bringing intergenerational social mobility research into...

To understand current inequality socialmobility researchers must bridge a long-

standing gap between theory and practice thatincreasingly distorts social mobility and strati-f ication research f indings1 A gap exists

because in theory class background (ie child-hood class position) is a family-level variablebut the conventional research practice equatesclass background solely with a fatherrsquos classposition This assumes that mothersrsquo econom-ic participation is not common or important toclass background and that father-headed fami-lies are the norm Yet in the United States ratesof labor force participation among mothers havesteadily increased since intergenerational classmobility models were first developed Figure 1illustrates this trend

The bias that comes from excluding mothersrsquoclass characteristics is increasingly importantbut not widely recognized or understoodScholarly debate over the conceptualization andmeasurement of family-level class position haswaned since the early 1990s with a generalconsensus that the conventional mobility

Bringing Intergenerational Social MobilityResearch into the Twenty-first CenturyWhy Mothers Matter

Emily BellerUS Government Accountability Office

Conventional social mobility research which measures family social class background

relative to only fathersrsquo characteristics presents an outmoded picture of familiesmdasha

picture wherein mothersrsquo economic participation is neither common nor important This

article demonstrates that such measurement is theoretically and empirically untenable

Models that incorporate both mothersrsquo and fathersrsquo characteristics into class origin

measures fit observed mobility patterns better than do conventional models and for both

men and women Furthermore in contrast to the current consensus that conventional

measurement strategies do not alter substantive research conclusions analyses of cohort

change in social mobility illustrate the distortions that conventional practice can

produce in stratification research findings By failing to measure the impact of mothersrsquo

class the current practice misses a recent upturn in the importance of family background

for class outcomes among men in the United States The conventional approach suggests

no change between cohorts but updated analyses reveal that inequality of opportunity

increased significantly for men born since the mid-1960s compared with those born

earlier in the century

AMERICAN SOCIOLOGICAL REVIEW 2009 VOL 74 (August507ndash528)

Direct all correspondence to Emily Beller(ebellerberkeleyedu) This research was support-ed by a National Science Foundation DissertationImprovement Grant (0403401) An earlier version ofthis article was presented at the 2007 meeting of theInternational Sociological Association ResearchCommittee 28 on Social Stratification and Mobility(RC28) Montreal Canada I thank Claude FischerHarry Ganzeboom Caroline Hanley Michael HoutJerome Karabel Sam Lucas Robert Mare DonaldTreiman Jane Zavisca and four anonymous review-ers for helpful comments advice and suggestions onthis research The views expressed are solely thoseof the author and do not represent the views of theUS Government Accountability Office

1 The concept of social mobility rests on the ideathat social positions exist and can be differentiatedfrom one another but there are a variety of ways to

do so Here I define social position in terms of occu-pational classes but the ideas discussed are applica-ble to other definitions (eg socioeconomic status)

Delivered by Ingenta to Universidad Politecnica de Valencia

Tue 08 Sep 2009 193307

research practice remains adequateFurthermore there are practical incentives suchas data considerations to follow the conven-tional practice For example until 1994 theGeneral Social Survey did not ask respondentsabout their mothersrsquooccupations In this articleI demonstrate empirically that fathersrsquo class isan increasingly poor proxy for family socialclass background in the United States andresearch conclusions can be distorted when it isused as such

Given practical considerations such as lim-ited data on mothersrsquo occupations it is impor-tant to emphasize that adequately defining andmeasuring class background to bring it in linewith theory is essential for substantive ratherthan simply methodological reasons (egimproving explained variance or model fit inand of itself) To illustrate this point I exam-ine changes in intergenerational class mobili-ty between recent birth cohorts in the UnitedStates and show that the typical measurementstrategy masks important findings Withoutupdating intergenerational class mobility mod-els to incorporate mothersrsquoclass characteristics

a recent significant upturn in the importanceof class background for menrsquos class destinationsin the United States is not evident The con-ventional approach indeed suggests there wasno change in the extent of mobility betweencohorts Updated analyses however reveal thatclass mobility declined for men born since the1960s compared with those born earlier in thecentury

BACKGROUND

STRUCTURAL MOBILITY AND SOCIAL

FLUIDITY

Intergenerational class mobility research ana-lyzes the strength of the association betweenindividualsrsquoclass background or childhood classposition (class origin) and their current indi-vidual or family-level class position (class des-tination) as well as patterns of immobility ormovement between particular origins and des-tinations Such intergenerational class mobili-ty has two components worth examination Thefirst often called social fluidity refers to theextent to which an individualrsquos chances of reach-

508mdashndashAMERICAN SOCIOLOGICAL REVIEW

Figure 1 Percent of Respondents with a Mother Who Worked Outside the Home by Year of Birth

Source General Social Surveys 1994 to 2006Delivered by Ingenta to Universidad Politecnica de Valencia

Tue 08 Sep 2009 193307

ing a particular class destination are constrainedby class background The second componentoften called structural mobility captures shiftsin the distribution of class origins and destina-tions that affect everyonersquos mobility regardlessof class background (eg upgrading of theeconomy toward better jobs) In the analyses thatfollow I hold structural mobility constant thuscentering attention on social fluidity Social flu-idity is generally considered indicative of equal-ity of opportunity more fluidity is thereforeldquobetterrdquo However perfect or very high socialfluiditymdashthe absence or near absence of asso-ciation between origins and destinationsmdashisneither plausible nor arguably desirable giventhat some of the processes leading to the inter-generational persistence in class position suchas inherited cognitive ability may be legitimate(Harding et al 2005 Roemer 2004)

Because there is no external benchmark suchas perfect mobility to aid in interpretation esti-mates of social fluidity mean little in and ofthemselves Yet such estimates become instruc-tive in comparative context For exampleresearchers compare the social fluidity levels ofdifferent periods or cohorts to assess whetherequality of opportunity is increasing or decreas-ing over time (eg Breen and Jonsson 2007Hout 1988) Researchers also often comparesocial fluidity levels between countries (egBreen 2004 Erikson and Goldthorpe 1992)The comparative nature of social mobilityresearch is an important reason why movingthe measurement of class background moreclosely in line with theory is a matter of morethan simply technical or methodological inter-est When measurement is biased the extent towhich it is biased can vary between the cohortscountries or other groups being compared Asa result researchers could misinterpret differ-ences in measurement error between groups assubstantive differences or a lack thereof insocial fluidity In this article I use the exampleof change in social fluidity between successivebirth cohorts about which little is known toillustrate this very concern

COMPARING SOCIAL FLUIDITY RATES OVER

TIME

Research on change over time in social fluidi-ty typically uses either a period or a cohortapproach In the period approach researchers

compare social fluidity levels between differentsurvey years (ie periods) In the cohortapproach survey data collected in different yearsor periods is pooled but respondent birth cohortis held constant Period-related shifts in mobil-ity apply to individuals across the board at agiven point in time independently of their birthcohort while cohort-related shifts in mobilitytrends arise from the different experiences ofindividuals born in specific cohorts Breen andJonsson (2007) argue convincingly that changesover time in social fluidity are more likely to becohort-driven than period-driven

Research on change over time between peri-ods and cohorts demonstrates that social fluid-ity increased over the course of the past centuryin the United States until about the mid-1980s(from a cohort perspective for individuals bornup until about 1960 DiPrete and Grusky 1990Featherman and Hauser 1978 Hout 1988)Trends in social class fluidity after the mid-1980s are unclear in part because changes to thecensus coding of occupations in the 1980s madeit impossible to directly compare new surveydata with older data (Vines and Priebe 1988)Some research suggests a possible slowing ofthe trend of increasing social fluidity after themid-1980s (Hout 1996) others predict a con-tinued trend of increasing fluidity due to a grow-ing proportion of individuals raised in non-intactfamilies who appear more mobile than theirpeers raised in intact families (Biblarz andRaftery 1999)

DEFINING SOCIAL CLASS

My own analyses of the possibilities suggestedabove employ Erikson and Goldthorpersquos (1992)class schema which is widely used in socialmobility research and often called the EGPclass schema in reference to an early explicationof it (Erikson Goldthorpe and Portocarero1979) The theoretical basis for the EGP classschema has been linked to the Weberian viewthat classes can be meaningfully differentiatedaccording to the market resources and conse-quently the life chances of their members (seeBreen 2005)2 Classes then are not defined

WHY MOTHERS MATTER IN MOBILITY RESEARCHmdashndash509

2 Erikson and Goldthorpe (199237) cite bothWeber and Marx as sources for the principles ofclass differentiation on which the EGP schema isbased


Tue 08 Sep 2009 193307

with respect to particular workplace tasks rolesor experiences per se but are defined accord-ing to the resources that are consequences ofwork Correspondingly the EGP schema definesclasses primarily in terms of the types ofemployment relationships that characterizethemmdashwith the logic that different employmentrelationships entail different rewards opportu-nities and constraintsmdashand the authors empha-size that class experiences are not restricted tothe workplace (Erikson and Goldthorpe1992236)

The EGP schema starts with a basic distinc-tion between the self-employed and employ-ees among employees a key additionaldistinction is made between two types of rela-tionships with employers Employment in a pro-fessional or service context in which employeeshave some degree of autonomy and advanta-geous resources such as employment securitycareer advancement prospects pensions andsalary increments is distinguished from employ-ment regulated by labor contracts which isunder close supervision in return for piecewages and lacks the advantages of the servicerelationship (see Goldthorpe [2000] for dis-cussion of how occupational conditions lead tothese two employment relationships)

The distinction between occupations involv-ing service versus labor-contract relationshipsis further refined in two ways First profes-sional occupations involving a service rela-tionship are divided into two classes (higher-versus lower-level professionals managers andadministrators) based on the extent of expect-ed advantageous resources Similarly occupa-tions regulated through labor contracts aredivided into two classes of skilled and unskilledmanual workers in recognition that some ben-eficial modifications to the labor contract arelikely for skilled workers Second occupationsin which the distinction between service versuslabor-contract employment relationships isblurred are also included in intermediate class-es (eg one class includes the administrativepositions that support professional bureaucra-cies another covers supervisory manual andtechnical occupations) The fully elaboratedschema includes 11 class categories that may becollapsed into fewer categories for research pur-poses (Erikson and Goldthorpe 199235ndash47)

CONCEPTUALIZING ANDMEASURING CLASS AT THE FAMILYLEVEL

Erikson and Goldthorpe as noted above argueagainst a workplace centered view of class theyalso maintain that the family rather than theindividual worker is the unit of class ldquofaterdquo(Erikson and Goldthorpe 1992233) While classexperiences follow from family membersrsquoinvolvement in different types of employmentrelationships they are not limited to the work-place they also include for example broaderexperiences of economic security or insecurityaffluence or poverty and prospects for improve-ment in economic circumstances Members ofa family who live together experience similarresources and future life chances The classposition of family members without occupa-tions including children and wives who arenot employed outside the home can thus beconsidered dependent on the resources accruedfrom the household headrsquos employment rela-tionship In addition spouses in dual-earnerfamilies whose individual occupations may besubstantially different in terms of employmentrelationships also share one class positionbecause of their shared experiences of resourcesor constraints While some argue that the ideaof a shared family-level class position should beabandoned in favor of measuring class in termsof individual occupations (Acker 1973Stanworth 1984) the individual approach is notpossible when applied to class origins (given thatchildren do not have occupations)

DEBATE OVER JOINT VERSUS

CONVENTIONAL MEASUREMENT OF SHARED

FAMILY CLASS POSITION

Scholars who agree theoretically that familiesshare both class experiences and a class posi-tion still debate how shared family class shouldbe measured (Sorensen 1994) Goldthorpe(1983 1984) initially argued that the sharedfamily-level class position is determined by afatherrsquos or husbandrsquos class (this was termed theconventional view of family class) This posi-tion sparked debate with others arguing for ajoint approach to measuring the shared familyclass position in which both spouses ifemployed contribute to family class (Brittenand Heath 1983 Davis and Robinson 19881998 Heath and Britten 1984) Research offers



