 1.32*** .47* .77*** .29 .44*
2. DEPENDENTVARIABLE = WORKINGCLASS INTERESTFACTORSCALE
MODELA (ORIGINMODEL)b
CAP PB M CAP PB M CAP PB M CAP PB M
PB .26 .30 .81 1.62**
M .61* .35 .29 -.01 .22 -.59 1.32* -.31
W .86** .60* .25 .35 .05 .06 1.50*** .69* 1.28*** 2.05 .43 .74*
MODELB (CURRENTLOCATIONMODEL)
CAP PB M CAP PB M CAP PB M CAP PB M
PB .71 .08 1.54* .21
M .72* .01 .08 .00 1.74*** .21 1.15 .95
W 1.79*** 1.09* 1.08*** .39 .31 .31 3.83*** 2.30*** 2.09*** 2.23* 2.03** 1.08***
MODELA + B (ORIGIN + CURRENTLOCATIONMODEL)
CAP PB M CAP PB M CAP PB M CAP PB M
PB Origin .19 .25 -.00 1.59**
M Origin .42 .23 .27 .02 -.36 -.36 1.46** -.13
W Origin .55 .36 .13 .30 .04 .02 .57 .58* .94** 1.95*** .36 .49
PB Now .68 .12 1.44* -.02
M Now .58 -.10 .08 -.04 1.73*** .29 1.14 1.16
WNow 1.63*** .95* 1.05*** .37 .25 .29 3.71*** 2.27*** 1.98*** 2.10* 2.12** .96***
MODELF (BIOGRAPHICTRAJECTORYMODEL)C
(1) (2) (3) (1) (2) (3) (1) (2) (3) (1) (2) (3)
(2) .29 .41 .55 .30
(3) 1.22*** .93** .51* .10 1.83*** 1.28** .49 .19
(4) 1.22*** .93*** .00 .45 .04 -.06 2.57*** 2.01*** .73* 1.45*** 1.15** .96**
<. p<.05* **
p<.01*** p<.001
a See Table 4 for full definition of models.
CAP= Capitalistclass; PB = PettyBourgeoisie;M = MiddleClass Wage Earners;W = WorkingClass Wage Earners.
b The coefficients in the triangularmatrix represent the difference between the column category and the row
category. The coefficients were calculated from three separate regressions in which the column categories are
successively the "left out categories".
c The numbersin this column refer to the levels in model F in Table 2:
Level (l)=a life entirely in the privileged classes.
Level (2)= upwardmobility from popularclasses to privileged classes.
Level (3)= downwardmobility from privileged classes to popularclasses.
Level (4)= a life entirely in the popularclasses.
TEMPORALITYAND CLASS ANALYSIS 79
Table 12. Comparisonof Coefficients of Models PredictingClass Identificationand Class Interests
Entriesin the table are differencesin coefficients between groups in equationsin Tables 10 and 1.a

