working for the woman? female managers and the gender wage …pnc/asr07.pdf · narrow the gender...

The past two decades have seen substantialincreases in the proportion of women in

management During this time womenrsquos rep-resentation in managerial occupations increasedfrom about one-third to one-half1 These posi-tions confer well-documented benefits includ-ing improved status wages autonomy and

overall work experience (England et al 1994Reskin and Ross 1992) In recent years a spateof empirical research has addressed womenrsquosaccess to managerial authority (Blum Fieldsand Goodman 1994 Huffman and Cohen2004a Reskin and McBrier 2000 Smith 2002)and the ldquoglass ceilingrdquomdashan unseen barrierbetween women and management or high-status positions (Cotter et al 2001 Hultin 2003Wright and Baxter 2000)

Although the question of access to manage-rial positions is critical to understanding per-sistent gender inequality in the labor market theincrease in womenrsquos managerial presence rais-es a broader question that is provocative andinherently sociological What happens to thestatus of a subordinate group when some of itsmembers attain positions from which they mightreduce inequality We use gender to gain insightinto this question Specifically we ask whetherthe increase in womenrsquos representation in man-agement ldquolifts all boatsrdquo by reducing genderinequality among nonmanagerial workers orwhether the benefits that accrue to female man-agers are limited only to those women Clearlythe actions of managers affect those below them(Wright 1997) Yet managersrsquo role in repro-ducing gender inequality is conspicuouslyunderstudied despite its relevance for persistentlabor market inequality (Hultin and Szulkin1999 Hultin and Szulkin 2003) and for broad-

Working for the Woman FemaleManagers and the Gender Wage Gap

Philip N Cohen Matt L HuffmanUniversity of North Carolina Chapel Hill University of California Irvine

Most previous research on gender inequality and management has been concerned with

the question of access to managerial jobs and the ldquoglass ceilingrdquo We offer the first large-

scale analysis that turns this question around asking whether the gender characteristics

of managersmdashspecifically the gender composition and relative status of female

managersmdashaffect inequality for the nonmanagerial workers beneath them Results from

three-level hierarchical linear models estimated on a unique nested data set drawn from

the 2000 Census suggest that greater representation of women in management does

narrow the gender wage gap Model predictions show however that the presence of

high-status female managers has a much larger impact on gender wage inequality We

conclude that the promotion of women into management positions may benefit all

women but only if female managers reach relatively high-status positions

AMERICAN SOCIOLOGICAL REVIEW 2007 VOL 72 (October681ndash704)

The authors contributed equally to this manuscriptand list their names alphabetically Direct corre-spondence to Philip N Cohen Department ofSociology University of North Carolina-Chapel Hill155 Hamilton Hall CB3210 Chapel Hill NC27599-3210 (pncuncedu) This research was sup-ported by a grant from the National ScienceFoundation (SES-0647265) The authors also thankthe ASR editors and anonymous reviewers for theirhelpful comments and suggestions In addition webenefited from comments by Kenneth AndrewsBarbara Entwisle Roberto Fernandez ArneKalleberg Joan S M Meyers Calvin Morrill TedMouw Judy Ruttenberg and the participants of theGender Work and Family research group at theUniversity of California Irvine Any remaining errorsor omissions are solely our own

1 This result is from our unpublished analysis ofdata from the Current Population Survey (availableupon request) Interestingly some of this changeoccurred at the same time that progress for USwomen stalled on many other fronts (CotterHermsen and Vanneman 2004)

Delivered by Ingenta to University of North CarolinaMon 12 Nov 2007 181631

er questions about the status of subordinategroups

We adopt a unique approach to this ques-tion Rather than analyzing managerial workersalone (eg Huffman and Velasco 1997) or sim-ply including them with nonmanagerial work-ers (eg Cohen and Huffman 2003a Jacobs1992) we analyze the wages of nonmanageri-al workers as a function of the gender compo-sition of their managersmdashthe managers in theirlocal industries Further we extend the existingliterature by considering whether the relativestatus of female managers affects the pattern ofgender inequality for the workers beneath themIf female managers influence gender inequali-ty the effect may depend on how highly placedthose female managers are (Denmark 1993)Although this point may seem prosaic carefulattention to the relative status of female man-agers allows us to make more nuanced obser-vations about the conditions under which labormarket benefits extend to ascriptively similarsubordinates Using data from the 2000 USCensus we estimate multilevel wage modelswith controls at three levelsmdashthe individualthe job and the local industrymdashto assess theimpact of female managers on gender inequal-ity These advances provide new leverage ontheoretically important yet unanswered ques-tions concerning the role of managers in gen-der stratification with implications for theinequality trajectories of subordinate groupsmore generally

MANAGERS AS AGENTS OFCHANGE OR COGS IN THEMACHINE

AGENTS OF CHANGE

Many researchers believe that gender inequal-ity at work results in part from the practices ofmanagersmdashoften assuming that these practicesare associated with managersrsquogender For exam-ple Cotter and colleagues (1997715) offer thisas one reason why women benefit from occu-pational integration in the local labor marketldquoAs more women in [positions of authority]make crucial decisions about salaries promo-tions hiring and firing gender differences inearnings should declinerdquo (emphasis added)Similarly Nelson and Bridges (1999) argue thatthe scarcity of women in authority positionssustains workplace gender inequality Managers

by definition have a certain level of organiza-tional authority and thus might be poised tohelp reduce inequality at lower organizationallevels For female managers to reduce workplaceinequality however two assumptions must holdFirst these women must be motivated to act inthe interests of subordinate women Secondthey must have the power to influence outcomesfor subordinates to affect gender inequality

Gender creates a potential common interestbetween female managers and subordinatesbased on homophily (Ibarra 1992 McPhersonSmith-Lovin and Cook 2001) or in Kanterrsquos(1977) term for the tendency of women to hireother women ldquohomosocial reproductionrdquo (seealso Elliott and Smith 2004 Pfeffer 1983) Sexsimilarity between subordinates and supervi-sors increases performance ratings by supervi-sors (Roth 2004 Tsui and OrsquoReilly 1989) andwomenrsquos evaluation of potential female job can-didates is less subject to pregnancy-related bias(Halpert Wilson and Hickman 1993)

More broadly women express stronger sup-port than men do for employer practices aimedat overcoming gender inequality The 1996General Social Survey asked for the level ofagreement with the statement ldquoBecause of pastdiscrimination employers should make specialefforts to hire and promote qualified womenrdquoEmployed women were 119 times more likelythan men to agree (595 versus 498 percentp lt 001 N = 1373) Among managers thedifference was larger with women 132 timesmore likely than men to agree (535 versus 404percent p = 068 N = 193) Not only are womenmore supportive of efforts toward workplaceequality in principle but manager bias againstsuch efforts among women is also less sug-gesting that the presence of female managersshould (if they have the power) promote genderequality2

682mdashndashAMERICAN SOCIOLOGICAL REVIEW

2 The difference in agreement between nonman-agers and managers was not significant for women(604 percent versus 535 percent agreeing p = 197N = 753) but substantial for men (515 percent ver-sus 404 percent p = 047 N = 620) Baunach (2002)analyzed the same data set and reports no significantgender difference but she used a subsample ofrespondents with insufficient power to identify thedifference (N = 313)


Ely (1995) argues that the presence of womenin the upper echelons of organizations reducesthe persistence of sex as a salient category forall workers thereby weakening some of thenegative consequences associated with genderimbalance (eg performance pressures stereo-typed role encapsulation and exclusion fromwork-related networks) She demonstrates thatthe demographic composition of those holdingpowerful positions in organizations can havesubstantial effects on all workers not just thoseholding positions at or near the top of organi-zational hierarchies

Supporting this view some empirical studieshave shown less inequality where women occu-py positions of authority Hultin and Szulkin(2003) offer the most direct test of the wageeffect of female managers using rare employ-erndashemployee linked data from Swedish private-sector work establishments They find a negativerelationship between the gender wage gapamong nonmanagerial workers and the propor-tion of women in managerial positions Thisrelationship which remains substantial in thepresence of controls for individual attributesestablishment characteristics and industry ispresent for both white- and blue-collar employ-ees Importantly Hultin and Szulkin (2003) dis-tinguish between higher-level decision makers(managers) and lower-level decision makers(supervisors) They find that the effect of the sexcomposition of supervisors on wage inequali-ty is stronger than that of managers3 Althoughtheir analysis shows that the level of authorityis an important consideration female managersat high levels in the hierarchy are not shown tohave a stronger effect than those at lower lev-els4