Tue 08 Sep 2009 193307

mixed support for both the conventional andjoint viewpoints leading to a conclusion that theconventional measurement of class may notstrongly distort research results (Sorensen1994) Conventional scholars remain in favor ofmeasuring the family-level class with respect toonly one spouse (rather than jointly based on twospousesrsquo occupations) but allow that the ldquodom-inantrdquo class positionmdashthat of the spouse withthe stronger labor force attachment and higherindividual class positionmdashcould potentiallydetermine family class position rather thannecessarily the husband (Erikson 1984 Eriksonand Goldthorpe 1992) They argue that the jointapproach to determining the family-level posi-tion while attractive in principle may blur classboundaries and create too many possible classpositions (Erikson and Goldthorpe 1992238)3

Importantly conventional practice has morepotential to distort social mobility research con-clusions than is clear from the previous debatewhich had important limitations Core ques-tions addressed in the debate regarding family-level social class include (1) how to measurethe proportion of mixed-class families to deter-mine the significance of the problem they mightpose to researchers and (2) how mixed-classfamilies might affect substantive researchresults particularly given findings that the sub-jective class identification class related behav-ior (eg voting) and life chances of marriedwomen can be better predicted by their hus-bandsrsquo than by their own occupations (Baxter1994 Erikson and Goldthorpe 1992Goldthorpe 1983 1984 Heath and Britten 1984Stanworth 1984)

One limitation of the prior debate is that itfocused on adultsrsquo class positions without alsoconsidering the position of childrenmdashdespite thefact that intergenerational class mobilityresearch is centrally concerned with the influ-ence of family class position on childrenrsquos futurelife chances For example as Erikson andGoldthorpe (1992250 note 16) noteresearchers critiquing the conventional view ofclass destinations nonetheless rely on conven-tional measures of class origins Given a focuson how class affects future life chances theadequacy of different approaches to measuringthe shared family class position must be evalu-

ated from the perspective of children as well asadults

Another limitation of the prior debate is thatit focused on families in which adults wereemployed in different classes (ldquomixed-classrdquofamilies) over class-consistent families or fam-ilies with only one spouse in the labor market(single-earner families) Families with only oneemployed spouse or with both spousesemployed in the same class were not consideredproblematic Furthermore the class position ofa dual-earner family with both spousesemployed in a particular class was consideredequivalent to the position of a single-earnerfamily with one spouse employed in that classSorensen (199443) characterizes this assump-tion as surprising noting that a major reason fordeveloping new measures of familiesrsquo classpositions is the hypothesis that womenrsquos employ-ment makes a difference for familiesrsquo materialcircumstances and life chances That is if thejoint perspective is correct it should logicallyapply not only to mixed-class but also to class-consistent families For example if individualspouse class characteristics jointly define fam-ily class position dual-earner families whereboth adults are employed in the higher profes-sional class might be expected to have a moreadvantaged class position than would familieswith one spouse employed in the higher pro-fessional class and another spouse who is eithernot in the labor market or is employed in a lessadvantaged occupation The idea that eachspousersquos employment relationships could pro-duce cumulative class resources or constraintsilluminates the implications of the convention-al versus joint measurement

THEORETICAL IMPLICATIONS OF

CONVENTIONAL AND JOINT MEASUREMENT

OF FAMILY CLASS ORIGIN

Breen and Jonsson (2007) propose a theoreti-cal model of social mobility in which arrival ata particular class destination depends on class-related parental resources that can be eitherdirectly (eg genetics or property) or indirect-ly transmitted between generations The role ofindirect transmission in this process reflects theidea that parentsrsquo class experiences and conse-quent class-related resources influence theextent to which the next generation can accu-mulate assets such as higher education which


3 See Breen and Rottman (1995) and Sorensen(1994) for more detailed reviews of this debate


Tue 08 Sep 2009 193307

in turn generate particular returns in class des-tinations This model helps clarify the theoret-ical implications of conventional and jointmeasurement of the shared family class positionwith respect to class origin in particular Jointmeasurement of family class origin accounts forthe possibility that each parentrsquos employmentrelationships result in class-related resourcesand assets that may accumulatemdashregardless ofwhether the parentsrsquo occupations fall into dif-ferent classes or are class-consistent By con-trast the conventional measurement practiceassumes that net of the key (fatherrsquos or higher)class position a second parentrsquos employmentdoes not result in additional class-relatedresources If this assumption is incorrect con-ventionally measured class position could beunderstood to serve as a proxy for a more com-plex set of family class resources that are prod-ucts of both the measured class position and anunmeasured second parentrsquos class position

EVIDENCE OF A CUMULATIVE IMPACT OF

PARENTSrsquo CLASS-RELATED RESOURCES

Theorized mechanisms of the intergenerationalclass transmission process are consistent withthe joint view of family class origin Class-related economic resources clearly accumulateand play a key role in indirect and direct trans-mission of resources and assets (Conley 2001Hill and Duncan 1987) but economic resourcesare only part of the story (Mayer 1997) Class-related noneconomic resources such as occu-pational prestige and parent education (oftentermed cultural resources) might also play a rolein indirect transmission Individual parent cul-tural resources could accumulate many theorizethat advantaged parents provide children withadvantageous cultural resources through inter-active processes (Bourdieu and Passeron 1977Lareau 2003) For example middle- and upper-class parents may intentionally cultivate chil-drenrsquos social skills such as addressing andnegotiating with authority figures (Lareau2003) have greater knowledge of educationalbureaucracies (Deil-Amen and Rosenbaum2003 Lareau 1989 Lucas 1999) and hold highaspirations for their children (Hauser Tsai andSewell 1983 Sewell Haller and Portes 1969)The impact of cultural resources could add upparticularly if both parents spend time withtheir children

Prior research indeed provides empirical evi-dence that both parentsrsquo class characteristicsinfluence childrenrsquos class-related resourcesassets and eventual class destinations even ifthe parentsrsquo class characteristics are the sameAmong employed parents for instance bothparentsrsquo occupations independently shape chil-drenrsquos educational outcomes (Kalmijn 1994Korupp Ganzeboom and Van Der Lippe 2002)just as both parentsrsquo education levels do (Mare1981) Models of occupational mobility betterpredict class destinations for both sexes(Khazzoom 1997) when the models includemothersrsquo occupations Although many mothersdo not have occupations outside the home thetheorized role of parentndashchild interactions inthe intergenerational transmission process rais-es the question of whether the joint view offamily class origin might apply even to single-earner familiesmdashthat is homemaker mothersmay contribute their own class resources despitenot having an individual employment-basedclass position Dynamic views of class(Marshall Roberts and Burgoyne 1996 Plutzerand Zipp 2001) posit that a series of experiencessuch as childhood class background educationand previous employment or unemploymentspells contribute to onersquos class and associatednoneconomic class resources In either caseand given the theorized role of noneconomic orcultural class resources in the mobility processit is reasonable to test whether nonemployedparents may contribute to the transmission ofclass-related resources rather than assume theydo not

IMPLICATIONS FOR INTERGENERATIONAL

SOCIAL MOBILITY RESEARCH

The theoretical and empirical evidencedescribed above suggests that parental classresources may jointly determine family class ori-gin Because the prior debate over the joint ver-sus conventional measurement strategies didnot problematize class-consistent familiesassortative marriage patterns (which result in ahigh prevalence of class-consistent families)may appear to justify conventional measure-ment Conventional practice however wouldactually be less problematic if marriage wererandom with respect to class class-based assor-tative marriage patterns mean that the meas-urement error produced by the conventional



Tue 08 Sep 2009 193307

practice is not random it probably changes overtime and differs among groups

If parent class characteristics and associatedclass resources indeed jointly determine thefamily class position conventional estimatesof the strength of fatherndashchild association inclass position will include the correlated butunmeasured effects of mothersrsquoclass resourceson the process Because the correlation betweenmothersrsquoand fathersrsquo individual class positionsis not perfect the conventional measurement offamily class origin will underestimate the totaloriginndashdestination association Furthermore ifmarital sorting by class differs between com-parison groups (eg nations cohorts orracialethnic groups) the use of conventionalorigins measures could lead analysts to erro-neously interpret changes in the degree of meas-urement error to be substantive differencesbetween groups in social fluidity levels

ANALYTIC STRATEGY

DATA AND METHODS

The data from which I draw are a compilationof the available years of the General SocialSurvey (GSS) that include mothersrsquo occupa-tional datamdash1994 1996 1998 2000 20022004 and 2006 (Davis Smith and Marsden2007) Occupational data collected in these sur-vey years were recorded in 1980 basis censuscodes I conduct separate analyses for men andwomen and I restrict analyses to respondentswho were ages 25 to 64 and in the labor forceat the time of the survey I also restrict the analy-ses to respondents with valid data for both theirown and two parentsrsquo (or parental figuresrsquo)occupations or in the case of mothers home-maker status Missing data for parent occupa-tion due to item nonresponse is minimalaffecting approximately 25 percent of the oth-erwise eligible sample However the GSS likemost surveys does not ask about the occupa-tions of noncustodial parents Therefore with-out making the strong assumption thatnoncustodial parents do not shape class back-ground respondents who lived with a singleparent (or in an institution in which case no par-ent occupation data was collected) at age 16 can-not be included in analyses that focus oncomparing two-parent versus one-parent meas-ures of family class position To partially addressthis limitation I include single-parent families

in the final portion of the analysis which exam-ines change over time in social fluidity ratherthan comparing one-parent versus two-parentmeasures of class origins

I adopt a six-category version of the Eriksonand Goldthorpe class schema described aboveto define classes I generated the EPG classeson the basis of the GSS International StandardClassification of Occupations (ISCO) 88 codesfor occupations together with self-employmentinformation from a widely used conversionalgorithm (Ganzeboom and Trieman 2003)4

The class categories are the following5

I Professionals administrators officials andmanagers higher level

II Professionals administrators officials man-agers lower level

IIIab Routine nonmanual and service workers high-er and lower levels

IVab Self-employed with or without employeesVVI Technical specialists and supervisors of man-

ual workers skilled manual workers andVIIab Semiskilled and unskilled manual workers

nonfarm and farm

Separately for men and women6 I organizethese data into a three-way intergenerationalclass mobility table by cross-classifying themotherrsquos class category variable by the fatherrsquosclass category variable by the respondentrsquos classcategory variable (Table A1 in the Appendixshows the distributions of class positions among


4 The GSS data from 1994 through 2006 includedetailed occupational information recorded in 1980basis US census codes The GSS also converts these1980 basis census codes into ISCO 88 codes Analgorithm (unpublished available from the author) todirectly convert occupational data recorded in 1980sbasis census codes into EGP classes produces simi-lar results Erikson and Goldthorpe (1992315ndash16)encountered some problems applying their schemato United States occupational data recorded in 1960basis census codes but the 1980 basis census codesdo not present these problems

5 Apart from a small number of farm workers(class VIIb) I exclude agricultural classes due tolimited data Some research combines class IIIb withclass VII instead of IIIa this alternate categorizationdoes not change the key results

6 A study of the validity of the class schema reportsit to be valid for both men and women (Evans andMills 1998)


Tue 08 Sep 2009 193307

male and female respondents of the appropri-ate age and labor force status and those of theirmothers and fathers)

I use Goodmanrsquos (1979) log-multiplicativeRC association model (also called the RC-IImodel) to analyze the mobility tables describedabove To illustrate the RC model consider asimpler two-way contingency table such as theconventional intergenerational mobility tableof fatherrsquos class (i) by class destination (j) TheRC model simultaneously estimates row scores(i) that rank fatherrsquos class (origin) categoriesand column scores (j) that rank class destina-tion categories along with an intrinsic associ-ation parameter () The association parameterconveys the overall strength of the relationshipbetween the ranked class origin and destina-tion categories and it is interpreted similarly toa regression coefficient in that a larger valuemeans greater association (Hout 1983)7