Swedish Male Swedish Female Swedish Male vs


Models vs US Male vs US Male Swedish Female
1. DEPENDENTVARIABLES= SIMPLECLASS IDENTITY(groupedlogistic regressions)
Model A (ORIGINMODEL)
Cap PB M Cap PB M Cap PB M
PB .15 .27 -.12
M Origin -.07 -.21 -.28 -.54 .21 .33
W Origin .62 .48 .69** .62 .35 .89** .01 .12 -.20
Model B (CURRENTLOCATIONMODEL)
Cap PB M Cap PB M Cap PB M
PB Location .17 - 1.95* 2.12**
M Location .48 .31 -.69 1.26 1.17 -.95
W Location 1.07** .90 .59** -.39 1.56* .30 1.46* -.66 .30
Model F (BIOGRAPHICTRAJECTORYMODEL)
F(1) F(2) F(3) F(1) F(2) F(3) F(1) F(2) F(3)
F(2) .67* .90* - .23
F(3) .89* .22 .36 - .54 .53 .76
F(4) .98** .31 .09 .87* - .04 .51 .11 .34 -.42
2. DEPENDENTVARIABLE= SALIENT CLASS IDENTITY(OLS regression)
Model A (ORIGINMODEL)
Cap PB M Cap PB M Cap PB M
PB Origin .19 .26 -.17
M Origin .05 -.15 -.01 -.38 .06 .23
W Origin .53* .34 .49** .57* .21 .59** -.04 .13 -.10
Model B (CURRENTLOCATIONMODEL)
Cap PB M Cap PB M Cap PB M
PB Location .22 - 1.19* 1.40*
M Location .24 .02 -.40 .79 .64 -.76
W Location .74** .53 .50** -.04 1.15* .36* .78 -.63 .14
Model F (BIOGRAPHICTRAJECTORYMODEL)
F(1) F(2) F(3) F(1) F(2) F(3) F(1) F(2) F(3)
F(2) .49* .54* -.04
F(3) .74** .25 .30 .24 .44 .48
F(4) .71*** .22 -.02 .62** .08 .32 .10 .14 -.34
3. DEPENDENTVARIABLE= ADDITIVE WORKINGCLASS INTERESTSCALE
Model A (ORIGINMODEL)
Cap PB M Cap PB M Cap PB M
PB Origin .22 .80* -.58
M Origin -.04 .27 .61 -.19 -.65 -.08
W Origin .38 .16 .43* .82* .02 .22 -.43 .14 .21
Model B (CURRENTLOCATIONMODEL)
Cap PB M Cap PB M Cap PB M
PB Location .75 -1.18 1.94**
M Location .78* .03 -.05 1.14* .83 -1.11
W Location 1.14*** .37 .36* -.42 .76 -.38* 1.56** -.39 .74***
Model F (BIOGRAPHICTRAJECTORYMODEL)
F(1) F(2) F(3) F(1) F(2) F(3) F(1) F(2) F(3)
F(2) -.13 .27 -.39
F(3) .03 .16 -.57 -.84** .60 1.00**
F(4) .46* .59* .43 -.18 -.44 .40 .64* 1.03** .04

4. DEPENDENTVARIABLE= WORKINGCLASS INTERESTFACTORSCALE


Model A (ORIGINMODEL)
Cap PB M Cap PB M Cap PB M
PB Origin .55 1.36* -.82
M Origin -.39 -.94* .71 -.65 -1.10 -.28
W Origin .64 .09 1.02** 1.19* -.17 .48 -.55 .26 .54
80 SOCIOLOGICALTHEORY
Table 12. (Continued)

Swedish Male Swedish Female Swedish Male vs


Models vs US male vs US Male Swedish Female
Model B (Location)
Cap PB M Cap PB M Cap PB M
PB Location .83 -.50 1.33
M Location 1.04* .20 .44 .94 .59 -.74
W Location 2.04*** 1.21* 1.01*** .44 .94 .00 1.60 .27 1.01**
Model F (BIOGRAPHICTRAJECTORYMODEL)
F(1) F(2) F(3) F(1) F(2) F(3) F(1) F(2) F(3)
.26 .01 .25
F(3) .62 .35 -.73 -.74 1.34* 1.09*
F(4) 1.35*** 1.08** .73 .23 .22 .96* 1.12** .87* -.23
a
The standarderrorof the difference is coefficients has been calculatedas the squareroot of the sum of the squares
of each coefficient. The correspondingt-ratiodivides the difference in coefficients by this standarderror. Since this is
a quite conservative test of differences in equations, and since we have specific directionalhypotheses about these
differences, we have reportedone-tailed tests.
* p<.05 ** p<.01 *** p<.001 (one-tailed tests)