A number of other studies are also consistentwith the contention that female managers reduce

gender inequality but most rely on narrow sam-ples or settings An exception is a study of threeUS cities which finds that promotions fromsupervisor to manager occur more frequentlyunder conditions of ascriptive similarity(raceethnicity and gender) with immediatesupervisors (Elliott and Smith 2004) Narrowerstudies have shown for example that Californiastate agencies with more female managersexhibited less gender segregation in the 1970sand 1980s (Baron Mittman and Newman1991) and savings and loans with women inmanagement are more likely to hire women intomanagerial roles (Cohen Broschak andHaveman 1998) Additionally Carrington andTroske (1995) find a strong link between thegender of business owners and the gender com-position of their employees

A series of studies investigating higher edu-cation settings shows that female administrators(Kulis 1997) or a female president (PfefferDavis-Blake and Julius 1995) are associatedwith less gender segregation (see also Konradand Pfeffer 1991) In the legal profession lawfirms whose corporate clients have manywomen in leadership positions show a greaterincrease in female partners (Beckman andPhillips 2005) and female decision makers tendto fill more vacancies with women (Gorman2005) Finally prime-time television shows withfemale producers executive producers andwriters have a higher percentage of female majorcharacters (Glascock 2001 Lauzen and Dozier1999)

These studies imply that there is less genderinequality under conditions of greater femalerepresentation (and higher status) in manage-ment5 This may result from several distinctmechanisms including increased access to orga-nizational resources and power homophily pref-erences support for equity efforts and weakersex-based biases against female workers

WORKING FOR THE WOMANmdashndash683

3 Due to data limitations this portion of theiranalysis was only performed on the blue-collar sub-sample

4 Although we do not link workers to managersdirectly as do Hultin and Szulkin (2003) our studyis unique in that it includes controls at three levelsmdashthe individual the job and the local industry Assuch it accounts for variation in gender inequalityacross larger social contexts which purely organi-zation-based analyses do not (Cohen and Huffman2003b)

5 Our emphasis on inequality effects differentiatesthis study from research on gender differences inleadership styles but the pursuit of that question inthe management and psychology literature providessome evidence of a less authoritarian orientationamong female leaders which may benefit femalesubordinates (for reviews see Dobbins and Platz1986 van Engen and Willemsen 2004)


COGS IN THE MACHINE

Although there are reasons to believe femalemanagers might reduce inequality the under-lying motivation and power assumptions aredebatable The motivations of female managersmay be affected by two potential sources ofloyalty or identity their female peers in subor-dinate class positions and their managerial peersand superiors Ely (1995) argues that to assumefemale managers are sympathetic to the womenbelow them essentializes gender while in prac-tice gender is situationally enacted6 Class is onesource of distinction that might prevent theexpression of collective identity among women(Young 1994) In fact a selection process mayoperate such that female workers are promotedinto management partly for their affinity withthe existing hierarchy The disproportionate pro-motion of women who are ldquoteam playersrdquo maylimit the potential for female managers to actagainst inequality

Further some women share menrsquos biasedviews of womenrsquos work (Deaux 1985) Forexample women and men in college similarlydevalue the merit of female job applicants whoseresumes reflect motherhood status (Correll andBenard 2005) With regard to networking evenif female or minority managers are likely topass on job leads or other job-related informa-tion to subordinates research suggests suchcontacts may not be systematically beneficial(Huffman and Torres 2002 Mouw 2003)

The managerial power assumption is alsopotentially flawed It is not obvious that man-agers especially those in bureaucratic organi-zations are able to act autonomously on thebasis of their own or womenrsquos interests Insteadthey may be compelled to act under the man-dates of routinization efficiency or profitabil-itymdashor according to the prejudices of thosehigher up the hierarchy This counterargumentwas summarized by Merton (1940560) as

ldquoauthority || adheres in the office and not inthe particular person who performs the officialrolerdquo In fact as Charles and Grusky (2004)show the increase in managerial integrationafter the 1970s occurred during a period ofgrowing bureaucratization which implies lim-its on the discretionary power of lower-levelmanagers Kanter (1977) argues that femalemanagers in particular occupy weak structuralpositions In Ridgewayrsquos words they are ldquohand-icapped by their lower power and by interac-tional gender mechanismsrdquo (1997227)Affirmative action programs have been moresuccessful at integrating lower and middle lev-els of management (Brenner Tomkiewicz andSchein 1989) and to some extent womenrsquosincreasing managerial presence reflects ldquotitleinflationrdquo (see Jacobs 1992)mdashthe reclassifica-tion of previously nonmanagerial workers asmanagers with little increase in pay or author-ity

Finally it is possible that due to a baselinesectoral segregation women are typically man-agers in workplaces with lower quality jobsBoth Shenhav and Haberfeld (1992) and Pfefferand Davis-Blake (1987) find lower earningsfor both men and women in workplaces withmore female managers or administrators Iffemale managers are concentrated in organiza-tions with more female workers then any pos-itive effect of female manager attitudes orbehavior may be swamped by negative genderconcentration effects

In summary female managers may enhancethe labor market prospects of the women whowork below them Their homophilous prefer-ences or affiliations might promote equalityand they may have less to gain from discrimi-nation and therefore be more motivated to helpother women Additionally women may be moreaware than men of discriminatory practices andless susceptible to cognitive processes leadingto gender bias Any of these processes maysmooth the social and organizational path offemale subordinates In contrast bureaucracymarket pressures divided loyalties past dis-crimination or the mandates of those morepowerful may render the ascriptive characteris-tics of managers largely moot with regard toinequality


6 Research in psychology shows that to the extentthere are gender differences in moral reasoning asadvanced by Gilligan (1982) they are context-dependent (Ryan David and Reynolds 2004) andconditioned on among other factors socioeconom-ic status For our purposes however we note thatJaffee and Hyde (2000) find greater gender differ-ences in moral reasoning at higher levels of socialclass


MANAGERSrsquo RELATIVE STATUS

Regardless of how female managers mightreduce inequality their impact may depend crit-ically on their status (Denmark 1993) Evenhighly motivated female managers workingtogether may not be able to influence genderinequality if they are relatively powerless Boththe identityloyalty issue and the question ofmanagerial power highlight the possibility ofsubstantial interactions between the represen-tation of women and their relative statusAlthough disparate studies have investigatedthe effect of female representation among man-agers at different levels (eg university presi-dents screenwriters law firm partners) Hultinand Szulkinrsquos (2003) is the only one to directlytest for the effect across different levels withinone setting and their ability to address the issueis limited by their data which includes onlyundifferentiated manager and supervisor cate-gories In contrast we use a continuous meas-ure of vertical segregation (explained below) totap the relative status of female managers

LOCAL INDUSTRIES

Most of the research in this area is concernedwith the direct effects of managers within organ-izations Studies have thus been designed todraw from managers and workers who are asclosely linked as possible exemplified by thework of Hultin and Szulkin (2003) and Elliottand Smith (2004) Women in positions ofauthority however may change gender dynam-ics at various levels of proximity among imme-diate subordinates within their organizations ingeneral across organizations and across larg-er social contexts We know that gender inequal-ity varies systematically across larger socialcontexts including metropolitan labor markets(Cohen and Huffman 2003a Cotter et al 1997)and national industries (Fields and Wolff 1995Wharton 1986) This variation is not capturedwhen analysis is limited to direct examinationof organizations (Cohen and Huffman 2003b)

One way the processes reproducing inequal-ity across contextual levels would be linked isif female managers in one workplace affectproximate organizations For example anemployerrsquos decision to hire women creates acompetitive advantage relative to those who donot (Ashenfelter and Hannan 1986) especiallywhen women are paid less than men (Neumark

and Stock 2006) Such competition is one waythat corporate practicesmdashwhich presumablyinclude those related to gender inequalitymdashareadapted by different actors within an organiza-tional field Additionally employment practicescan be adopted by organizations in an effort toincrease legitimacy or to appear in compliancewith a changing legal environment (DiMaggioand Powell 1983 Tomaskovic-Devey andStainback 2007)

DiMaggio and Powellrsquos (1983) theory hasprompted an extensive literature on how to oper-ationalize the reference groups or fields fororganizational behavior (Greve 2005 MassiniLewin and Greve 2005 Strang and Soule1998) Employing organizations are part oflabor markets that are local (represented bymetropolitan areas) industries that are nation-al (represented by categorization schemes withvarious degrees of detail) and the intersectionof the two local industries We believe the localindustrymdashthe aggregate of organizations thatproduce a common product within a commonlocal labor marketmdashis an appropriate startingpoint for our questions This approach drawsfrom work on local industrial dominance bySouth and Xu (1990) who argue that localindustries are sites of intersection between struc-tural forces and individual attainment process-es These units combine functional commonalitywith social proximity capturing the interactionof these fields