While typically used to analyze grouped datasuch as mobility tables the RC model can beextended to incorporate individual-level covari-ates using various techniques such as includingstereotype ordered regression (SOR) parameters(Breen 1994 DiPrete 1990 Hendrickx andGanzeboom 1998) I incorporate SOR param-eters in RC models in some analyses to controlfor age which becomes important in models that

compare fluidity between cohorts The SORparameter is analogous to the RC associationparameter () in that it parsimoniously express-es the overall effect of an independent variablek on all categories of the dependent variable ina single parametermdashwhere the RC associationparameter indexes the strength of associationbetween i and j the SOR parameter indexesthe strength and direction of association betweenk and j

The mobility tables I analyze are somewhatsparse due primarily to clustering of women incertain classes so I assess overall model fitusing the Pearson chi-squared goodness of fitstatistic X2 rather than the likelihood-ratio good-ness of fit statistic L2 but I compare nestedmodels using L2 (Agresti and Yang 1987)8 Toincorporate the GSS case weight variable with-out distorting these model fit statistics thecounts in the mobility tables are the unweight-ed frequencies and the models include weightvectors containing average cell weights (Cloggand Eliason 1987) X2 and L2 are appropriate fitstatistics given grouped data but they are notapplicable to the models with SOR parametersgiven that such models include individual-leveldata I also use the BIC criterion (Raftery 1995)in concert with the other fit statistics as appli-cable to adjudicate among models Given thesample sizes marginal differences in BIC (offewer than approximately 10 points) can beconsidered equivalent (Wong 1994) I use theLEM program (Vermunt 1997) for the analysesof grouped data and Stata for the SOR analy-sis

MEASURES OF CLASS ORIGIN

I analyze several different measures of class ori-gin that fall into three categories those based on


7 I use log-multiplicative RC association modelsrather than log-linear models because the associationparameter of the RC model is readily interpretableas a descriptor of the overall strength of associationbetween class origins and destinations This featureof the RC model is key to illustrating the conse-quences of various origins measures One limitationof the RC model however is that in summarizing ori-gin and destination categories in terms of rankedscores it analyzes only one hierarchical dimensionof origin-destination association (multidimensionalRC(m) models are possible but are not as easilyinterpretable) I replicated the analyses presented inthis article using both log-linear models which do notimpose a unitary hierarchical dimension of associa-tion and Erikson and Goldthorpersquos (1992) core socialfluidity model which includes multiple nonhierar-chical and hierarchical log-linear parameters todescribe origin-destination association with sub-stantively similar results (see Tables S1 and S2 in theOnline Supplement on the ASR Web sitehttpwww2asanetorgjournalsasr2009toc070html)

8 There are 15 to 22 cells that contain samplingzeros for men and 22 to 36 cells for women in theanalyses of grouped data (the number of samplingzeros and the total number of cells varies dependingon whether and how homemaker mothers are includ-ed the total number of cells in the mobility tablesranges from 216 to 324) There are no zero marginsTo detect potential problems due to sparseness Iexamined the standard errors of the log-linear param-eters none are unusually large I did not add a con-stant (eg 5) to the cell counts


Tue 08 Sep 2009 193307

only one parentrsquos class position (these includeconventional father-only and dominance meas-ures of class origins) joint measures of class ori-gin determined by both parentsrsquo individual classpositions and joint measures of class origin deter-mined by both parentsrsquo individual class positionsthat also include interaction effects between par-

ent gender and parent class position Table 1 pre-sents equations for RC association models that fitthe partial association between class destinationand various measures of class origin net of dummyvariables for ldquodiagonalrdquo immobility effects Theseimmobility parameters capture respondentsrsquo ten-dency to cluster along the diagonal cells of the


Table 1 Equations RC Association Models with Various Measures of Class Origin

One-Parent Measures

01 Father-Only a Log Fhij = 0 + hM + i

F + hiMF + j

D + 1Dij + uij

02 Mother-Only b Log Fhij = 0 + hM + i

F + hiMF + j

D + 2Dhj + uhj

03 Higher Class Dominance c e Log Fhij = 0 + hM + i

F + hiMF + j

D + 1Dhij + uhij

04 Lower Class Dominance d e Log Fhij = 0 + hM + i

F + hiMF + j

D + 2Dhij + uhij

Joint-Parent Measures

05 Mother + Father a b f Log Fhij = 0 + hM + i

F + hiMF + j

D + 1Dij + 2Dhj + uhij

06 Higher Class + Lower Class c d e g Log Fhij = 0 + hM + i

F + hiMF + j

D + 1Dhij + 2Dhij + uhij

07 Equal Mother + Father a b h Log Fhij = 0 + hM + i

F + hiMF + j


08 Equal Higher + Lower Class c d i Log Fhij = 0 + hM + i

F + hiMF + j


Joint-Parent Measures with Interactions

09 Full Interaction a b e Log Fhij = 0 + hM + i

F + hiMF + j


10 Class Interaction a b e Log Fhij = 0 + hM + i

F + hiMF + j


Extensions of Selected Equations to Control for Cohort and Age

11 Father-Only a j Log Fghij = 0 + gC + h

M + iF + j

D + ghCM + gi

CF + hiMF + ghi

CMF +

gjCD + 1cDij+ j (0ui + 1(cohort)ui + 2(age)ui + 3(age2)ui + B1(age) +

B2(age2))

12 Mother + Father a b f j Log Fghij = 0 + gC + h

M + iF + j

D + ghCM + gi

CF + hiMF + ghi

CMF +

gjCD + 1cDij + 2cDhj+ 3cDij + j (0uhi + 1(cohort)uhi + 2(age)uhi +

3(age2)uhi + B1(age) + B2(age2))

Notes h indexes motherrsquos class (M) i indexes fatherrsquos class (F) j indexes destination class (D) and g indexesbirth cohort (C) For identification parameters sum to zero in all models in all models as applicable origin anddestination scores are identified using the constraints iui = huh = hiuhi = jj = 0 and iui

2 = huh2 = hiuhi2 =

jj2 = 1

a where 1Dij = 1 if i = j 0 otherwiseb where 2Dhj = 1 if h = j 0 otherwisec where 1Dhij = 1 if the higher of h or i = j 0 otherwised where 2Dhij = 1 if the lower of h or i = j 0 otherwisee where origin scores uhi are constrained as shown in the Online Supplement Table S3f where uhi = ui + uhg where uhi = uhi as defined in equation 3 + uhi as defined in equation 4h where uhi = mean ui uh from equation 5i where uhi = mean uhi as defined in equation 3 uhi as defined in equation 4j where 1cDij and 2cDhj are cohort-specific 1Dij and 2Dhj 3cDij is single-earner family cohort-specific 1Dij0 is the baseline association between ui or uhi and j 1 gives the impact of birth cohort on the association 2

+ 3 give the impact of age on the association and B1 + B2 give the association between age and class destina-tion


Tue 08 Sep 2009 193307

mobility table where origin = destination over andabove what RC models would otherwise predict9

The one-parent measures of class origin arethe conventional father-only model a mother-only model a higher class dominance model inwhich the higher class position solely definesthe family class position regardless of parentgender (Erikson 1984) and a lower class dom-inance model in which the lower class positionsolely defines the family class10 The higherand lower class dominance models constrainthe origin scores u such that the scores are basedon the parent with the higher or lower classposition (Table S3 in the Online Supplementshows the design matrices for these constraints)

The joint-parent measures of class originsinclude a mother + father model and a lower +higher class model (the latter differentiates par-ents by relative individual class position ratherthan by gender [Korupp et al 2002]) Thesemodels combine the mother-only and father-only models and the higher and lower classdominance models described above In addi-tion two related models the equal mother +father and equal lower + higher class models areconstrained versions of the former models theyspecify that the effects of the fatherrsquos and moth-errsquos class or the higher and lower class are heldequal in determining the joint family position

Finally two additional models jointly meas-ure class origin with respect to both parentsrsquoindividual class positions and also include par-ent interaction effects The first is a full inter-action model that allows each combination ofparent class and gender to result in a uniqueclass origin category The second is the class

interaction model which is a constrained ver-sion of the full interaction modelmdashit includesinteraction effects between pairs of parent classpositions but not between parent class posi-tion and parent gender In other words in theclass interaction model the origin score for afamily with a professional mother and a self-employed father is held equal to the score for afamily with a professional father and a self-employed mother (see Table S3 in the OnlineSupplement for the design matrix for the con-straints) By contrast in the full interactionmodel the scores for these two combinations ofparent class positions are free to differ

ANALYSIS STEPS

ADJUDICATING BETWEEN CLASS ORIGIN MEAS-URES The first component of the analysis adju-dicates between all the measures of class origindescribed above using the RC association mod-els shown in Table 1 for men and women whowere raised in dual-earner families (ie respon-dents who reported both a motherrsquos and a fatherrsquosoccupation) I begin by assessing whether mod-els that use the conventional father-only classorigin measure adequately fit the motherrsquos classby fatherrsquos class by class destination mobilitytable If conventional measures do not providean adequate fit the next question is whetherjoint-parent measures provide any improvementin overall fit and if so which joint-parent meas-ure is preferable (Breen [200547ndash49] suggeststhis type of empirical strategy) Finally I eval-uate whether the use of different origins meas-ures affects substantive analytical resultsmdashmostimportantly the estimated degree of social flu-idity as indexed by the association betweenclass destinations and each measure of classorigins11

CAN HOMEMAKER MOTHERS BE INCORPORATED

INTO JOINT-PARENT MEASURES OF CLASS ORIGINIn the second step of the analysis I evaluatewhether including respondents with homemakermothers about one-third of all respondentschanges the results I examine whether the meas-


9 The inclusion of diagonal immobility parametersis standard when RC models are used to analyzemobility tables (eg Gerber and Hout 2004Goodman and Clogg 1992) but their inclusion affectsthe interpretation of the association parameters whichno longer index the total origin-destination associa-tion I estimated all models without immobilityparameters with the same results in regard to relativemodel fit however none of the models that lackcontrols for fatherndashson diagonal immobility fit themobility data well among men

10 Ideally the dominance models should also takeinto account which parent has the more enduringattachment to the labor force While doing so mightimprove the performance of these measures it is notpossible in this analysis

11 I standardized the association parameters sothat they are comparable across models with differ-ent numbers of origin categories (Clogg and Shihadeh199451)


Tue 08 Sep 2009 193307

ures of class origin that were preferred forrespondents raised in dual-earner familiesremain preferred for a larger population thatincludes respondents raised in single-earnerfamilies (ie with homemaker mothers) Theequations are the same as those used in theprior analysis and shown in Table 1 but with amodification to the immobility parametersmdashthefather-respondent immobility parameter 1 ispermitted to vary in strength for respondentsraised in dual-earner versus single-earner fam-ilies

I evaluate three alternative ways of incorpo-rating respondents with homemaker mothersFirst I employ a hybrid of the father-only modeland the mother + father model In this hybridmodel the joint mother + father approachapplies only to respondents with employedmothers while the fatherrsquos class alone definesclass origin for respondents with homemakermothers In other words a motherrsquos positionenters the model only for mothers with an occu-pation the effect of homemaker mothers is setequal to zero Second I fit a model that speci-fies that homemaker mothers make up a uniformclass category that may shape class origin ThirdI fit a model that specifies that homemakermothers may influence class origin differentlydepending on their class resources as indexedby their education levels