the contrast between capitalists and workers Class and Gender in National
in the Swedish male equationpredictingclass Differences in Consciousness
interestconsciousness (additive scale) is 2.49
whereas it is only 1.35 for Americanmen. In It is beyond the scope of this paperto system-
the more general working class interest scale, aticallyexplorethe role of genderin the models
the contrasts between workers and all other we have been exploring.Nevertheless,the ef-
classes and between capitalists and all other fects of genderare sufficientlystrikingin these
classes are all around twice as large among resultsto warrantsome comment.In particular,
Swedish men as among Americanmen. What thereis a sharpcontrastbetweengenderdiffer-
these figures indicate is that not only does ences and nation differencesin these models:
class location more systematically predict nationaldifferencesare particularlymarkedin
class interestconsciousness in Sweden than in theoverallexplanatory powerof themodels;gen-
the United States, but the class structureis der differencesare markedin both the explana-
more ideologically polarized in Sweden.23 tory power of the equationsand the patternsof
Finally, the relationship between class coefficients associatedwith the models. These
identityand class interestsis much strongerin contrastsare summarizedin Table 13. We have
Sweden than in the United States. In the alreadyexaminedthe nationaldifferencesin ex-
zero-order relationship between identity and planatorypower of these models. Let us now
interestconsciousness (equation2 in table 9), look more closely at the gendercontrasts.
the R2 in the Swedish equations is about 8% Within both Sweden and the United States,
for women and 16%for men, but a negligible the explanatorypower of all the equations is
1%for both men and women in the US. In the much higher for men than for women. The R2
multivariate equation, the coefficient for in the basic models we have investigated are
identity net of class location in the Swedish typically in the 5-8% range for US men but
sample is roughly three times greaterthan in generally less than 1%for US women, and in
the US sample. the 15-20% range for Swedish men but only
Taking these various results together, it is 6-10% for Swedish women. At least as
reasonable to conclude that both class loca- measured with the variables and models in
tion and class trajectory are more powerful this study, both class location and class
determinantsof class consciousness in Swe- trajectory are more systematically linked to
den than in the United States. The results are dimensions of class consciousness among
thus strongly supportiveof Hypothesis 4. men than among women.
The gender contrasts in the pattern of
23 The textureof this
coefficients is equally striking.24 If we
greaterideological polarizationin
Sweden than in the United States is explored systemati-
cally in Wright (1985: chapter 6). See also Carolyn 24
Howe, (1987: chapter9). Since none of these models has more than a trivial
TEMPORALITYAND CLASS ANALYSIS 81
Table 13. Summaryof GenderContrastsAcross Countries

US Men Swedish Men US Women Swedish Women


EXPLANATORY POWER
OF EQUATIONS(R2)a
location -- interests .08(.06)b .18(18) .01(.00) .05(.05)
location - identity .04 .12 .01 .08
trajectory- identity .04 .14 .01 .10
trajectory-- interests .06(.05) .12(.13) .00(.00) .03(.05)
identity - interests .01(.02) .16(.19) .01(.01) .08(.08)
location + identity - interests .08(.07) .26(.28) .01(.01) .10(.11)
PA7TERNSOF COEFFICIENTS
location and origins - interests location mattersmuch neitherlocation origins mattermuch
more than origins; or origins has more than locations;
no significant coefficients significanteffects no significantcoeffi-
for origins cients for locations
trajectory-, identity mobility into popularclasses patternof coefficients mobility into and
associated with stronger similar to men, out of popular
working class identity than but much weaker classes have identical
mobility out of popular levels of working
classes class identity
aThe R2 figures have been roundedoff to the nearestpercent.
b
The figures in parenthesesrefer to the six itemfactor scale; the figures not in parenthesesrefer to the simple
additive scale.