We thus expect local industries to displayless internal variation in gender-related practicesthan either metropolitan labor markets or nation-al industries as a whole This implies that agiven restaurant for example is likely to resem-ble other restaurants in its local area more thanit resembles either restaurants in the entire coun-try or all employers in the local labor marketAlthough we cannot fully demonstrate this pat-tern we provide a simple illustration based onone key variable for establishments the gendercomposition of managers7

Using data from all large US private-sectorfirms collected by the US Equal EmploymentOpportunity Commission (EEOC) in 2002 we


7 This is the measure used by Ashenfelter andHannan (1986) who examine the pattern of genderrepresentation in management for banks across localmarkets


measured the similarity of establishments onmanagerial gender composition across 275 met-ropolitan labor markets 301 national indus-tries and 10131 local-industry cells formedby the intersection of labor markets and indus-tries First we calculated the natural logarithmof the percentage female among ldquoofficers andmanagersrdquo at each establishment We then cal-culated the standard deviation in the naturallogarithm for each contextual unitmdasheach met-ropolitan area national industry and local indus-try Finally we averaged the standard deviationsacross these contextual units The means ofthese standard deviations are 145 for metro-politan areas 122 for national industries and103 for local industries8 In terms of the gen-der composition of managers establishments onaverage are indeed more similar to others with-in their local industries than they are to thosewithin their entire local labor markets or nation-al industries as a whole This supports the pre-sumption that gender-related organizationaldynamics generally cluster or reproduce moretightly within local industries than within theselarger units

Of course we remain interested in within-organization effects of managerial gender com-position on gender inequality Despite reasonsto suspect larger processes this more directeffect remains the most plausible pathway bywhich female managers influence gendered out-comes as has been shown in the limited researchthat uses such linked data In the absence of suchdata on a generalizable scale however we takeheart from the results of our EEOC exercisewhich imply that for a given worker the gendercomposition of the local industryrsquos managers is

likely to approximate that of her own estab-lishment at least compared to the correspon-dence obtained at the level of the local labormarket or national industry We thus considerlocal-industry managerial composition as aproxy for establishment management charac-teristics

HYPOTHESES

Previous research has shown that the overallwage gap largely results from between-job andbetween-occupation inequality whereby female-dominated jobs and occupations pay less thanotherwise comparable male-dominated lines ofwork (Cohen and Huffman 2003a England etal 1994 Huffman and Velasco 1997 Tam1997) in part because female-dominated jobsoffer fewer training opportunities (Tomaskovic-Devey and Skaggs 2002) However inequalitywithin job and occupational categories also con-tributes to wage inequality Clearly manageri-al composition and relative status could shapewage inequality through either route as man-agers might influence both sorting and the train-ing and rewards processes Rather than makepredictions without justification from prior the-ory or research our models account for bothpossibilities We also do not consider more com-plex interactions (eg McCall 2001) such asthose involving the conditional effects of man-agersrsquo race and gender

Our analysis concerns the extent to whichwages for men and women are sensitive to boththe representation of women among managersand the relative status of those female man-agers We test two hypotheses beginning withwhether the gender wage gap is smaller in localindustries where there are more female man-agers Specifically we test

Hypothesis 1 There is less gender wageinequality in local industries with a high-er proportion of female managers

This hypothesis addresses the associationbetween the representation of women in man-agerial positions and wage inequality If femalemanagers tend to cluster at the bottom of man-agerial hierarchies though their mere repre-sentation in management may be insufficient toalter wage inequality Therefore we test theinteraction between the representation and therelative status of female managers


8 Our data are from the 2003 EEOC files basedon required filings by all private-sector establish-ments with 50 or more employees and smaller firmsif they are federal contractors The calculations arebased on about 170000 individual establishmentswith any managerial workers located in identifiablemetropolitan areas (as used below) We set loggedpercent female to 0 where there were less than 1 per-cent female managers (the results were substantive-ly the same when we used unlogged percentages) Thedifferences in mean standard deviations were high-ly significant at conventional levels Details are avail-able from the authors See Cohen and Huffman(2007) and Robinson and colleagues (2005) for morerecent analyses of this data set


Hypothesis 2 There is less gender wage inequal-ity in local industries (a) that have morefemale managers and (b) in which femalemanagers on average hold higher-statuspositions

To test these hypotheses one must combinedata from several sources and organize them intoa multilevel structure Despite large data sets andsophisticated methods as is typical in large-scale studies of labor market inequality we can-not offer causal tests of these hypotheses Ratherwe test whether the data are consistent with thepatterns predicted by these hypothesesFollowing Reskin (200314) we believe thatldquoalthough contextual effects are not themselvesmechanisms they are proxies for mechanismsthat vary across settingsrdquo Next we describeour data collection and manipulation as well asour statistical modeling strategy

DATA MEASURES AND MODELS

DATA

We investigate our hypotheses by analyzingdata at three conceptual levels individual work-ers jobs and local industries We nest individ-ual workers in ldquojobsrdquo defined as three-digitoccupation by three-digit industry by metro-politan area cells (Cohen and Huffman 2003aHuffman and Cohen 2004b) Our innovation isthat employed respondents are separated intomanagerial and nonmanagerial jobsNonmanagerial workers are the subjects of ourwage analysis Each nonmanagerial job is nest-ed within a local industrymdasha three-digit indus-try in a metropolitan labor market Managercharacteristics are drawn from the managerialworkers in each local industry and used as inde-pendent variables

Our primary data source is the combined2000 Census 5- and 1-percent Public UseMicrodata Samples (PUMS) We analyze thewages of metropolitan nonmanagerial civilianworkers ages 25 to 54 years We restrict oursample to those who are employed in jobs withat least 10 people and local industries with atleast 10 managers in each of at least two man-agerial occupationsmdashthis is necessary for cal-culating female managersrsquo relative status (seebelow) This process yields a sample of approx-imately 132 million workers nested in 29294local jobs which are in turn nested in 1318 local

industries (representing 155 industries in 79metropolitan labor markets)9 Additional meas-ures for characteristics of local industries basedon their metropolitan areas are drawn from theCensus 2000 Summary Files The Dictionary ofOccupational Titles (DOT) which includesmeasures of occupational skills and require-ments is the final source of data (Tomaskovic-Devey and Skaggs 2002) Appending thesemeasures to occupations in the 2000 Censusrequired converting occupation codes from the1990 scheme to the new coding schemeemployed in 200010


9 We exclude workers who were self-employed inmilitary-specific occupations or in the armed forcesbecause their wages and promotions are not deter-mined by local managers We also exclude thosewith wages outside the range of $1 to $300 per hourlegislators for whom we have no occupational char-acteristics and those with no specific metropolitanarea identified (mostly rural workers) The age restric-tion is applied after job cell characteristics are cal-culated (so that all workers contribute to jobcharacteristics not only those for whom the out-come is analyzed) The job and local industry restric-tions reduce the final sample from 345 to 132million workers mostly by removing workers fromsmaller labor markets Notable differences in thefinal sample include more foreign-born workersmore concentration in the Northeast and West regionsand more highly educated workers Workers exclud-ed by job and local industry restrictions have loggedwages 18 lower than those in the final sample butthis is reduced to less than 05 when adjusted forobserved individual characteristics We have no rea-son to suspect that our selection criteria introduce sys-tematic biases with regard to our hypotheses but wecannot rule out that possibility

10 The DOT database contains information fornearly 13000 occupations corresponding to about500 occupations in the three-digit codes used by the1990 Census We matched 2000 Census occupationsto DOT occupations using a crosswalk file from theNational Occupational Information CoordinatingCommittee and calculated mean scores across DOToccupations for each Census occupation We usedNOICC Master Crosswalk v 43 (revised November15 1999) and a f ile titled ldquoDOTCEN00rdquo(ldquoCrosswalk linking the 2000 Census occupations tothose from the Dictionary of Occupational Titlesrdquo)both accessed from the National Crosswalk ServiceCenter (httpwwwxwalkcenterorg) on March 102006


MEASURES

LEVEL 1 INDIVIDUALS In our models thedependent variable is the natural logarithm of thehourly wage which is annual earnings dividedby hours worked At the individual level weinclude a binary variable for gender (female =1) and four dummy variables to representrespondentsrsquoethnicity (coded 1 if the respondentis Black Latino Asian or other ethnicity)11