CHANGE IN SOCIAL FLUIDITY BETWEEN SUC-CESSIVE BIRTH COHORTS In the final componentof the analysis I evaluate the importance ofusing adequate class origin measures forresearch conclusions about trends in social flu-idity I use a cohort perspective to analyzechange over time comparing social fluidity lev-els for men and women born in 1945 to 19541955 to 1964 and 1964 to 197912 To control fordifferences in the age distributions of the sur-vey respondents between cohorts I extend theRC models used previously by including SORparameters for age and age-squared The equa-

tions for these extensions of the key modelsused in the first two steps of the analysis areshown in Table 1 and specify that net of cohort-specific immobility parameters and cohort dif-ferences in the distribution of class origins anddestinations (eg cohort change in structuralmobility) both class origin and respondent agemay affect class destination Furthermore thestrength of the association between class originand destination may vary by birth cohort andage

The cohort-specific analyses include respon-dents with homemaker mothers and use meth-ods for incorporating homemaker mothers thatthe second step of the analysis demonstrateswork best for each gender for men homemakermothers make up one class category for womenhomemaker mothers are differentiated by edu-cation level Additionally to examine cohortchange in social fluidity for a more compre-hensive population the final models in this sec-tion also include respondents raised insingle-parent families Given the problem ofmissing data for noncustodial parent occupationone of the more straightforward ways to includenoncustodial parents in class origin measures isto add ldquoclassrdquo categories representing noncus-todial mothers and fathers Instead of six indi-vidual class positions for fathers for examplethere are now sevenmdashone of which capturesnoncustodial fathers (whose occupational classpositions are not known) The model thereforeestimates an origin score representing the impactof having a noncustodial parent alongside scoresfor having parents of particular class positionsthis accounts for the average impact of non-custodial parentsrsquo positions

RESULTS

EVALUATING CONVENTIONAL VERSUS

JOINT-PARENT CLASS ORIGIN MEASURES

In prior debates about joint versus convention-al measurement of class proponents of the con-ventional approach argued that empiricalsupport for the joint approach to measuringfamily class must go beyond demonstrating asignificant net impact of wivesrsquo individual classon the family class position making an empir-ically compelling case for the joint view requiresdemonstrating that conventional measurementis empirically inadequate and that joint meas-urement can change the substance of research


12 I exclude respondents from survey years priorto 1994 when the GSS began to collect data onmothersrsquo occupations This restriction prevents theinclusion of respondents born during the first half ofthe twentieth century I define the youngest cohortmore broadly than the others to obtain a comparablesample size


Tue 08 Sep 2009 193307

findings (see Sorensen 1994) The results ofthe first step of the analysis demonstrate thatboth of these standards are met in the case offamily class origin

Table 2 shows the results of mobility modelsemploying the various one-parent and joint-parent class origin measures13 The model sig-nificance statistics (Model p) illustrate whichmodels best summarize the actual patterns

observed in the data (if Model p is significantat p lt 05 the modelsrsquo predictions differ sig-nificantly from the observed mobility data andmodel fit is not considered adequate) None ofthe models employing one-parent class originmeasures (Models 7 through 10 in Table 2)accurately summarize observed mobility pat-terns apart from the lower-class dominancemodel among women By contrast all of themodels using joint-parent measures (Models 1through 6 in Table 2) do accurately account forthe patterns observed in the data These resultsindicate that each parentrsquos class resources shapefamily class origin and using one parentrsquos classposition as a proxy for the family-level classposition is not an empirically adequateapproach

The next question then is which of the mul-tiple joint-parent models all of which ade-quately summarize observed mobility are the


Table 2 Fit Statistics and Parameters for RC Association Models with Various One-Parent andJoint-Parent Class Origin Measures

p p Model Model Description X2 L2 df BIC vs 1 vs 9 p 1 2

Men

01 Full Interaction 1595 1662 134 ndash8913 0 07 26 31 0302 Class Interaction 1653 1723 149 ndash10036 98 0 17 23 31 0303 Mother + Father 1778 1860 159 ndash10689 76 0 15 21 31 0304 Higher Class + Lower Class 1892 1934 159 ndash10615 35 0 05 22 17 2205 Equal Mother + Father 1782 1861 164 ndash11082 92 0 21 21 30 0306 Equal Higher + Lower Class 1937 1989 164 ndash10954 34 0 06 20 18 2107 Higher Class Dominance 2652 2610 165 ndash10412 0 0 22 1808 Lower Class Dominance 2439 2428 165 ndash10594 0 0 21 2509 Father-Only 2280 2319 165 ndash10703 0 0 18 3010 Mother-Only 3649 3540 165 ndash9482 0 0 23 0511 Quasi-independence 3096 3194 174 ndash10539 0 0 6012 Independence 4989 4935 175 ndash8876 0 0

Women

01 Full Interaction 1349 1455 134 ndash9192 0 46 29 02 0702 Class Interaction 1464 1586 149 ndash10254 60 0 54 25 02 1003 Mother + Father 1716 1836 159 ndash10798 05 0 23 20 03 0804 Higher Class + Lower Class 1671 1767 159 ndash10867 18 0 31 21 02 1505 Equal Mother + Father 1716 1837 164 ndash11195 15 0 33 20 03 0806 Equal Higher + Lower Class 1721 1847 164 ndash11184 12 0 32 18 03 2107 Higher Class Dominance 2123 2175 165 ndash10936 0 01 18 ndash0108 Lower Class Dominance 1865 1964 165 ndash11147 01 12 19 2109 Father-Only 2203 2259 165 ndash10851 0 0 19 0510 Mother-Only 2108 2233 165 ndash10878 0 01 20 0311 Quasi-independence 2834 2935 174 ndash10891 0 0 2412 Independence 3025 3106 175 ndash10799 0 0

Notes N = 2676 (men) N = 2824 (women) is the association parameter and 1 and 2 are model-specificimmobility parameters defined in Table 1

13 Among the one-parent measures the higher-class dominance model suggested as an improvementover the conventional father-only model has a poor-er overall fit than the father-only model among menand notably a poorer fit than the lower-class domi-nance model among both men and women Thefather-only model is strongly preferred over the moth-er-only model for men but the two are equivalent forwomen


Tue 08 Sep 2009 193307

best While L2 comparisons show that all of themodels that employ joint-parent class originmeasures (Models 1 through 6 in Table 2) pro-vide a significantly better account of mobilitythan does the father-only model the full andclass interaction models are needlessly com-plex compared with more parsimonious joint-parent models such as the mother + fathermodels (the BIC criterion which emphasizesmodel parsimony rejects the full and class inter-action models compared with the father-onlymodel) Both the higher + lower class and moth-er + father models remain as appropriate joint-parent measures of class origin for theremainder of the study I focus on the mother +father conceptualization because it is morestraightforward than the higher + lower classmeasure

Not only do conventional origin measuresfail to serve as an adequate proxy for family-class position from the perspective of accu-rately summarizing observed mobility patternsbut the results of this analysis also illustratehow their use could distort the substance ofmobility research findings Generally mobili-ty studies address questions about the extent ofequality of opportunity in a society by examin-ing the estimated extent of social fluidity orclass immobility The parameters in Table 2show that the extent of class immobility esti-mated by the father-only model approximatesthat of the mother + father models14 Turning toestimated social fluidity on the other hand the parameters in Table 2 for the father-onlymodel are smaller than the parameters esti-mated by the joint-parent models from the samedata15 This is particularly true for men The esti-mated association between class origins anddestinations is 15 percent higher among men and5 percent higher among women given either ofthe mother + father models compared with theconventional model16

That the different models estimate somewhatdifferent levels of association given identicaldata supports the possibility further evaluatedbelow that when both parentsrsquo resources deter-mine family class position but it is measuredwith respect to fathers only measurement errordue to ignoring mothersrsquo resources coulddepress the estimated intergenerational associ-ation in class position potentially distortingresearch findings about social fluidity levelsThis is of particular concern because the extentof the measurement error and consequent dis-tortion of estimated social fluidity likely variesbetween comparison groups such as nationsor cohorts

CAN HOMEMAKER MOTHERS BE

INCORPORATED INTO JOINT-PARENT


For simplicity the first step of the analysisexcludes respondents who reported that theirmothers worked in the home rather than in thelabor force (ie 33 percent of otherwise eligi-ble respondents) The results of this second stepof the analysis shown in Table 3 demonstratethat the preferred joint-parent measures from theprior analysis can be extended to a populationthat includes single-earner families The joint-parent measures continue to provide a moreaccurate depiction of observed mobility in thislarger population than do conventional classorigin measures

First it is important to reevaluate the con-ventional father-only model in the context of thecurrent sample Despite the fact that one-thirdof the sample reported that their fathers were theonly employed parent the father-only model(Model A1 in Table 3) does not adequately sum-marize observed mobility patterns By com-parison the hybrid of the father-only modeland the mother + father model (where the jointmother + father approach applies only torespondents with employed mothers while thefatherrsquos class alone defines class origin forrespondents with homemaker mothers) pro-vides a significantly better account of mobili-ty compared with the poorly fitting father-only


14 This is not the case when comparing higherclass dominance with the higher + lower class modelthe latter estimates nearly twice as much class immo-bility as the former

15 LEM does not provide standard errors for the parameters that the association is significantly high-er than 0 is demonstrated by the rejection of the inde-pendence model (Model 12 in Table 2)

16 The father-only model also underestimates theassociation compared with the mother-only model

Note however that the mother-only model does notinclude the father-respondent immobility parame-ter which accounts for the difference

Font andsize arecorrectas-is


Tue 08 Sep 2009 193307

model However while it comes close thishybrid mother + father model (Model A2 inTable 3) still does not quite fit the mobilitytable (although among men the more parsimo-nious hybrid equal mother + father model[Model A3] does fit)

Considering that the mother + father modeladequately depicts mobility when the analysisis restricted to respondents raised in dual-earn-er families this change for the worse in the fitof the model to the data suggests that home-maker mothers may indeed shape family classorigin To further investigate this idea ModelsA4 and A5 in Table 3 include homemaker asa possible class category for mothers Amongmen including the single broad class of home-maker mothers adds little to the picture ofclass originmdashthe effect of the homemakermother class category is very close to zeroeven though it was not constrained to zero asin Models A2 and A3 Among women how-

ever the models that include a class of home-maker mothers (Models A4 and A5) are mar-ginally preferred over models that set theeffects of homemaker mothers to zero (ModelsA2 and A3)

One interpretation of the finding that mod-els ignoring homemaker mothers fit the mobil-ity table equally well among men and almostas well among women as models that includea class category for homemaker mothers is thatthe joint view of family class may apply only todual-earner families Another possible inter-pretation however is that the measurementerror inherent in the single broad category ofhomemakers with diverse class resources coulddepress the observable effects of such resourcestoward zero The second interpretation is con-sistent with the results of the models in termsof overall fitmdashnone of the models summarizethe mobility data as well as the same models inthe previous analysis which was restricted to


Table 3 Methods for Including Homemaker Mothers

p ModelModel Description X2 L2 df BIC vs1 p 1a

1b2

Men

A Treating the Homemaker Mother Class Category as a Single Categorymdash1 Father-Only 2775 2825 195 ndash13380 0 19 32mdash2 Employed Mother + Father 2266 2350 188 ndash13274 0 03 21 25 46 05mdash3 Equal Employed Mother + Father 2271 2363 193 ndash13676 0 05 22 27 48 02mdash4 Mother + Father 2265 2350 187 ndash13191 0 03 21 25 46 05mdash5 Equal Mother + Father 2272 2364 192 ndash13592 0 04 22 27 49 02B Differentiating the Homemaker Mother Class Category by Education Levelmdash1 Father-Only 3919 4049 255 ndash17142 0 19 32mdash2 Mother + Father 2949 3067 245 ndash17293 0 02 24 29 42 04mdash3 Equal Mother + Father 2967 3080 251 ndash17779 0 03 23 27 40 05