examine men separately,the basic patternof ables have consistently large and significant
coefficients in the models tends to be quite coefficients, whereas the coefficients for
similar in the United States and Sweden, in origins are nearly all statisticallyinsignificant
spite of the fact that the coefficients are and much smaller in absolute magnitude.
generally much larger in Sweden than in the The pattern of coefficients for Swedish
United States and the explanatorypower of women, in contrast, is very different from
the Swedish models considerably greater. men in both countries. In the class identity
Thus, for example, in the trajectorymodels models for Swedish women (Table 10), the
predictingclass identity (Table 10), for men two mixed trajectories (from exploiting to
in both countries, category 4 in model F (a popular classes and from popularto exploit-
life entirely within the popular classes) does ing classes), are virtually indistinguishable,
not differ significantly from category 3 and both have a significantly lower probabil-
(people mobile into the popularclasses) in the ity of working class identification than a
probabilityof having a workingclass identifi- trajectoryentirely within the popularclasses,
cation, whereas at least in the simple class whereas as already noted, for men in both
identity equation, category 3 does differ countries there is no difference between
significantlyfrom category 2 (mobility out of trajectories 3 and 4, while "downward"
the popularclasses). Similarly, in the current mobility generally differs from "upward"
locations model predicting class interest mobility.25Similarly, whereas among men in
consciousness (Table 11), for men in both both Sweden and the United States there is
countries the working class has the highest relatively little difference between the petty
level of working class interestconsciousness, bourgeoisie and middle class in the class
the capitalist class has an antiworkingclass location model predicting class interest con-
interests consciousness, and the petty bour- sciousness(Table 11), amongSwedishwomen
geoisie and middle class wage earners have the petty bourgeois location is much less
rather similar intermediate coefficients. Fi- proworking class than middle .class wage
nally, in the additive model of origins and earnersfor the additiveinterestscale. What is
locations (model A + B in Table 11), for men
in both countries the current locations vari-
25
American women have similar patternsto men in
both countries, but because the coefficients are so much
effect on class consciousness for Americanwomen, all of weakerand the standarderrorslarger, we have much less
the discussion which follows will focus on Swedish confidence in the descriptive accuracy of these coeffi-
women comparedwith Americanmen and Swedish men. cients.
82 SOCIOLOGICALTHEORY
particularly striking for the class interest extent that this is true, then the class interests
results is the patterns for the origins + of marriedwomen will be determinedless by
location model (Model A+B in Table 11). the class of their own job and more by the
For Swedish women the coefficients for the class of their husbands. Furthermore,over
origins dummy variables are all large and time this will also mean that their class
highly significant in the contrastwith capital- identity will depend more on their family-
ist class origins, while, in contrast,for men in class trajectorythan simply on their individ-
both countries, none of the coefficients for the ual trajectory (as defined in this study).
class origins contrastswith the capitalistclass Individual class location and trajectory will
are statistically significant. therefore tend to have lower explanatory
We do not have a definitive interpretation power among women than among men. And,
of these rather complex interactions of depending upon how the individual's class
gender, country and class. We can, however, variables are causally tied to the correspond-
offer some preliminarysuggestions both for ing family class variables, the patterns of
why individual class location and trajectory coefficients for individual class variables
variablesseem to have less explanatorypower would also be expected to be different for
for women than for men in general, and women than for men.
second, for why the patterning of effects Research by Wright (1988) on the effects
seems so different for men and (at least) of family class composition on class identity
Swedish women. supports this interpretation.In that study he
There are two possible lines reasoning for found that in the United States the class
explaining these gender differences. First, the identity of marriedwomen in the labor force
lower causal efficiency of class trajectoryand is almost entirely determined by the class