White is the omitted raceethnicity We usedummy variables to capture differences in edu-cational attainment (less than high school somecollege bachelorrsquos degree or masterrsquos degree orabove) whether the respondent is married for-merly married (divorced widowed or separat-ed) and foreign born We also include dummyvariables to control for disability status and thepresence of children younger than six years inthe household as well as whether the respon-dents do not speak English well and whetherthey are currently attending school Continuousvariables measure potential labor market expe-rience (age minus years of education minus 5)and its square in addition to number of own chil-dren in the household

LEVEL 2 JOBS At the job level our inde-pendent variable of primary interest is percentfemale To account for nonlinear effects of per-cent female we include percent femalesquared12 We measure other demographic char-

acteristics of jobs including percent Black per-cent Latino and percent Asian We also controlfor the percent part-time employed in the jobFrom the DOT we include three measures stan-dard vocational preparation (SVP) generaleducational development (GED) and physicalstrength (STR) SVP measures the amount oftraining time needed to learn the techniquesand obtain the information necessary for aver-age job performance (high values of this scalerepresent a longer period of time required toacquire the skills) SVP can be thought of as ameasure of occupation-specific human capital(Tomaskovic-Devey and Skaggs 2002) whileGED measures ldquothe typical requirement of theoccupation for schooling that is not vocationallyspecif icrdquo (England Hermsen and Cotter20001742) GED is calculated as the mean ofvalues required for mathematics language andreasoning preparation Finally STR is codedto reflect strength requirements for each occu-pation ranging from 1 (sedentary) to 5 (veryheavy) This variable is intended to capture themanual nature of occupations which featuresprominently in the gender division of labor(Charles and Grusky 2004)

LEVEL 3 LOCAL INDUSTRIES At the localindustry level our key independent variables arepercent female and the relative status of femalemanagers In the absence of a direct measure ofdecision-making authority we identify man-agers as those in ldquomanagement occupationsrdquo inthe occupational classifications of the federalgovernment (US Census Bureau 2003) Thisclassification clearly is a relevant if imperfectmeasure of authority13 Percent female among


occupation percent female and median earnings at thenational level Closer examination revealed a 4th-power fit in the bivariate relationship between jobgender composition and average wages With theintroduction of controls at the individual level (whichalso captures womenrsquos lower average individualwages) the best fit was cubic and with controls atthe job level added the best fit fell to quadratic Thisdid not change with the addition of local-industrycontrols Therefore we model the job gender com-position effect as quadratic

13 We checked the Multi-City Study of UrbanInequality for three large US cities and found that65 percent of workers in managerial occupationsreport having the authority to hire and fire others

11 Because of overlapping racial and ethnic iden-tification in the 2000 Census we use a descendingselection to reach mutually exclusive categories inthe following order Latino Black Asian otherWhite Latinos are thus coded as such regardless oftheir responses to the race question and Whites arethose who selected no Latino ethnicity or other raceThis conforms to the recommendation of the feder-al Office of Management and Budget with regard tocivil rights enforcement (Goldstein and Morning2002)

12 Consistent with many findings in the literature(eg Cohen and Huffman 2003a Cotter et al 1998England et al 1994) we found a linear effect onwages of the gender composition of jobs in modelswith controls at all levels Cotter and colleagues(2004) however present evidence from the 2000Census that calls this simple relationship into ques-tion cubic and 4th-power fits for women and menrespectively in the bivariate relationship between


managers in each local industry is simplyobtained from the PUMS files using the samecriteria for selection we use for nonmanagerialworkers at level 1

To measure female managersrsquo status relativeto male managersmdashour proxy for decision-making powermdashwe use a measure of verticalsegregation the Index of Net Difference (ND)as described by Lieberson (1976) ND is the dif-ference between the likelihood that a random-ly chosen man is employed in a higher-statusmanagerial occupation than a randomly chosenwoman and the opposing probability that a ran-domly selected woman works in a higher-statusmanagerial occupation than a randomly chosenman Specifically ND is given by

ND = 100 (MiCFi ndash FiCMi)

Mi and Fi equal the proportion of males andfemales respectively in managerial occupationi CFi equals the cumulative proportion offemales in managerial occupations ranked belowmanagerial occupation i and CMi equals theanalogous cumulative proportion of men WhenND equals zero men and women are equallylikely to occupy high-status occupations NDequals 1 when all women are in higher-statusoccupations than men and when ND equals ndash1all women are in lower-status positions thanmen

Calculating ND requires an ordinal rankingof managerial occupations Unfortunatelyauthority over other workers although theoret-ically central to analyses of class dynamics atwork (Wright 1997) is not a prominent concernin most occupational research We know of nosource that reports the relative decision-makingauthority or status of managerial occupationsDespite detailed descriptions of thousands ofoccupations neither the DOT nor the morerecent federal ONet descriptor system meas-ure authority directly One approach is todichotomously code occupations from theCensus as having authority or not based on theappearance of the words ldquomanagerrdquo ldquosupervi-sorrdquo or ldquoadministrationrdquo in the occupation title(England et al 1994) This does not however

rank occupations according to the level ofauthoritymdashchief executives and food servicemanagers for example are both simply codedas possessing authority

In the absence of a direct measure of author-ity we first select occupations in the federal sys-temrsquos ldquomanagement occupationsrdquo category thenapply a measure based on both skills and train-ing (representing expertise) and earnings (rep-resenting recognition and rewards) Our measureof the status of each managerial occupation isthe average of two factors (1) the average of z-scores for years of education SVP and GED and(2) the z-score for earnings Managerial occu-pations are thus ranked according to equalweightings of the skills and training required(education SVP and GED) and also averageearnings14 In the resulting authority rankingnatural-science managers are highest and gam-ing managers are lowest15

We control for other important characteristicsof local industries that could affect genderinequality Among managers we include man-agers as a percentage of all workers This isintended to capture the level of rationalizationor bureaucratization of the local industry whichare traits associated with reduced reliance onascriptive characteristics in determining workpositions and rewards (Jackson 1998 Reskin2003)

We also include characteristics of local labormarkets at this level drawing data from thePUMS and 2000 Census Summary Files16 To


(compared to 30 percent of those in other occupa-tions) and 48 percent reported influencing the rate ofpay of others (compared to 22 percent of those in non-managerial occupations)

14 Some managerial occupations are specific to oneindustry (eg funeral directors) but most occuracross at least several industries (eg human resourcemanagers and chief executives) To calculate statuswe combined managers from all industries in alllabor markets

15 Because women are heavily represented in somehigher-status managerial occupations (eg medicaland health service managers and educational admin-istrators) and poorly represented in some lower-sta-tus positions (eg construction managers) our statusmeasure and percent female are only weakly corre-lated (r = ndash16) Across all local industries femalemanagersrsquo relative status is higher where their rep-resentation in management is greater (r = 09)

16 In principle labor market variables could con-stitute a fourth level of the analysis However becauseour hypotheses do not focus on local labor marketdynamics per se we include these variables at level


capture other aspects of local gender dynamicswe control for the overall level of occupation-al gender segregation and the demand for femalelabor Both of these measures are computedacross the 33 major occupational categoriesreported in the Summary Files The segregationmeasure uses the index of dissimilarity (Duncanand Duncan 1955) This measure has a signif-icant association with the gender gap in earn-ings across US labor markets such that thosewith higher levels of segregation have lowerrelative wages for women regardless of whetherthey work in male- or female-dominated occu-pations or jobs (Cohen and Huffman 2003bCotter et al 1997) The demand measure reflectsthe number of women who would be employedif local occupations had the gender compositionobserved nationally divided by the actual num-ber of employed women Labor markets withhigher levels of demand for female labor alsohave significantly less gender wage inequality(Cotter et al 1998) Our measure is similar tothat used by Cotter and colleagues (1998)although with less occupational detail Thesetwo controls are especially important if we areto avoid spurious effects whereby wage gaps andwomenrsquos managerial representation are bothmore favorable as a result of larger factorsaffecting all women in the local labor market17

To capture other local economic conditionswe include the unemployment rate and industrialcomposition measured by percentage in man-ufacturing among all workers Region is con-trolled with dummy variables for the SouthWest and Midwest (Northeast is omitted) Wealso include demographic characteristics thenatural logarithm of population size the per-

centage Black and percentage Latino and therate of in-migration measured by the percent-age of local residents who moved to the metro-politan area in the previous five years