Women

A Treating the Homemaker Mother Class Category as a Single Categorymdash1 Father-Only 2791 2847 195 ndash13386 0 20 05mdash2 Employed Mother + Father 2321 2408 188 ndash13242 0 02 21 ndash05 27 02mdash3 Equal Employed Mother + Father 2331 2426 193 ndash13640 0 03 21 ndash04 30 01mdash4 Mother + Father 2264 2343 187 ndash13223 0 03 21 ndash05 26 04mdash5 Equal Mother + Father 2277 2372 192 ndash13610 0 04 21 ndash03 30 03B Differentiating the Homemaker Mother Class Category by Education Levelmdash1 Father-Only 3616 3764 255 ndash17463 0 20 05mdash2 Mother + Father 2654 2827 245 ndash17568 0 18 27 ndash02 17 03mdash3 Equal Mother + Father 2710 2870 251 ndash18024 0 18 25 ndash05 15 04

Notes N = 4066 (men) N = 4123 (women) is the association parameter In Model 1 1a represents father-respondent immobility In Models 2 to 5 1a represents father-respondent immobility in dual-earner families and1b represents father-respondent immobility in single-earner families (ie families with homemaker mothers) 2

represents mother-respondent immobility in all models See Table S4 in the Online Supplement for class originand destination scores for selected models


Tue 08 Sep 2009 193307

respondents raised in dual-earner families Thissuggests that homemaker mothers shape classorigin but in a fashion that is not well measuredby the inclusion of a single homemaker classcategory A second series of models (B in Table3) therefore differentiates the homemaker cat-egory by education level as a proxy for home-maker mothersrsquo class resources Among mendifferentiating the homemaker mother class cat-egory with respect to education still does notadequately describe mobility patterns Amongwomen on the other hand the joint-parent mod-els in which homemaker mothers are differen-tiated by education level (Models B2 and B3 inTable 3) fit the mobility table well overallMothersrsquononemployment based class resourcesmay be more salient for women than for men17

For both men and women the results confirmthat the substantive findings of the previous analy-sis continue to apply when respondents with home-maker mothers are included in the mobility dataset Joint-parent models provide a substantiallybetter summary of observed mobility than does theconventional model although the diversity with-in the homemaker category may be problematicwith respect to model fit The findings also recon-firm those in the previous analysis that the father-only model produces a lower estimate ofintergenerational association in class positionFinally in this step of the analysis the inclusionof mothersrsquo characteristics has implications forthe estimated immobility between fatherrsquos class andclass destination Both fatherndashson andfatherndashdaughter class immobility is stronger insingle-earner families than in dual-earner familiesThis weaker fatherndashrespondent immobility whenmothers are employed outside the home furtherillustrates how mothersrsquo employment matters forintergenerational mobility

CHANGE IN SOCIAL FLUIDITY BETWEEN

SUCCESSIVE BIRTH COHORTS

The results thus far demonstrate that class ori-gin measures that jointly capture both parentsrsquo

class positions provide the most accurate pictureof intergenerational mobility Furthermoredefining class origin with reference to thefatherrsquos position alone understates intergenera-tional inequality Consequently conventionalclass origin measures could very well producemisleading comparisons of social fluidity lev-els between groups This final section of theanalysis evaluates the research consequencesof class origin measurement choices for detect-ing whether social fluidity levels have increaseddecreased or stayed constant across birthcohorts

Model 1 in Table 4 the conventional father-only model indicates that no significant changehas occurred in social fluidity levels between thecohorts for either gender (similarly diagonalimmobility parameters show no significantchange between cohorts) This provides a mis-leading picture of social fluidity trends howeverfor when joint-parent class origin measures areemployed in Model 2 conclusions about changein social fluidity are dramatically different Themother + father model shows that origin-desti-nation association has strengthened over thecohorts and therefore social fluidity hasdeclined particularly for the most recentcohort18 As Figure 2 illustrates the conven-tional father-only class origin measure (Model1) underestimates the origin-destination asso-ciation for the younger cohorts masking areduction in fluidity between cohorts that isrevealed by joint-parent class origin measures(Model 2)19 The reduction in fluidity experi-enced by the most recent cohort compared withthe earliest cohort is statistically significant formen but not for women20


17 I tested origins measures that differentiatedhomemaker mothers with respect to their husbandsrsquoclass (in other words I tested an interaction betweenthe homemaker category and husbandrsquos class) Theresults are similar to differentiating homemaker moth-ers by their education

18 I employ the mother + father model rather thanthe more parsimonious equal mother + father modelin the cohort analysis to capture how the relativeimportance of parentsrsquo class characteristics haschanged between cohorts

19 Similarly the conventional model is unable todetect a change in cohort fluidity whereas joint-par-ent origin models detect such change when cohortchange is assessed without controlling for age

20 Note that the models for men and women dif-fer in that homemaker mothers make up one broadclass category in menrsquos models but are differentiat-ed by education in womenrsquos models in accordancewith results from the previous analysis section Thedegree of measured association is therefore not direct-


Tue 08 Sep 2009 193307


Table 4 Cohort Change in Origin-Destination Association (Social Fluidity)

1 Father- 2 Mother + 3 Model 2 + Only Model Father Model Single Parents

Model Coef SE Coef SE Coef SE

Men

Baseline Association () 21 04 21 04 23 04mdashCohort 2(1955 to 1964)a 0 06 04 06 04 05mdashCohort 3(1965 to 1979)a 06 07 16 08 12dagger 07mdashAge a 02 02 04 02 04 02mdashAge2 a 0 0 00 0 00 0Baseline FatherndashSon Immobility 34 08mdashCohort 2(1955 to 1964)a 0 11mdashCohort 3(1965 to 1979)a 17 11Baseline FatherndashSon Immobility Dual-Earner Families 36 10 37 10mdashCohort 2(1955 to 1964)a ndash02 13 ndash03 13mdashCohort 3(1965 to 1979)a 02 13 ndash01 13Baseline FatherndashSon Immobility Single-Earner Families 36 12 37 11mdashCohort 2(1955 to 1964)a 02 16 ndash01 16mdashCohort 3(1965 to 1979)a 48 17 45 17Baseline MotherndashSon Immobility ndash01 12 ndash01 11mdashCohort 2(1955 to 1964)a 01 15 0 14mdashCohort 3(1965 to 1979)a 18 15 25dagger 13Age ndash14 05 ndash18 05 ndash11 04Age2 00 0 00 0 00 0N 3580 3580 4406Log Likelihood ndash60009 ndash59631 ndash73171LR Chi Square 11839 (25) 12595 (32) 14933 (32)BICrsquo ndash9794 ndash9976 ndash12248

Women

Baseline Association () 22 04 26 04 30 04mdashCohort 2(1955 to 1964)a ndash01 06 06 06 03 06mdashCohort 3(1965 to 1979)a 02 08 14 09 06 08mdashAgea 03 02 02 02 0 02mdashAge2 a 0 0 0 0 0 0Baseline FatherndashDaughter Immobility 13 09mdashCohort 2(1955 to 1964)a 04 13mdashCohort 3(1965 to 1979)a ndash08 14Baseline FatherndashDaughter Immobility Dual-Earner Families ndash10 12 ndash06 11mdashCohort 2(1955 to 1964)a 22 15 13 15mdashCohort 3(1965 to 1979)a 03 16 0 15Baseline FatherndashDaughter Immobility Single-Earner Families 27 14 26dagger 14mdashCohort 2(1955 to 1964)a ndash09 20 ndash10 19mdashCohort 3(1965 to 1979)a ndash02 23 ndash02 23Baseline MotherndashDaughter Immobility 30 09 22 09mdashCohort 2(1955 to 1964)a ndash25 13 ndash13 11mdashCohort 3(1965 to 1979)a ndash29 13 ndash17 11Age ndash02 05 0 06 ndash01 05Age2 0 0 0 0 0 0N 3620 3620 4570Log Likelihood ndash54004 ndash53298 ndash67018LR Chi Square 20721 (25) 22133 (32) 26141 (32)BICrsquo ndash18672 ndash19511 ndash23444

Notes See Table S4 in the Online Supplement for the class origin and destination scores for these models Agewas re-centered around 40 Higher class destination scores are negative (see Table S4 in the Online Supplement)therefore negative coefficients for age indicate a positive relationship between age and higher class destinationa These variables denote interaction terms (eg to obtain the estimated association for respondents in Cohort 3add the coefficients for the baseline association [] and Cohort 3 [1965 to 1979])dagger p lt 10 p lt 05


Tue 08 Sep 2009 193307


Figure 2 Association Parameter Estimates by Cohort Net of Age Effects

ly comparable between men and women Whencohort change is assessed for women without dif-ferentiating homemaker mothers by education thepattern of declining social fluidity between cohortsis less apparent (and it is not statistically significant

in either case) Prior research did not find evidenceof differences between men and women in socialfluidity trends over time however this researchemployed a period comparison rather than distin-guishing between cohorts (Hout 1988)

Footnotesize and

font iscorrect as

is Thislayout

was thebest ofseveral

attemptsnone of

whichwere

ideal


Tue 08 Sep 2009 193307

Additional differences between men andwomen include the finding that fatherndashsondiagonal immobility is significantly higher formen raised in single-earner families in themost recent birth cohort compared with theearliest cohort Among women there is nochange in fatherndashdaughter immobility betweencohorts but motherndashdaughter immobility issubstantially reduced in the second and thirdcohort compared with the first While there isno apparent effect of age for women respon-dent age among men is associated not onlywith higher class position but also withstronger origin-destination association whenjoint-parent origins measures are employedThis means that family class origins and classdestinations are more tightly linked amongolder than younger men within each cohort21

This relationship between menrsquos age andgreater intergenerational association is notdetectable when conventional origins meas-ures are used in Model 1

Considering the dramatic results of Model2 which indicate rising inequality particular-ly among men it is important to ask whetherthe finding is representative of a broader pop-ulation that includes respondents raised in sin-gle-parent families Some research suggeststhat respondents raised in non-intact familieshave higher rates of social mobility than dotheir peers (Biblarz and Raftery 1999)Therefore when considering population-levelestimates of social fluidity an increasing pro-portion of respondents raised in single-parentfamilies in more recent cohorts could coun-teract the declining social fluidity among thoseraised in intact families Model 3 in Table 4evaluates whether the results of Model 2 aremaintained when the population analyzedincludes respondents raised in single-parentfamilies Again among men the results ofModel 3 demonstrate significantly decliningsocial fluidity between cohorts for this morerepresentative population of respondentsAmong women however all the cohorts appear

less fluid after including respondents raised insingle-parent families

DISCUSSION AND CONCLUSIONS

The theoretical argument and empirical analy-ses presented in this article clearly demonstratethe importance of adequate measurement offamily-level class background to social mobil-ity research Conventional measurement whenapplied to family class origin assumes that netof the key measured parentrsquos class position asecond parentrsquos class-related resources do notaffect the family class position Yet as theo-rized mechanisms of intergenerational persist-ence in class would suggest this study showsthat parentsrsquo class resources jointly determinefamily class origin With this in mindmdashtogeth-er with the view that class is determined byresources that follow from employment rela-tionshipsmdashit may be helpful in the socialmobility research context to think beyond theidea of classes as necessarily bounded posi-tions that individuals or families occupy Ratherclass origins could be conceived of as sets offamily-level economic cultural and other class-related resources that shape childrenrsquos mobili-ty chances and are consequences of employmentrelationships occupational conditions and otherclass-related experiences of adults in the fam-ily

Empirically this study demonstrates thatjoint-parent measures of class origin capturemobility patterns significantly better than doconventional measures of class origin Althoughconventional measurement may provide a con-venient proxy for family-level class it is increas-ingly inadequate and fails to capture significantdeclines in social class fluidity among men bornbetween 1965 and 1979 (compared with earli-er cohorts) in the United States This decline asmy analyses show is a function of the growingassociation between mothersrsquo class and sonsrsquoclass destinations This example of change insocial fluidity between cohorts illustrates thatinadequate measurement of family class origincan distort research conclusions Moreover theimplications extend well beyond this particularexample and apply to comparative mobilityresearch more generally For example if moth-ersrsquoclass remains unmeasured and marital sort-ing by class differs cross-nationally differencesin measurement error could be misinterpreted