class location could primarilybe a measure- location of their husband'sjobs ratherthan by
ment problem. Perhaps the same formal their own jobs, whereas in Sweden the two
criteria for class divisions have a different jobs have roughly equal impact on the class
class content among men than among women. identity of married women. In contrast, the
It could be, for example, that because of class locations of wives' job had either no
gender discriminationthe productiveassets of effect on husband'sclass identity (in the US)
"middle class" women (skills/credentials)are or a much smaller effect than the husband's
devalued and are thus unable to be converted own job (in Sweden). Wright interpretsthe
into exploitation. This would imply that the contrast between Swedish and American
distinctionbetween working class and middle married women as reflecting the greater
class among women is a weakerclass division dependency of American women than Swed-
than among men. This, in turn, would both ish women on the economic fate of their
reduce the explanatory power of class- husbands, both because wage inequalities
centered variables (both location and trajec- between men and women are greater in the
tory) and potentially change the patterningof United States than in Sweden and because the
their coefficients. redistributive polices of the state make
An alternativeapproachto explainingthese women in general less dependent upon their
patternsof results is to search for some other husbands in Sweden. This greater economic
mechanismthat acts to block the effectivity of dependency of women on men in the United
these variables among women (or enhance States undermines the causal effects of
their effectivity among men). One likely women's individual class variables on their
candidatefor this is the relationshipbetween consciousness. But even in Sweden, married
gender, family structureand class. As many women remain more economically dependent
people have argued, because of deeply upon their husbands than married men are
entrenchedpatternsof gender inequality, the upon their wives, and thus individual class
material interests of married women are variables have less explanatory power for
generally more tied to the economic fate of Swedish women than for Swedish men.
their husbands than are the materialinterests
of marriedmen tied to the economic fate of CONCLUSION
their wives. This is obvious in the case of
full-time housewives, but as Goldthorpe This paperhas attemptedto join structuraland
(1983) has argued, it is also true for most processual approacheswithin class analysis.
women in the paid labor force as well. To the Much of the analysis is certainly open to
TEMPORALITYAND CLASS ANALYSIS 83
criticism. Processualapproachesare generally impossible to define the class content of
built around a fairly rich inventory of class biographicaltrajectories,communities, work-
constitutiveexperiencesin communities,work- place experiences, etc. withouthaving a prior
places, schools, etc. Our use of processual elaborationof a structuralclass concept. But
approaches, in contrast, is based on a fairly this conceptual primacy does not imply
impoverishedoperationalizationof class tra- explanatoryprimacy in the lives of individu-
jectory, relying entirely on the linkage als.
between class origins and class location. Third, all of these micro-mechanismsof
Similarly, contemporary structural ap- consciousness formationare heavily mediated
proaches to class generally move beyond by the macro-institutionalcontext in which
simple working class vs. middle class divi- they operate. Where political and economic
sions among wage earners,but because of the conflicts are heavily organized along class
limitations in our data and sample size, we lines, these micro-mechanismsare likely to
had to rely on this fairly crude class model. be reinforced;where class has been margina-
What is more, as our results for gender lized in institutional forms of collective
reveal, even within this simple class structural organization,then the micro-mechanismswill
model, our results may be significantly be weaker. Again, this reflects a fundamental
affected by ignoring the problem of the insight of processual theorists of class: the
location of families within the class structure. effectivity of the individual experiences that
In spite of these limitations, we believe that are constitutive of class depends upon the
the data we have presented broadly support historical context within which those experi-
the general orienting propositions of this ences occur.
paper. In particular, the following general Finally, and not as part of the original
conclusions should be stressed: agenda of this paper, all of the results we