Descriptive statistics at each level of theanalysis appear in Table 1 which shows that menin our sample have average logged wages of 284($1713) compared to 267 ($1442) for womenfor a gender wage ratio of 84 Managerial com-position ranges from 2 to 90 percent with meansof 34 percent for men and 48 percent for womenFemale managersrsquo relative status has a mean ofndash075 and ranges from ndash72 to 79

ILLUSTRATIVE EXAMPLES

Several examples clarify our data structure andits applicability to our hypotheses Table 2 showsa comparison across four local industries restau-rants and computer systems in the Los Angelesand New York metropolitan areas18 Our sam-ple has 1887 managers in Los Angeles restau-rants Of these food service managers are themost common (N = 1450) This local industryalso has 44 nonmanagerial occupations (jobs)with 10 workers or more and a total of 10422workers the plurality of whom are cooks (N =2928)

Our analytic strategy works on the assump-tion that the behavior of managers in the localindustrymdashwhich is influenced by their gendercomposition and the relative status of the womenamong themmdashmay influence the wages of thenonmanagerial workers under them In therestaurant example we see that there are morefemale managers in Los Angeles (424 percent)than in New York (334 percent) The femalemanagers in New York however have higherrelative status than those in Los Angeles most-ly because they are less concentrated in the low-status food service manager occupation TheND score for New York (021) is thus higherthan that for Los Angeles (ndash047) Among non-managerial workers New York has fewer women(407 versus 454 percent) and the gender wageratio among nonmanagerial workers is higher inNew York where women earn 95 percent of


3 which allows us to use the HLM software (seebelow) for computing and greatly simplifies com-putation and interpretation

17 In our sample female representation in man-agement is negatively correlated with occupationalsegregation (r = ndash06 p lt 05) but not with demandFemale managersrsquo relative status on the other handis higher in local labor markets with more occupa-tional segregation (r = 06 p lt 05) and lower inthose with more female demand (r = ndash08 p lt 01)This implies that in more integrated markets andthose with more female-dominated occupationalstructures women are more likely both to be man-agers and to be crowded into low-level managerialpositions

18 We chose the restaurant industry because it hasthe most local cases in our data set and the comput-er system industry as a male-dominated contrast



Table 1 Descriptive Statistics for Variables Used in the Analysis (by gender)

Men Women Min Max

IndividualsmdashHourly Wage (natural log) 284 267 0 570mdashLess than High School 18 11 0 1mdashHigh School Complete 23 22 0 1mdashSome College 27 32 0 1mdashBA 20 23 0 1mdashMA or Higher 12 13 0 1mdashIn School 08 10 0 1mdashWork Disability 17 15 0 1mdashWhite 58 60 0 1mdashBlack 11 15 0 1mdashAsian 08 08 0 1mdashLatino 20 15 0 1mdashOther RaceEthnicity 02 02 0 1mdashNever Married 29 25 0 1mdashMarried 57 56 0 1mdashFormerly Married 14 18 0 1mdashOwn Children in Household 90 90 0 12mdashChildren Under 6 in Household 23 19 0 1mdashEnglish (not spoken well) 17 11 0 1mdashForeign Born 30 24 0 1mdashPotential Experience 1889 1923 0 48mdashPotential Experience Squared 43273 44885 0 2304JobsmdashProportion Female 28 69 0 1mdashProportion Part-Time 13 23 0 1mdashProportion Black 11 15 0 1mdashProportion Latino 20 15 0 1mdashProportion Asian 07 08 0 1mdashGeneral Educational Development 326 354 100 6mdashSpecific Vocational Preparation 566 564 100 822mdashStrength 227 182 100 5Local IndustriesmdashManager Percent Female 3385 4780 233 9000mdashFemale Manager ND ndash07 ndash08 ndash72 79

mdashND Percent Female Interaction ndash252 ndash343 ndash5172 4521mdashPercent Managers 1019 995 120 3772mdashMA Gender Segregation 38 38 34 48mdashMA Female Labor DemandSupply 101 101 92 104mdashNortheast 26 30 0 1mdashMidwest 15 14 0 1mdashSouth 26 27 0 1mdashWest 32 29 0 1mdashMA Population (ln) 1581 1587 1192 1687mdashMA Percent Black 1320 1359 49 3995mdashMA Percent Latino 1670 1632 63 4766mdashMA Unemployment 589 589 345 1198mdashMA Percent in Manufacturing 1242 1228 366 3941mdashMA Percent In-migration 3002 2975 2274 3562

Source 2000 US Census Public Use Microdata files and other sources (see text)Notes MAs are (Consolidated) Metropolitan Statistical Areas ND is the index of net difference showingwomenrsquos status relative to men 1 = all men in top positions and 1 = all women in top positions (see text)All gender differences are significant at p lt 001 except Asian (ns) and own children in household (p lt 05)



Table 2 Managerial and Nonmanagerial Occupations in Four Local Industries

Los Angeles New York T-tests

Percent Percent PercentN Female Wage N Female Wage Female Wage

Restaurants and Other Food ServicesmdashLargest Managerial OccupationsmdashmdashChief Executives 31 226 6430 30 300 6506mdashmdashGeneral amp Operations Managers 179 302 1930 154 299 1975mdashmdashMarketing amp Sales Managers 61 443 3097 26 500 2190mdashmdashHuman Resources Managers 110 482 1533 65 446 1241mdashmdashFood Service Managers 1450 437 1527 1681 329 1846 mdashmdashTotal (including occupations not shown) 1887 424 1733 2014 334 1931 mdashmdashGender ND ndash047 021mdashLargest Nonmanagerial OccupationsmdashmdashCashiers 827 809 1100 701 783 1124mdashmdashWaiters amp Waitresses 2529 660 1218 2824 616 1244 mdashmdashFood Preparation Workers 444 480 1018 551 427 1017mdashmdashSupervisors Food Prep amp Service Workers 717 499 1441 627 408 1401 mdashmdashBartenders 304 359 1286 452 467 1362 mdashmdashCooks 2928 295 1188 2384 188 1112 mdashmdashAttendants amp Bartender Helpers 401 145 1012 303 274 982 mdashmdashChefs amp Head Cooks 482 140 1404 1215 111 1419mdashmdashDishwashers 255 137 923 305 112 842mdashmdashDriverSales Workers amp Truck Drivers 316 130 1198 237 80 1202mdashmdashTotal (including occupations not shown) 10422 454 1219 11025 407 1224 mdashmdashmdashWomen 4730 1166 4487 1187mdashmdashmdashMen 5692 1265 6538 1250mdashmdashmdashGender Wage Gap 922 950

Computer Systems Design and Related ServicesmdashLargest Managerial OccupationsmdashmdashChief Executives 89 180 5535 108 185 6340mdashmdashGeneral and Operations Managers 45 222 4413 57 193 4839mdashmdashComputer amp Information Systems Managers 115 252 3236 216 245 4028 mdashmdashFinancial Managers 24 458 3218 44 386 3819mdashmdashMarketing amp Sales Managers 84 369 3069 132 394 4092 mdashmdashTotal (including occupations not shown) 589 289 3749 893 315 4306 mdashmdashGender ND ndash140 ndash196mdashLargest Nonmanagerial OccupationsmdashmdashSales Representatives 87 276 3741 131 313 4324mdashmdashComputer Support Specialists 92 315 2412 128 258 2721mdashmdashComputer Programmers 352 239 3340 765 242 3430mdashmdashComputer Software Engineers 428 196 3403 665 224 3792 mdashmdashComputer Scientists amp Systems Analysts 296 206 3110 617 201 3610 mdashmdashNetwork Systems amp Data Comm Analysts 155 174 2199 238 172 3365 mdashmdashTotal (including occupations not shown) 2104 300 2882 3586 298 3294 mdashmdashmdashWomen 632 2533 1069 2847 mdashmdashmdashMen 1472 3032 2517 3483 mdashmdashmdashGender Wage Gap 835 817

Source 2000 US Census Public Use Microdata filesNotes Full names are Los Angeles-Riverside-Orange County and New York-Northern New Jersey-Long IslandManagerial occupations sorted by status nonmanagerial occupations sorted by percent female Wages are meansND is the index of net difference (see text) p lt 05 (two-tailed)


menrsquos average wages compared to 922 per-cent in Los Angeles

The computer systems industry in these twolabor markets is much more male dominatedhas higher wages than restaurants and featuresmore striking gender inequality In this casefemale managers are more common in NewYork than in Los Angeles but New York hasmore women in lower-status marketing managerjobs and fewer in higher status positions Amongnonmanagerial workers the gender wage ratiois lower in New York than in Los Angeles (82versus 84) The similar gender distributionacross jobs suggests this gap mostly reflectsinequality within jobs rather than job segrega-tion