21 Hendrickx and Ganzeboom (1998) found theopposite relationship among men in the Netherlandsmore years of work experience led to lower origin-destination association Controlling for years of workexperience rather than age does not change the resultsof this analysis


Tue 08 Sep 2009 193307

as substantive differences between countries insocial fluidity levels On the flip side substan-tive cross-national differences in social fluidi-ty could also be masked22

Given the evidence offered regarding a dra-matic decline in social fluidity among recentbirth cohorts future research should examinethe social processes that may account for it Forexample womenrsquos careers and economic par-ticipation became more intangibly valued andrecognized during this period which potential-ly affected the extent to which children viewmothers as career role models In addition thereare at least two compelling explanations relat-ed to trends in increasing inequality betweenfamilies that could play a role in explaining therecent decline in social fluidity First family eco-nomic resources became more unequally dis-tributed due to rising income inequality Secondnoneconomic cultural family class resourcesalso became more unequally distributed due toincreasing educational assortative marriage

With respect to the first of the possibilitiesnoted above the explanation of rising eco-nomic inequality between families theyoungest cohort in this study grew up afterthe substantial increase in income inequalitythat occurred in the United States during the1970s and 1980s Greater income inequalitymight have heightened the advantages and dis-advantages of class origin thereby strength-ening the linkage between origins and classdestinations Indeed a similar pattern occursin studies of intergenerational income mobil-ity when income inequality in the childrenrsquosgeneration exceeds that of the parentsrsquo gener-ation intergenerational income mobilitydecreases (Solon 2001) For the rise in incomeinequality to be a plausible explanation fordeclining social class (as opposed to income)mobility depends however on whether incomeinequality increased more between as opposedto within class categories and also on theextent of assortative marriage by income indual-earner families

Like economic resources a key culturalclass resourcemdashparental educationmdashalsobecame more unequally distributed amongfamilies between the successive birth cohortsanalyzed in this study Educational assortativemarriage has increased since the 1960s fol-lowing a period of decline While this trend wasinitially due to an increasing propensity amongthose with a college degree to marry one anoth-er by the 1970s there was also a strong declinein the extent to which individuals with lowlevels of education married upward withrespect to education (Schwartz and Mare2005) Individuals born in the most recentbirth cohort analyzed in this study were thusmore likely than their predecessors to haveparents with similar levels of education Thedistribution of parental education between fam-ilies therefore became more unequal To theextent that education is related to occupation-al class position this pattern might correspondwith an increasing social divide between fam-ily class origin categories

While this article provides initial stepstoward modernizing family-based stratifica-tion research I do not measure class destina-tions at the family-level and the current lineof inquiry will be logically incomplete until thesame scrutiny I apply to class origin is extend-ed to class destination Future research mightevaluate for instance how inequalities of fam-ily class origins may or may not be com-pounded given family-level as opposed toindividual measurement of class destinations

Finally it is worth noting that limited datapresents a substantial challenge for updatingthe practice of intergenerational social mobil-ity research and stratification research morebroadly This article for instance illustrates theimportance of measuring both parentsrsquo classresources to adequately define class originyet most surveys do not collect information onnoncustodial parent occupation and many stilldo not collect mothersrsquo occupations For thetime being stratification and mobility researchshould of course proceed despite data limi-tations This article demonstrates that the bestpractice is to measure class origin as jointlydetermined by both parentsrsquo class characteris-tics but this is often not possible with currentdata sources Researchers can move forwarddespite limited data by on the one hand meas-uring class origin as comprehensively as pos-


22 In addition the argument presented here in favorof family-level measurement of childhood socialposition is not necessarily limited in applicability tothe particular class schema employed It could alsoapply to other class categorizations or methods of dif-ferentiating between social positions


Tue 08 Sep 2009 193307

REFERENCES

Acker Joan 1973 ldquoWomen and Social StratificationA Case of Intellectual Sexismrdquo American Journalof Sociology 78936ndash45

Agresti Alan and Ming-Chung Yang 1987 ldquoAnEmpirical Investigation of Some Effects of

Sparseness in Contingency Tablesrdquo ComputationalStatistics and Data Analysis 59ndash21

Baxter Janeen 1994 ldquoIs Husbandrsquos Class EnoughClass Locations and Class Identity in the UnitedStates Sweden Norway and Australiardquo AmericanSociological Review 59220ndash35


sible when possible and on the other hand bycarefully considering in the interpretation ofresearch findings the potential for omittedvariable bias given that important aspects ofchildrenrsquos class origins may remain unmea-sured

Emily Beller received her PhD in 2006 from theUniversity of California-Berkeley Department ofSociology and is currently a Senior Analyst with theUS Government Accountability Office focusing onhealth issues Her areas of research interest includesocial mobility and stratification families educationand health and illness

Table A1mdashClass Distributions of Fathers Mothers and Respondents (Weighted Percent)

Fathers Mothers Men Women

Respondents Raised in Two-Parent Families (N = 8189)mdashI 2263 504 2529 1939mdashII 1201 1440 1871 3231mdashIIIab 479 2456 828 3065mdashIVab 1096 503 782 609mdashVVI 2489 427 1926 299mdashVIIab 2472 1334 2065 858mdashHome 3335

Respondents Born 1945 to 1954 (N = 2522)mdashI 1834 358 3013 1976mdashII 905 1120 1886 3083mdashIIIab 475 2233 534 2946mdashIVab 970 525 911 747mdashVVI 2276 428 1568 321mdashVIIab 2275 1492 2087 927mdashHome 3705mdashNoncustodial 1265 139




Tue 08 Sep 2009 193307

Biblarz Timothy J and Adrian Raftery 1999 ldquoFamilyStructure Educational Attainment and Socio-eco-nomic Success Rethinking the Pathology ofMatriarchyrdquo American Journal of Sociology105321ndash65

Bourdieu Pierre and Jean-Claude Passeron 1977Reproduction in Education Society and CultureLondon UK Sage Publications

Breen Richard 1994 ldquoIndividual Level Models forMobility Tables and Other Cross-ClassificationsrdquoSociological Methods and Research 23147ndash73

mdashmdashmdash ed 2004 Social Mobility in EuropeOxford UK Oxford University Press

mdashmdashmdash 2005 ldquoFoundations of a Neo-WeberianClass Analysisrdquo Pp 31ndash50 in Approaches to ClassAnalysis edited by E O Wright Cambridge UKCambridge University Press

Breen Richard and Jan O Jonsson 2007ldquoExplaining Change in Social FluidityEducational Equalization and EducationalExpansion in Twentieth-Century SwedenrdquoAmerican Journal of Sociology 1121775ndash1810

Breen Richard and David B Rottman 1995 ClassStratif ication A Comparative PerspectiveHertfordshire UK Harvester Wheatsheaf

Britten Nicky and Anthony Heath 1983 ldquoWomenMen and Social Classrdquo Pp 46ndash60 in Gender Classand Work edited by E Gamarnikow D MorganJ Purvis and D Taylorson London UKHeinemann

Clogg Clifford C and Scott R Eliason 1987 ldquoSomeCommon Problems in Log-Linear AnalysisrdquoSociological Methods and Research 168ndash44

Clogg Clifford C and Edward S Shihadeh 1994Statistical Models for Ordinal Variables ThousandOaks CA Sage Publications

Conley Dalton 2001 ldquoCapital for College ParentalAssets and Postsecondary Schoolingrdquo Sociologyof Education 7459ndash72

Davis James A Tom W Smith and Peter V MarsdenGeneral Social Surveys 1972ndash2006 2007[CUMULATIVE FILE] [Computer f ile]ICPSR04697-v1 Chicago IL National OpinionResearch Center [producer] Storrs CT RoperCenter for Public Opinion Research University ofConnecticutAnn Arbor MI Inter-universityConsortium for Political and Social Research [dis-tributors] 2007-05-31

Davis Nancy J and Robert V Robinson 1988 ldquoClassIdentification of Men and Women in the 1970s and1980srdquo American Sociological Review53103ndash112

mdashmdashmdash 1998 ldquoDo Wives Matter Class Identitiesof Wives and Husbands in the United States1974ndash1994rdquo Social Forces 761063ndash86

Deil-Amen Regina and James E Rosenbaum 2003ldquoThe Social Prerequisites of Success Can CollegeStructure Reduce the Need for Social Know-

Howrdquo The Annals of the American Academy ofPolitical and Social Science 586120ndash43

DiPrete Thomas A 1990 ldquoAdding Covariates toLoglinear Models for the Study of Social MobilityrdquoAmerican Sociological Review 55757ndash73

DiPrete Thomas A and David Grusky 1990ldquoStructure and Trend in the Process ofStratification for American Men and WomenrdquoThe American Journal of Sociology 96107ndash143

Erikson Robert 1984 ldquoSocial Class of Men Womenand Familiesrdquo Sociology 18500ndash514

Erikson Robert and John H Goldthorpe 1992 TheConstant Flux Oxford UK Clarendon Press

Erikson Robert John H Goldthorpe and LuciennePortocarero 1979 ldquoIntergenerational ClassMobility in Three Western Societies EnglandFrance and Swedenrdquo The British Journal ofSociology 30415ndash41

Evans Geoffrey and Colin Mills 1998 ldquoAssessingthe Cross-Sex Validity of the Goldthorpe ClassSchema Using Log-linear Models with LatentVariablesrdquo Quality and Quantity 32275ndash96

Featherman David L and Robert M Hauser 1978Opportunity and Change New York AcademicPress

Ganzeboom Harry B G and Donald Trieman 2003ldquoThree Internationally Standardised Measures forComparative Research on Occupational StatusrdquoPp 159ndash93 in Advances in Cross-NationalComparison A European Working Book forDemographic and Socio-Economic Variables edit-ed by J H P Hoffmeyer-Zlotnik and C WolfNew York Kluwer Academic Press

Gerber Theodore P and Michael Hout 2004ldquoTightening Up Class Mobility during RussiarsquosMarket Transitionrdquo American Sociological Review69677ndash703

Goldthorpe John H 1983 ldquoWomen and ClassAnalysis In Defense of the Conventional ViewrdquoSociology 17465ndash88

mdashmdashmdash 1984 ldquoWomen and Class Analysis AReply to the Repliesrdquo Sociology 18491ndash99

mdashmdashmdash 2000 ldquoSocial Class and the Differentiationof Employment Contractsrdquo Pp 206ndash229 in OnSociology Numbers Narratives and theIntegration of Research and Theory Oxford UKOxford University Press

Goodman Leo 1979 ldquoMultiplicative Models of theAnalysis of Mobility Tables and Other Kinds ofMixed-Classification Tablesrdquo American Journal ofSociology 84804ndash819

Goodman Leo A and Clifford C Clogg 1992ldquoReview New Methods for the Analysis ofOccupational Mobility Tables and Other Kinds ofCross-Classificationsrdquo Contemporary Sociology21609ndash622

Harding David J Christopher Jencks Leonard MLopoo and Susan E Mayer 2005 ldquoThe ChangingEffects of Family Background on the Incomes of



Tue 08 Sep 2009 193307

American Adultsrdquo Pp 100ndash144 in UnequalChances Family Background and EconomicSuccess edited by S Bowles H Gintis and MGroves New York and Princeton NJ Russell SageFoundation and Princeton University Press

Hauser Robert M Shu-Ling Tsai and William HSewell 1983 ldquoA Model of Stratification withResponse Error in Social and PsychologicalVariablesrdquo Sociology of Education 5620ndash46

Heath Anthony and Nicky Britten 1984 ldquoWomenrsquosJobs Do Make a Difference A Reply toGoldthorperdquo Sociology 18475ndash90