First, class trajectories seem to be more have observed of both a processual and
importantthan class locations in shaping the structural nature vary sharply by gender.
more culturaldimensions of class conscious- While this does not mean that class mecha-
ness, whereas class locations are relatively nisms are inherentlygendered in the sense of
more important for shaping the more nar- having no independenteffectivity in theirown
rowly interest-centereddimensions. This is in right, it does mean that the empirical
keeping with the view that class interest investigation of the effects of both class
consciousness dependsmore upon the percep- location and class trajectorymust include an
tions of one's future class situation than on analysis of interactionswith gender.
one's past, whereas class identity is more
heavily shaped by one's biography in class REFERENCES
based communities, social networks and
Bourdieu, Pierre. 1985. "The Social Space and tle
interactions. Structural and processual ap- Genesis of Groups," Theoryand Society, 14:723-744.
proaches to class, therefore, seem most Bowles, Sam and Herb Gintis. 1986. Capitalism and
appropriatefor explaining different dimen- Democracy. New York: Basic Books.
sions of consciousness. Brubaker,Rogers. 1985. "RethinkingClassical Theory:
the Sociological Vision of Pierre Bourdieu" Theory
Second, the perceptionof class interests is and Society, 14:745-775.
shaped at least as much by class identities as Carchedi, G. 1977. On the Economic Identificationof
it is by class locations. While structural Social Classes. London:Routledge & Kegan Paul.
approachesto class may be correctin positing Centers, Richard. 1949. The Psychology of Social
a set of material interests tied to class Classes. Princeton:PrincetonUniversity Press.
Cohen, G.A. 1978. Karl Marx's Theory of History: a
structures which create tendencies towards Defense Princeton:PrincetonUniversityPress.
class interestconsciousness, the actualpercep- Duncan, Otis Dudley. 1966. "MethodologicalIssues in
tion of those interests depends as much or the Analysis of Social Mobility," in Social Structure
more upon class identity as upon class and Mobility in Economic Development, edited by
N.J. Smelser and S.M. Lipset, Chicago: Aldine.
location itself. In' this sense processual Elster, Jon. 1985. Making Sense of Marx. Cambridge:
theorists may be correct in emphasizing the CambridgeUniversity Press.
primacy of process over structurein explain- Goldthorpe,John. 1983. "Womenand Class Analysis: a
ing empirical variations in class conscious- Defense of the Conventional View" Sociology,
ness and class behavior. There remains, of 17:465-488.
Halaby, Charles and Michael Sobel, 1979. "Mobility
course, a conceptual primacy of structural Effects in the Workplace" American Journal of
over processual class concepts since it is Sociology, 85:385-416.
84 SOCIOLOGICALTHEORY
Howe, Carolyn. 1987. Class Structure and Class Przeworski,Adam. 1985. Capitalismand Social Democ-
Formation in Advanced Capitalism: a Comparative racy. Cambridge:CambridgeUniversity Press.
Study of Sweden and the United States, PhD Seeman, Melvin. 1977. "Some Real and Imaginary
Dissertation, Departmentof Sociology, University of Consequencesof Social Mobility: A French American
Wisconsin, Madison. Comparison." American Journal of Sociology,
Jackman, Mary. 1972. "The Political Orientationof the 82:757-782.
Socially Mobile in Italy: a Re-Examination." The Thompson, KennethH. 1971. "UpwardSocial Mobility
AmericanSociological Review 37:213-22. and Political Orientation: a Re-evaluation of the
Jackman,Mary R. and RobertW. Jackman. 1983. Class Evidence" The American Sociological Review.
Awareness in the United States. Berkeley: University
36:223-235.
of CaliforniaPress.
Knoke, David. 1973. "IntergenerationalMobility and the Wright, Erik Olin, Cynthia Costello, David Hachen and
Political PartyPreferencesof Men." AmericanJournal Joey Sprague. 1982. "The American Class Structure"
The AmericanSociological Review, 47:709-726.
of Sociology, 78:1448-1468.
Lukacs, George (1971 [1922]). History and Class Wright, Erik Olin. 1982. "The ComparativeProject on
Consciousness. Cambridge:MIT Press. Class Structure and Class Consciousness: an Over-
Oilman, Bertell. 1986. "How to Study Class Conscious- view" Technical paper # 1, The ComparativeProject
ness" paper presented at the annual meetings of the on Class Structureand Class Consciousness, Depart-
American Sociological Association, New York, Au- ment of Sociology, Universityof Wisconsin, Madison.
gust, 1986. Wright, Erik Olin. 1985. Classes. London:NLB/Verso.
Poulantzas, Nicos. 1975. Classes in Contemporary Wright, Erik Olin. 1988. "Women in the class
Capitalism. London:New Left Books. structure,"forthcoming,Politics & Society.

Potrebbero piacerti anche