Continuing the restaurant example our dataset is adequate to analyze 58 local industriesthat is the restaurant industry in 58 local labormarkets Figure 1 shows the relationshipbetween manager percent female and the gen-der wage ratio among nonmanagerial workersin these local industries with the largest met-ropolitan areas highlighted For illustration wesplit the local industries at the median genderND score (ndash047) and show each half in a sep-arate plot In the low-ND marketsmdashwherefemale managers are in lower-status positionsrelative to male managersmdashthere is no rela-tionship between percent female and the genderwage ratio In the high-ND markets howeverlocal industries with more women in manage-ment exhibit a smaller gender gap In this indus-try the pattern across labor markets suggests thatthe effect of higher female representation amongmanagers may be conditional on their attainmentof higher-status managerial positions

These examples also illustrate an issue wementioned abovemdashthat female managers maybe ghettoized in the same industries wherefemale workers are concentrated producing anegative relationship between female manage-rial concentration and average wages for bothfemale and male workers19 Table 3 displays

the distribution of male and female workersacross categories of female manager represen-tation as well as median wages and femalemanagersrsquo relative status The table shows that318 percent of men but only 65 percent ofwomen work in local industries with fewer than20 percent female managers typified by theconstruction industry On the other hand morethan one-quarter of women but only 8 percentof men work in local industries where 60 per-cent or more of the managers are female typi-fied by hospitals Thus the gender of workersand managers is clearly related Note that themost common situation involves female mana-gerial representation between 20 and 40 percent(eg restaurants) This is the group of localindustries in which female managers have thelowest relative status (ND = ndash14) and the gen-der gap in median wages is greatest with non-managerial women earning just 75 percent of themale wage We now investigate these relation-ships on a larger scale taking into account pos-sible confounding factors at the levels of theperson job and local industry

STATISTICAL MODELS

Our data have a nested structure with individ-ual workers nested within jobs which in turn arenested within local industries We therefore usethree-level hierarchical linear models(Raudenbush and Bryk 2002) which are justi-fied both on statistical and substantive groundsFirst hierarchical models avoid downwardlybiased estimation of standard errors when dataare nested (Guo and Zhao 2000 Raudenbushand Bryk 2002) Additionally they provide theflexibility to specify cross-level interactioneffects on which our hypotheses are based Forexample is the individual gender effect strongeror weaker depending on the gender compositionof the job and its local managers

Specifically in our individual-level modellogged wages are a function of a model inter-cept (average wages for men) a female effect(within-job gender inequality) and controlsFormally our individual-level model can beexpressed as

Yijk = 0jk + 1jk(femaleijk) + 2jka1ijk

+ || + mjkamijk + eijk

where Yijk is the logged wage of person i injob j in local industry k and 0jk is the inter-


19 In fact the data show a small positive bivariatecorrelation between female managerial representationand (logged) wages for both men (r = 09) and women(r = 07) A closer inspection though shows this isbecause a cluster of male-dominated blue-collar jobssuch as construction have relatively low wages andalmost no female managers


cept for job j in local industry k Next 1jk isthe individual-level effect of gender amijk

denote the M individual-level control vari-ables and 2jk through mjk are the associat-

ed individual-level regression coefficientsFinally ejk is the level-1 random effect Thecoeff icient 1jk is of particular interestbecause it represents the net within-job wage


Figure 1 Restaurant Industry Manager Percent Female and the Gender Wage Gap AmongNonmanagerial Workers by Female Managersrsquo Relative Status in 58 Labor Markets

Notes Local industries split at median ND (ndash047) N = 29 in each figure Correlations and regression lines weight-ed by sample size Nrsquos range from 51 (Santa Barbara CA) to 2886 (New York) the largest 10 labor markets arelabeled Wage ratio is womenrsquos median wage as percent of menrsquos


gap To simplify interpretation of the resultsall variables except gender are centeredaround their grand means Thus 0jk repre-sents the average wage for men at the meanof the control variables

Our job-level model estimates both the level-1 intercept and the level-1 female effect as afunction of job percent female (and its square)and our job-level controls Each level-1 coeffi-cient relating individual characteristics to wagescan be modeled as either a fixed or randomeffect across jobs We allow only the level-1intercept and the coefficient for the femaledummy variable to vary across jobs Thus ourlevel-2 model is

0jk = 00k + 01k(job femalejk) +02k(job female2

jk) + 03kX1jk + || +0qkXqjk + r0jk


jk) + 13kX1jk +|| + 1qkXqjk + r1jk

where 00k is the intercept for the job-levelmodel in local industry k In turn 01k and 02kare the effects of job proportion female and itssquare on 0jk Likewise 10k is the job-levelintercept for the effect of being female 1jkwhile 11k and 12k are the effects of job pro-portion female and its square on the level-1effect of being female (these are cross-levelinteraction effects) Finally X1jk through Xqjkdenote the Q control variables in each job-levelmodel These control variables are centeredaround their grand means The level-2 errorterms are denoted by r0jk and r1jk These error

terms mean that the coefficients for the inter-cepts and job composition are within-local-industry effects just as the intercept and gendercoefficient in the individual-level model arewithin-job effects

Finally each level-2 coefficient relating jobcharacteristics to level-1 effects on wages canbe modeled as either a random or a fixed effectacross local industries In our models only thelevel-2 intercept and the job proportion female(and its square) coefficients are permitted tovary across local industries Thus our level-3model is

00k = 000 + 001(mngr femalek) + 002(mngr NDk) + 003

(mngr femalek mngr NDk) + || + 00sWsk + u00k

01k = 010 + 011(mngr femalek) + 012(mngr NDk) + 013(mngr femalek

mngr NDk) + || + 01sWsk + u01k

02k = 020 + 021(mngr femalek) +022(mngr NDk) + 023(mngr femalek







Table 3 Nonmanagerial Worker Distribution and Wages by Manager Percent Female in LocalIndustries

Worker Distribution Median Wage

Manager Median WagePercent Female ND Total Men Women Total Men Women Ratio Most Common Industry

le 19 ndash022 194 318 65 1500 1538 1288 84 Construction20 to 39 ndash140 257 310 203 1692 1923 1442 75 Restaurants40 to 59 ndash074 377 292 465 1563 1779 1451 82 Kndash12 Schoolsge 60 ndash050 172 80 268 1538 1700 1496 88 HospitalsTotal ndash068 1000 1000 1000 1577 1731 1442 83

Source 2000 US Census Public Use Microdata filesNotes Worker distribution and wages are measured at the individual level Manager ND is the median for localindustries within each category weighted by the number of workers in each local industry Most common indus-tries are those with the most total workers in each category shown for illustration ND is the index of net differ-ence (see text)




where 000 010 020 100 110 and 120 are thelevel-3 intercepts in models of the level-2 coef-ficients 001 011 021 101 111 and 121 arethe effects of the gender composition of man-agers on the level-2 coefficients Likewise 002021 022 102 112 and 122 denote the effectsof the status of female managers (ND) on thelevel-2 coefficients Finally 003 013 023103 113 and 123 are manager percent femaleby manager relative status interaction termsCoefficients for the S level-3 control variables(W) are denoted by the remaining terms They

are centered around their grand means Thelevel-3 error terms are given by u for each of thek local industries

RESULTS

Results from the hierarchical linear modelappear in Table 4 We show the coefficientsonly for key variables complete results areavailable from the authors The variance com-ponents for several models are presented in theAppendix

Recall that in our models gender is associat-ed with individualsrsquowages through three distinctpathways individual gender the gender of theco-workers in their jobs and the gender of the


Table 4 Coefficients for Three-Level Hierarchical Linear Regressions for Logged Wages OnIndividual Job and Local-Industry Characteristics

Level 1Individual Level 2 Level 3Variables Job Variables Local-Industry Variables Coefficient t-statistic

INTERCEPT 2888 23685Manager Percent Female ndash002 ndash643Index of Net Difference ndash056 ndash122Manager Percent Female Net Difference 002 166

Job PercentFemale 124 227

Manager Percent Female 001 52Index of Net Difference 324 146Manager Percent Female Net Difference ndash018 ndash301

Job PercentFemale2 ndash385 ndash600

Manager Percent Female 002 109Index of Net Difference ndash268 ndash99Manager Percent Female Net Difference 017 255