Hendrickx John and Harry B G Ganzeboom 1998ldquoOccupational Status Attainment in theNetherlands 1920ndash1990 A Multinomial LogisticAnalysisrdquo European Sociological Review14387ndash403

Hill Martha and Greg Duncan 1987 ldquoParentalFamily Income and the Socioeconomic Attainmentof Childrenrdquo Social Science Research 1639ndash73

Hout Michael 1983 Mobility Tables SeriesQuantitative Applications in the Social SciencesNewbury Park CA Sage

mdashmdashmdash 1988 ldquoMore Universalism Less StructuralMobility The American Occupational Structure inthe 1980srdquo American Journal of Sociology931358ndash1400

mdashmdashmdash 1996 ldquoSpeed Bumps on the Road toMeritocracy The American Occupational Structurein the 1980srdquo Working Paper Survey ResearchCenter Berkeley CA

Kalmijn Matthijs 1994 ldquoMotherrsquos OccupationalStatus and Childrenrsquos Schoolingrdquo AmericanSociological Review 59257ndash75

Khazzoom Aziza 1997 ldquoThe Impact of MothersrsquoOccupations on Childrenrsquos OccupationalDestinationsrdquo Research in Social Stratificationand Mobility 1557ndash90

Korupp Sylvia E Harry B G Ganzeboom andTanja Van Der Lippe 2002 ldquoDo Mothers MatterA Comparison of Models of the Influence ofMothersrsquo and Fathersrsquo Educational andOccupational Status on Childrenrsquos EducationalAttainmentrdquo Quality and Quantity 3617ndash42

Lareau Annette 1989 Home Advantage SocialClass and Parental Intervention in ElementaryEducation New York Falmer Press

mdashmdashmdash 2003 Unequal Childhoods Class Raceand Family Life Berkeley CA UC Press

Lucas Samuel R 1999 Tracking InequalityStratification and Inequality in American HighSchools New York Teachers College Press

Mare Robert D 1981 ldquoChange and Stability inEducational Stratificationrdquo American SociologicalReview 4672ndash97

Marshall Gordon Stephen Roberts and CaroleBurgoyne 1996 ldquoSocial Class and Underclass inBritain and the United Statesrdquo The British Journalof Sociology 4722ndash44

Mayer Susan E 1997 What Money Canrsquot BuyFamily Income and Childrenrsquos Life ChancesCambridge MA Harvard University Press

Plutzer Eric and John F Zipp 2001 ldquoClass Genderand the Family Unit A Dynamic Model ofStratification and Class Politicsrdquo Social ScienceResearch 30426ndash48

Raftery Adrian E 1995 ldquoBayesian Model Selectionin Social Researchrdquo Sociological Methodology25111ndash63

Roemer John E 2004 ldquoEqual Opportunity andIntergenerational Mobility Going BeyondIntergenerational Income Transition MatricesrdquoPp 48ndash57 in Generational Income Mobility inNorth America and Europe edited by M CorakCambridge UK Cambridge University Press

Sewell William H Archibald O Haller andAlejandro Portes 1969 ldquoThe Educational andEarly Occupational Attainment Processrdquo AmericanSociological Review 3483ndash92

Schwartz Christine R and Robert Mare 2005ldquoTrends in Educational Assortative Marriage from1940ndash2003rdquo Demography 42641ndash26

Solon Gary 2001 ldquoMobility Within and BetweenGenerationsrdquo Pp 153ndash68 in The Causes andConsequences of Increasing Inequality edited byF Welsh Chicago IL Chicago University Press

Sorensen Annemette 1994 ldquoWomen Family andClassrdquo Annual Review of Sociology 2027ndash47

Stanworth Michelle 1984 ldquoWomen and ClassAnalysis A Reply to John Goldthorperdquo Sociology18159ndash70

Vermunt Jeroen K 1997 LEM 10 A GeneralProgram for the Analysis of Categorical DataTilburg Netherlands Tilburg University

Vines P and J A Priebe 1988 ldquoThe Relationshipbetween the 1970 and 1980 Industry andOccupation Classification Systemsrdquo WashingtonDC US Government Printing Office

Wong Raymond Sin-Kwok 1994 ldquoModel SelectionStrategies and the Use of Association Models toDetect Group Differencesrdquo Sociological Methodsand Research 22460ndash91



Tue 08 Sep 2009 193307

research practice remains adequateFurthermore there are practical incentives suchas data considerations to follow the conven-tional practice For example until 1994 theGeneral Social Survey did not ask respondentsabout their mothersrsquooccupations In this articleI demonstrate empirically that fathersrsquo class isan increasingly poor proxy for family socialclass background in the United States andresearch conclusions can be distorted when it isused as such

Given practical considerations such as lim-ited data on mothersrsquo occupations it is impor-tant to emphasize that adequately defining andmeasuring class background to bring it in linewith theory is essential for substantive ratherthan simply methodological reasons (egimproving explained variance or model fit inand of itself) To illustrate this point I exam-ine changes in intergenerational class mobili-ty between recent birth cohorts in the UnitedStates and show that the typical measurementstrategy masks important findings Withoutupdating intergenerational class mobility mod-els to incorporate mothersrsquoclass characteristics

a recent significant upturn in the importanceof class background for menrsquos class destinationsin the United States is not evident The con-ventional approach indeed suggests there wasno change in the extent of mobility betweencohorts Updated analyses however reveal thatclass mobility declined for men born since the1960s compared with those born earlier in thecentury

BACKGROUND

STRUCTURAL MOBILITY AND SOCIAL

FLUIDITY

Intergenerational class mobility research ana-lyzes the strength of the association betweenindividualsrsquoclass background or childhood classposition (class origin) and their current indi-vidual or family-level class position (class des-tination) as well as patterns of immobility ormovement between particular origins and des-tinations Such intergenerational class mobili-ty has two components worth examination Thefirst often called social fluidity refers to theextent to which an individualrsquos chances of reach-


Figure 1 Percent of Respondents with a Mother Who Worked Outside the Home by Year of Birth

Source General Social Surveys 1994 to 2006Delivered by Ingenta to Universidad Politecnica de Valencia

Tue 08 Sep 2009 193307




TIME









Tue 08 Sep 2009 193307












Tue 08 Sep 2009 193307













Tue 08 Sep 2009 193307











Tue 08 Sep 2009 193307



ANALYTIC STRATEGY

DATA AND METHODS










nonfarm and farm







Tue 08 Sep 2009 193307












Tue 08 Sep 2009 193307





One-Parent Measures


F + hiMF + j

D + 1Dij + uij


F + hiMF + j

D + 2Dhj + uhj


F + hiMF + j

D + 1Dhij + uhij


F + hiMF + j

D + 2Dhij + uhij



F + hiMF + j



F + hiMF + j



F + hiMF + j



F + hiMF + j




F + hiMF + j



F + hiMF + j




M + iF + j

D + ghCM + gi

CF + hiMF + ghi

CMF +


B2(age2))


M + iF + j

D + ghCM + gi

CF + hiMF + ghi

CMF +




2 = huh2 = hiuhi2 =

jj2 = 1




Tue 08 Sep 2009 193307






ANALYSIS STEPS









Tue 08 Sep 2009 193307






RESULTS







Tue 08 Sep 2009 193307








Men


Women





Tue 08 Sep 2009 193307
















Tue 08 Sep 2009 193307








1b2

Men


Women





Tue 08 Sep 2009 193307














Tue 08 Sep 2009 193307





Men


Women




Tue 08 Sep 2009 193307





Footnotesize and

font iscorrect as

is Thislayout


attemptsnone of

whichwere

ideal


Tue 08 Sep 2009 193307











Tue 08 Sep 2009 193307










Tue 08 Sep 2009 193307

REFERENCES















Tue 08 Sep 2009 193307


































Tue 08 Sep 2009 193307































Tue 08 Sep 2009 193307




TIME









Tue 08 Sep 2009 193307












Tue 08 Sep 2009 193307













Tue 08 Sep 2009 193307











Tue 08 Sep 2009 193307



ANALYTIC STRATEGY

DATA AND METHODS










nonfarm and farm







Tue 08 Sep 2009 193307












Tue 08 Sep 2009 193307





One-Parent Measures


F + hiMF + j

D + 1Dij + uij


F + hiMF + j

D + 2Dhj + uhj


F + hiMF + j

D + 1Dhij + uhij


F + hiMF + j

D + 2Dhij + uhij



F + hiMF + j



F + hiMF + j



F + hiMF + j



F + hiMF + j




F + hiMF + j



F + hiMF + j




M + iF + j

D + ghCM + gi

CF + hiMF + ghi

CMF +


B2(age2))


M + iF + j

D + ghCM + gi

CF + hiMF + ghi

CMF +




2 = huh2 = hiuhi2 =

jj2 = 1




Tue 08 Sep 2009 193307






ANALYSIS STEPS









Tue 08 Sep 2009 193307






RESULTS







Tue 08 Sep 2009 193307








Men


Women





Tue 08 Sep 2009 193307
















Tue 08 Sep 2009 193307








1b2

Men


Women





Tue 08 Sep 2009 193307














Tue 08 Sep 2009 193307





Men


Women




Tue 08 Sep 2009 193307





Footnotesize and

font iscorrect as

is Thislayout


attemptsnone of

whichwere

ideal


Tue 08 Sep 2009 193307











Tue 08 Sep 2009 193307










Tue 08 Sep 2009 193307

REFERENCES















Tue 08 Sep 2009 193307


































Tue 08 Sep 2009 193307































Tue 08 Sep 2009 193307












Tue 08 Sep 2009 193307













Tue 08 Sep 2009 193307











Tue 08 Sep 2009 193307



ANALYTIC STRATEGY

DATA AND METHODS










nonfarm and farm







Tue 08 Sep 2009 193307












Tue 08 Sep 2009 193307





One-Parent Measures


F + hiMF + j

D + 1Dij + uij


F + hiMF + j

D + 2Dhj + uhj


F + hiMF + j

D + 1Dhij + uhij


F + hiMF + j

D + 2Dhij + uhij



F + hiMF + j



F + hiMF + j



F + hiMF + j



F + hiMF + j




F + hiMF + j



F + hiMF + j




M + iF + j

D + ghCM + gi

CF + hiMF + ghi

CMF +


B2(age2))


M + iF + j

D + ghCM + gi

CF + hiMF + ghi

CMF +




2 = huh2 = hiuhi2 =

jj2 = 1




Tue 08 Sep 2009 193307






ANALYSIS STEPS









Tue 08 Sep 2009 193307






RESULTS







Tue 08 Sep 2009 193307








Men


Women





Tue 08 Sep 2009 193307
















Tue 08 Sep 2009 193307








1b2

Men


Women





Tue 08 Sep 2009 193307














Tue 08 Sep 2009 193307





Men


Women




Tue 08 Sep 2009 193307





Footnotesize and

font iscorrect as

is Thislayout


attemptsnone of

whichwere

ideal


Tue 08 Sep 2009 193307











Tue 08 Sep 2009 193307










Tue 08 Sep 2009 193307

REFERENCES















Tue 08 Sep 2009 193307


































Tue 08 Sep 2009 193307































Tue 08 Sep 2009 193307













Tue 08 Sep 2009 193307











Tue 08 Sep 2009 193307



ANALYTIC STRATEGY

DATA AND METHODS










nonfarm and farm







Tue 08 Sep 2009 193307












Tue 08 Sep 2009 193307





One-Parent Measures


F + hiMF + j

D + 1Dij + uij


F + hiMF + j

D + 2Dhj + uhj


F + hiMF + j

D + 1Dhij + uhij


F + hiMF + j

D + 2Dhij + uhij



F + hiMF + j



F + hiMF + j



F + hiMF + j



F + hiMF + j




F + hiMF + j



F + hiMF + j




M + iF + j

D + ghCM + gi

CF + hiMF + ghi

CMF +


B2(age2))