FEMALE ndash126 ndash798Manager Percent Female ndash0003 ndash68Index of Net Difference 026 40Manager Percent Female Net Difference ndash002 ndash115

Job PercentFemale ndash270 ndash372

Manager Percent Female 004 209Index of Net Difference ndash456 ndash147Manager Percent female Net Difference 023 272

Job PercentFemale2 327 434

Manager Percent Female ndash004 ndash217Index of Net Difference 475 146Manager Percent Female Net Difference ndash023 ndash285

Source 2000 US Census Public Use Microdata files and other sources (see text)Notes This model includes control variables at all three levels (see Table 1) Coefficients in the lower panel onthe individual female coefficient are cross-level interactions showing effects on the wage difference betweenmen and women p lt 05 p lt 01 p lt 001 (two-tailed tests)


managers in their local industries At level 1 thegender of individuals is associated with wage dif-ferences within jobs At level 2 the gender com-position of jobs is associated with the averagewages of men and women across those jobswithin local industries This effect may be dif-ferent for women and men and may be nonlin-ear At level 3 the gender characteristics ofmanagers are associated with overall wage lev-els as well as with the job- and individual-leveleffects The variables at each level are labeled inthree columns in Table 4 Because the numberof interactions complicates the interpretation ofthe results we only briefly discuss the coeffi-cients before moving to a plot of predicted wagesfor men and women under plausible scenarios

For men the predicted wage at the mean ofall control variables is the model intercept 2888in logged dollars or $1796 per hour This reflectsan all-male job with no female managers and anet difference of zero because those variables arenot centered The individual-level gender effectis ndash126 which means that women in the samejob are predicted to earn 2888 ndash 126 = 2762or $1583 for a gender wage ratio of 88 Toextend the example simply if 50 percent of themanagers in that local industry are women thosemanagers are of equal status as the male man-agers (ND = 0) and the job remains all-malethen the predicted wage for men would be themodel intercept plus the manager percent femaleeffect 2888 + (ndash002 50) = 2788 or $1625To calculate the predicted wage for women insuch a job we start with menrsquos wage then addthe individual gender effect as modified by themanager composition of 50 percent female2788 + [ndash126 + (ndash0003 50)] = 2747 or$1411 for a gender wage ratio of 87

This simple example is not realistic howev-er because it reflects men and women workingin an all-male job with equal status betweenmale and female managers When we assessthe gender gap at different levels of job andmanagerial gender composition and femalemanagersrsquo relative status a more conclusiveresult emerges Specifically managerial com-position and relative status largely work on thegender gap through the effects of job composi-tion20 Note the significant effects of manager

percent female and its interaction with ND onthe level-2 variables in the lower half of thetable The net result is that female managers areassociated with a reduced gender wage gapespecially when those female managers holdrelatively high-status positions We will illustratethe results graphically rather than walk througha summary of many interactions and nonlineareffects

Figure 2 shows predicted wages (the y-axis)for male and female workers in jobs of typicalgender composition (30 percent female for men70 percent female for women) working undermanager compositions ranging from 0 to 80percent female (the x-axis) with relative gen-der status one standard deviation above andbelow the mean (one line for each) The pre-dictions show menrsquos wages are consistentlylower where the percentage female among man-agers is higher (consistent with the suggestionthat women managers are concentrated in lesshighly-paid industries) Among female workerswages fall as a function of female managementwhere those managers are of low status butwages rise where female managers are of highstatus For our focus on the gender wage gapthe result is clear Net of controls at all three lev-els the higher presence of female managers inlocal industries is associated with a reducedgender wage gap among nonmanagerial work-ers only where those managers hold relativelyhigh-status positions In the figure the genderwage ratio is constant at 81 where female man-agersrsquo relative status is one standard deviationbelow the mean but narrows from 76 to 99when that status is one standard deviation abovethe mean and manager composition rises to 80percent closing the gender gap21

The results support both of our hypothesesFirst on average (ie midway between the highand low lines in Figure 2) the gender gap issmaller in local industries where a high pro-


20 The baseline effect of job composition is suchthat as female representation at the job level rises

menrsquos wages start slightly upward but then turnsharply downward after 16 percent female (the inflec-tion point is [ndash1 124][2 ndash385] = 16) The non-linearity is much less pronounced for womenresulting in a negative net effect of female job com-position

21 Such extreme cases are rare but do exist In thesample 90 percent of employees work under mana-gerial pools that are between 12 and 67 percentfemale


portion of the managers are women This effectis small though and results primarily from anegative effect on average (menrsquos) wages net offemale managersrsquo relative status womenrsquos wagesare largely unaffected by manager percentfemale The results also strongly support oursecond hypothesis which posits an interactionbetween the representation and relative status offemale managers Specifically we hypothesizeless gender wage inequality in local industrieswith (1) more female managers and (2) femalemanagers averaging higher-status positionsThis is consistent with our data The genderwage gap is smaller under female managersand this effect is much stronger when thosefemale managers are of relatively high statusThus although female managers may boostwomenrsquos wages relative to men the represen-tation of low-status female managers alone doesnot affect the gender wage gap It is represen-tation in upper-status managerial positions thatis associated with the gender wage gap

ENDOGENEITY

Although the statistical relationship is strongpotential sources of error or bias exist in ouranalysis In particular endogeneity and theomission of potentially relevant control vari-ables may be sources of bias although our

research design and choice of control variableshelp to minimize their effects Endogeneitycould confound our results to the extent thatunmeasured variables are driving both the gen-der wage gap and the representation of womenin management However we include controlsfor key indicators of gender inequality includ-ing gender segregation and demand for femalelabor in the local labor market as well as aproxy for the level of rationalization or bureau-cratization of the local industry which is like-ly to reduce the reliance on ascriptivecharacteristics in determining work-relatedrewards Differences across local industries inendogenous factors that drive both the wagegap and female representation in managementare likely to be captured by these controlsControls at the job and individual levels playsimilar roles In addition because the job-levelintercepts of our models are permitted to varyacross local industries and the individual-levelintercepts vary across jobs our analysis accountsfor unobserved factors influencing averagewages across local industries In essence thenour results reflect within-job and within-local-industry effects This improves confidence inour results


Figure 2 Predicted Wages for Men and Women by Manager Percent Female and Relative Status


ALTERNATIVE SPECIFICATION

An additional problem could result from ourdefinition of managerial authority Identifyingmanagers using occupational classifications isperhaps more straightforward than identifyingthose who are not managerial Some peopleidentified as nonmanagerial in our sample (egdoctors) may wield considerable authority insome cases more authority than those countedas managers (eg administrative service man-agers in doctorsrsquo offices) This represents thelong-standing problem of intermediate classlocations that preoccupies some analysts ofclass (eg Wright 1997) Excluding self-employed workers presumably helps reduce thenumber of such cases but many workers in pro-fessional occupations remain who have ambigu-ous authority relations

To verify that our results are not unduly influ-enced by such ambiguity we reestimated ourregression models but excluded from the non-managerial sample those in ldquobusiness and finan-cial operationsrdquo or ldquoprofessional and relatedrdquooccupations This reduced the worker sample bymore than one-third (39 percent of women and29 percent of men) In fact three of the largestoccupations in our original sample were exclud-ed elementary and middle school teachers reg-istered nurses and accountants and auditors(these three occupations alone accounted foralmost 10 percent of the original sample) Thelargest remaining nonmanagerial occupationswere unambiguously nonmanagerial secre-taries truck drivers and customer service rep-resentatives together compose 13 percent of thereduced sample Results from this alternativespecification (not shown but available fromthe authors upon request) are substantively iden-tical to those reported above All the centralcoefficients are of the same or larger magnitudeand remain statistically significant

DISCUSSION AND CONCLUSIONS

After analyzing a survey of managersrsquo attitudesnearly two decades ago Brenner and colleagues(1989668) concluded ldquoUnlike her male coun-terpart todayrsquos female manager would beexpected to treat men and women equally inselection promotion and placement decisionsrdquoUnfortunately as they noted women in man-agement mostly held lower- and middle-man-agement positions where their impact was

limited Despite two decades of movementtoward managerial integration we still do nothave conclusive evidence that the entry of somewomen into managerial positions has broughtmaterial benefits to the majority who remainbelow

To address this question we offer the firstlarge-scale analysis using nationally represen-tative data on workers tied to managers andincluding theoretically relevant variables at thelevel of the individual job and local industryWe consider the relative status of female man-agers as well as their numeric representation inmodels of the gender wage gap Specifically wetest two hypotheses (1) female workers earnmore when their local industries include morewomen among the managerial ranks and (2)such representation is more beneficial whenthe relative status of female managers is high-er