M + iF + j

D + ghCM + gi

CF + hiMF + ghi

CMF +




2 = huh2 = hiuhi2 =

jj2 = 1




Tue 08 Sep 2009 193307






ANALYSIS STEPS









Tue 08 Sep 2009 193307






RESULTS







Tue 08 Sep 2009 193307








Men


Women





Tue 08 Sep 2009 193307
















Tue 08 Sep 2009 193307








1b2

Men


Women





Tue 08 Sep 2009 193307














Tue 08 Sep 2009 193307





Men


Women




Tue 08 Sep 2009 193307





Footnotesize and

font iscorrect as

is Thislayout


attemptsnone of

whichwere

ideal


Tue 08 Sep 2009 193307











Tue 08 Sep 2009 193307










Tue 08 Sep 2009 193307

REFERENCES















Tue 08 Sep 2009 193307


































Tue 08 Sep 2009 193307































Tue 08 Sep 2009 193307











Tue 08 Sep 2009 193307



ANALYTIC STRATEGY

DATA AND METHODS










nonfarm and farm







Tue 08 Sep 2009 193307












Tue 08 Sep 2009 193307





One-Parent Measures


F + hiMF + j

D + 1Dij + uij


F + hiMF + j

D + 2Dhj + uhj


F + hiMF + j

D + 1Dhij + uhij


F + hiMF + j

D + 2Dhij + uhij



F + hiMF + j



F + hiMF + j



F + hiMF + j



F + hiMF + j




F + hiMF + j



F + hiMF + j




M + iF + j

D + ghCM + gi

CF + hiMF + ghi

CMF +


B2(age2))


M + iF + j

D + ghCM + gi

CF + hiMF + ghi

CMF +




2 = huh2 = hiuhi2 =

jj2 = 1




Tue 08 Sep 2009 193307






ANALYSIS STEPS









Tue 08 Sep 2009 193307






RESULTS







Tue 08 Sep 2009 193307








Men


Women





Tue 08 Sep 2009 193307
















Tue 08 Sep 2009 193307








1b2

Men


Women





Tue 08 Sep 2009 193307














Tue 08 Sep 2009 193307





Men


Women




Tue 08 Sep 2009 193307





Footnotesize and

font iscorrect as

is Thislayout


attemptsnone of

whichwere

ideal


Tue 08 Sep 2009 193307











Tue 08 Sep 2009 193307










Tue 08 Sep 2009 193307

REFERENCES















Tue 08 Sep 2009 193307


































Tue 08 Sep 2009 193307































Tue 08 Sep 2009 193307



ANALYTIC STRATEGY

DATA AND METHODS










nonfarm and farm







Tue 08 Sep 2009 193307












Tue 08 Sep 2009 193307





One-Parent Measures


F + hiMF + j

D + 1Dij + uij


F + hiMF + j

D + 2Dhj + uhj


F + hiMF + j

D + 1Dhij + uhij


F + hiMF + j

D + 2Dhij + uhij



F + hiMF + j



F + hiMF + j



F + hiMF + j



F + hiMF + j




F + hiMF + j



F + hiMF + j




M + iF + j

D + ghCM + gi

CF + hiMF + ghi

CMF +


B2(age2))


M + iF + j

D + ghCM + gi

CF + hiMF + ghi

CMF +




2 = huh2 = hiuhi2 =

jj2 = 1




Tue 08 Sep 2009 193307






ANALYSIS STEPS









Tue 08 Sep 2009 193307






RESULTS







Tue 08 Sep 2009 193307








Men


Women





Tue 08 Sep 2009 193307
















Tue 08 Sep 2009 193307








1b2

Men


Women





Tue 08 Sep 2009 193307














Tue 08 Sep 2009 193307





Men


Women




Tue 08 Sep 2009 193307





Footnotesize and

font iscorrect as

is Thislayout


attemptsnone of

whichwere

ideal


Tue 08 Sep 2009 193307











Tue 08 Sep 2009 193307










Tue 08 Sep 2009 193307

REFERENCES















Tue 08 Sep 2009 193307


































Tue 08 Sep 2009 193307































Tue 08 Sep 2009 193307












Tue 08 Sep 2009 193307





One-Parent Measures


F + hiMF + j

D + 1Dij + uij


F + hiMF + j

D + 2Dhj + uhj


F + hiMF + j

D + 1Dhij + uhij


F + hiMF + j

D + 2Dhij + uhij



F + hiMF + j



F + hiMF + j



F + hiMF + j



F + hiMF + j




F + hiMF + j



F + hiMF + j




M + iF + j

D + ghCM + gi

CF + hiMF + ghi

CMF +


B2(age2))


M + iF + j

D + ghCM + gi

CF + hiMF + ghi

CMF +




2 = huh2 = hiuhi2 =

jj2 = 1




Tue 08 Sep 2009 193307






ANALYSIS STEPS









Tue 08 Sep 2009 193307






RESULTS







Tue 08 Sep 2009 193307








Men


Women





Tue 08 Sep 2009 193307
















Tue 08 Sep 2009 193307








1b2

Men


Women





Tue 08 Sep 2009 193307














Tue 08 Sep 2009 193307





Men


Women




Tue 08 Sep 2009 193307





Footnotesize and

font iscorrect as

is Thislayout


attemptsnone of

whichwere

ideal


Tue 08 Sep 2009 193307











Tue 08 Sep 2009 193307










Tue 08 Sep 2009 193307

REFERENCES















Tue 08 Sep 2009 193307


































Tue 08 Sep 2009 193307































Tue 08 Sep 2009 193307





One-Parent Measures


F + hiMF + j

D + 1Dij + uij


F + hiMF + j

D + 2Dhj + uhj


F + hiMF + j

D + 1Dhij + uhij


F + hiMF + j

D + 2Dhij + uhij



F + hiMF + j



F + hiMF + j



F + hiMF + j



F + hiMF + j




F + hiMF + j



F + hiMF + j




M + iF + j

D + ghCM + gi

CF + hiMF + ghi

CMF +


B2(age2))


M + iF + j

D + ghCM + gi

CF + hiMF + ghi

CMF +




2 = huh2 = hiuhi2 =

jj2 = 1




Tue 08 Sep 2009 193307






ANALYSIS STEPS









Tue 08 Sep 2009 193307






RESULTS







Tue 08 Sep 2009 193307








Men


Women





Tue 08 Sep 2009 193307
















Tue 08 Sep 2009 193307








1b2

Men


Women





Tue 08 Sep 2009 193307














Tue 08 Sep 2009 193307





Men


Women




Tue 08 Sep 2009 193307





Footnotesize and

font iscorrect as

is Thislayout


attemptsnone of

whichwere

ideal


Tue 08 Sep 2009 193307











Tue 08 Sep 2009 193307










Tue 08 Sep 2009 193307

REFERENCES















Tue 08 Sep 2009 193307


































Tue 08 Sep 2009 193307































Tue 08 Sep 2009 193307






ANALYSIS STEPS









Tue 08 Sep 2009 193307






RESULTS







Tue 08 Sep 2009 193307








Men


Women





Tue 08 Sep 2009 193307
















Tue 08 Sep 2009 193307








1b2

Men


Women





Tue 08 Sep 2009 193307














Tue 08 Sep 2009 193307





Men


Women




Tue 08 Sep 2009 193307





Footnotesize and

font iscorrect as

is Thislayout


attemptsnone of

whichwere

ideal


Tue 08 Sep 2009 193307











Tue 08 Sep 2009 193307










Tue 08 Sep 2009 193307

REFERENCES















Tue 08 Sep 2009 193307


































Tue 08 Sep 2009 193307































Tue 08 Sep 2009 193307






RESULTS







Tue 08 Sep 2009 193307








Men


Women





Tue 08 Sep 2009 193307
















Tue 08 Sep 2009 193307








1b2

Men


Women





Tue 08 Sep 2009 193307














Tue 08 Sep 2009 193307





Men


Women




Tue 08 Sep 2009 193307





Footnotesize and

font iscorrect as

is Thislayout


attemptsnone of

whichwere

ideal


Tue 08 Sep 2009 193307











Tue 08 Sep 2009 193307










Tue 08 Sep 2009 193307

REFERENCES















Tue 08 Sep 2009 193307


































Tue 08 Sep 2009 193307































Tue 08 Sep 2009 193307








Men


Women





Tue 08 Sep 2009 193307
















Tue 08 Sep 2009 193307








1b2

Men


Women





Tue 08 Sep 2009 193307














Tue 08 Sep 2009 193307





Men


Women




Tue 08 Sep 2009 193307





Footnotesize and

font iscorrect as

is Thislayout


attemptsnone of

whichwere

ideal


Tue 08 Sep 2009 193307











Tue 08 Sep 2009 193307










Tue 08 Sep 2009 193307

REFERENCES















Tue 08 Sep 2009 193307


































Tue 08 Sep 2009 193307































Tue 08 Sep 2009 193307
















Tue 08 Sep 2009 193307








1b2

Men


Women





Tue 08 Sep 2009 193307














Tue 08 Sep 2009 193307





Men


Women




Tue 08 Sep 2009 193307





Footnotesize and

font iscorrect as

is Thislayout


attemptsnone of

whichwere

ideal


Tue 08 Sep 2009 193307











Tue 08 Sep 2009 193307










Tue 08 Sep 2009 193307

REFERENCES















Tue 08 Sep 2009 193307


































Tue 08 Sep 2009 193307































Tue 08 Sep 2009 193307








1b2

Men


Women





Tue 08 Sep 2009 193307














Tue 08 Sep 2009 193307





Men


Women




Tue 08 Sep 2009 193307





Footnotesize and

font iscorrect as

is Thislayout


attemptsnone of

whichwere

ideal


Tue 08 Sep 2009 193307











Tue 08 Sep 2009 193307










Tue 08 Sep 2009 193307

REFERENCES















Tue 08 Sep 2009 193307


































Tue 08 Sep 2009 193307































Tue 08 Sep 2009 193307














Tue 08 Sep 2009 193307





Men


Women




Tue 08 Sep 2009 193307





Footnotesize and

font iscorrect as

is Thislayout


attemptsnone of

whichwere

ideal


Tue 08 Sep 2009 193307











Tue 08 Sep 2009 193307










Tue 08 Sep 2009 193307

REFERENCES















Tue 08 Sep 2009 193307


































Tue 08 Sep 2009 193307































Tue 08 Sep 2009 193307





Men


Women




Tue 08 Sep 2009 193307





Footnotesize and

font iscorrect as

is Thislayout


attemptsnone of

whichwere

ideal


Tue 08 Sep 2009 193307











Tue 08 Sep 2009 193307










Tue 08 Sep 2009 193307

REFERENCES















Tue 08 Sep 2009 193307


































Tue 08 Sep 2009 193307































Tue 08 Sep 2009 193307





Footnotesize and

font iscorrect as

is Thislayout


attemptsnone of

whichwere

ideal


Tue 08 Sep 2009 193307











Tue 08 Sep 2009 193307










Tue 08 Sep 2009 193307

REFERENCES















Tue 08 Sep 2009 193307


































Tue 08 Sep 2009 193307































Tue 08 Sep 2009 193307











Tue 08 Sep 2009 193307










Tue 08 Sep 2009 193307

REFERENCES















Tue 08 Sep 2009 193307


































Tue 08 Sep 2009 193307































Tue 08 Sep 2009 193307










Tue 08 Sep 2009 193307

REFERENCES















Tue 08 Sep 2009 193307


































Tue 08 Sep 2009 193307































Tue 08 Sep 2009 193307

REFERENCES















Tue 08 Sep 2009 193307


































Tue 08 Sep 2009 193307































Tue 08 Sep 2009 193307


































Tue 08 Sep 2009 193307































Tue 08 Sep 2009 193307































Tue 08 Sep 2009 193307

bringing intergenerational social mobility research into...

Documents