Although we cannot draw causal conclusionsfrom our data our results are consistent with theargument that female managers do matter In themodels the representation of women in man-agement reduces the wage gap The interactionbetween managerial gender composition andfemale managersrsquo relative status howevershows that the relationship is much stronger inlocal industries where female managers hold rel-atively high-status positions The addition ofwomen at the low end of the managerial hier-archy may have weak effects on the genderwage gap but the potential effects of high-sta-tus female managers are much more positive

This finding highlightsmdashin a new waymdashthesignificance of the ldquoglass ceilingrdquo If our find-ings hold not only are qualified women blockedfrom upper-level managerial positions anddenied the benefits of those jobs but theirabsence has ripple effects that shape workplaceoutcomes for nonmanagerial women as well Tothe extent that women continue to cluster at thelow end of managerial hierarchies our find-ings may temper the optimism generated by therapid increase in the proportion of women inmanagement in the last several decades We arealso given pause by the finding that with all thecontrols in the model wages for men are lowerin local industries with more female managersThis suggests that female managers remain con-centrated in workplace settings with lowerwages across the board in ways that we cannotcapture with the variables used here On the



other hand our findings imply that inroadsmade by women into upper-status managerialpositions will ldquolift all boatsrdquo by also boostingthe wages of women employed in nonmanage-rial occupations All women may benefit fromthe desegregation of managerial occupationseven those who do not themselves attain suchpositionsmdashwhich compliments Cotter and col-leaguesrsquo (1997) finding that all women benefitfrom occupational desegregation

In addition our results are consistent withJacobsrsquos (1992) notion of ldquotitle inflationrdquomdashthereclassification of previously nonmanagerialworkers as managers with little change inauthority or wagesmdashin that the mere represen-tation of women in management has limitedeffects on the gender wage gap Importantly ouranalysis shows that simply looking at the per-centage of females among managers is notenough if one is interested in the effects of man-agerial access In this case including the rela-tive status of female managers is necessarymdashthis assertion garners strong support from ourmultivariate results

Our results are intrinsically sociological andbroadly provocative We hope they will enticeothers to investigate not only the role of man-agers but also analogous cases of subordinategroup members in positions of authority Forexample in studies of political alienation andparticipation research shows that the presenceof Black (Bobo and Gilliam 1990) and Latino(Pantoja and Segura 2003) elected leadersincreases political empowerment among mem-bers of those groups Female and minority polit-ical leaders appear to be more responsive to theconcerns of their ascriptively similar con-stituents (Bratton and Haynie 1999 Mansbridge1999) but the long-term implications of suchintegration remain to be seen In education stu-dents receive more positive evaluations andfeedback from teachers who match theirraceethnicity and gender but the evidence thatthis results in better educational outcomes is lim-ited and mixed (Butler and Christensen 2003Dee 2005 Downey and Pribesh 2004Ehrenberg Goldhaber and Brewer 1995)22 As

with our examination of labor markets thesestudies confront possible confounding effectsacross multiple levels of social interactionResearchers in these disparate fields shouldlearn from each other

Further theorizing and data collection effortsundoubtedly will help to overcome the weak-nesses in our case For example we are not ableto link workers and managers in their actualwork settings Just as workplace processes gen-erating inequality vary across organizationalsettings (Baron and Newman 1990 Huffmanand Velasco 1997) female managersrsquo ability toalter wage setting practices also may varymarkedly across workplace contexts On theother hand in our analysis the presence andstatus of female managers is measured at thelevel of local industries which may be the rel-evant organizational field within which employ-ers make crucial gender-related decisions Thispermits us to extend existing research on thegender wage gap to incorporate managerialcharacteristics using high-quality large-scaleCensus data that links population-based samplesof workers to managers We hope this innova-tion and these results will generate increasedattention to the role managers play in produc-ing and sustaining labor market inequality andby extension to the potential influence of sub-ordinate group members who attain positions ofauthority

Philip N Cohen is an associate professor of sociol-ogy at the University of North Carolina at Chapel Hilland a faculty fellow at the Carolina PopulationCenter His research concerns social inequality inlabor markets and families With Matt Huffman heis currently investigating the underrepresentation ofwomen and minorities in management and its con-sequences Another area of his research focuses onfamily structure and inequality within and betweenfamilies across social contexts including householdliving arrangements womenrsquos employment andunpaid labor

Matt L Huffman is an associate professor of soci-ology and Co-Director of Graduate Studies at theUniversity of California Irvine His research focus-es on race and gender inequality in work organiza-tions and across labor markets Among other thingsthis work examines the wage effects of segregation


addressing these issues should be high on the prior-ity list of those concerned with the progress of womenand minorities in the labor marketrdquo (p 560)

22 Ehrenberg and colleagues (1995548) note ldquoTherelationships between supervisors and employees isanalogous in important respects to that betweenteachers and studentsrdquo They conclude ldquoResearch


REFERENCES

Ashenfelter Orley and Timothy Hannan 1986 ldquoSex

Discrimination and Product Market Competition

The Case of the Banking Industryrdquo QuarterlyJournal of Economics 101149ndash74

Baron James N Brian S Mittman and Andrew ENewman 1991 ldquoTargets of Opportunity


gender and racial differences in access to manage-rial jobs and the consequences of changing mana-gerial composition for inequality among workers

APPENDIX

VARIANCE COMPONENTS

The hierarchical linear models permit us todecompose the total variance in wages intothree component parts (1) that which occursbetween individuals and the mean wages intheir jobs (2) that which occurs between meanwages in each job and the mean wages for alljobs in their local industries and (3) that whichoccurs between the mean wages for local indus-tries Results from the unconditional model(which includes no independent variables atany level) in Table A show that overall 723 per-cent of wage variation among nonmanagerialworkers occurs within jobs as we have definedthem 148 percent occurs between jobs butwithin local industries and 129 percent occursacross local industries

This construction differs from previous three-level wage decompositions (eg Cohen and

Huffman 2003a) in its use of local industries asthe third level rather than metropolitan labormarkets Not surprisingly variation across localindustries accounts for a larger share of the totalvariation than does metropolitan-area labor mar-kets The differences between average wages forthe restaurant industry versus the computer sys-tems industry within a given labor market wouldbe expected to be larger than differences inaverage wages between New York and LosAngeles overall The variance components alsoshow that our variables do much more to explainvariance between jobs and local industries (vari-ance components are reduced by about 80 per-cent each from the unconditional model to thefinal model) than they do to explain variationbetween individuals which is only reduced 8percent in the full model Finally Table A showsthat about three-fourths of the variance in theeffect of being female occurs across jobs butwithin industries In the final model 42 percentof that within-local-industry variance isexplained as the variance component is reducedfrom 0104 to 0068

Table AmdashVariance Components for Three-Level Hierarchical Linear Regressions

Unconditional UM +Model (UM) Female Dummy Full Model

Variance Percent Variance Percent Variance PercentComponent of Total Component of Total Component of Total

InterceptmdashIndividual 3459 723 3404 717 3191 923mdashJob 0708 148 0718 151 0146 42mdashLocal Industry 0620 129 0623 131 0118 34mdashTotal 1000 1000 1000FemalemdashJob 0104 751 0068 569mdashLocal Industry 0035 249 0051 431mdashTotal 1000 1000Job Percent Female (on intercept)mdashLocal Industry 0776 1000Job Percent Female Squared (on intercept)mdashLocal Industry 1031 1000Job Percent Female (on individual female effect)mdashLocal Industry 1055 1000Job Percent Female Squared (on individual female effect)mdashLocal Industry 1051 1000

Note All variance components have a two-tailed p-value of less than 05


Organizational and Environmental Determinantsof Gender Integration Within the California CivilService 1979ndash1985rdquo American Journal ofSociology 961362ndash401

Baron James N and Andrew E Newman 1990ldquoFor What Itrsquos Worth Organizations Occupationsand the Value of Work Done by Women andNonwhitesrdquo American Sociological Review55155ndash75

Baunach Dawn M 2002 ldquoProgress Opportunity andBacklash Explaining Attitudes Toward Gender-Based Affirmative Actionrdquo Sociological Focus35345ndash62

Beckman Christine M and Damon J Phillips 2005ldquoInterorganizational Determinants of PromotionClient Leadership and the Attainment of WomenAttorneysrdquo American Sociological Review70678ndash701

Blum Terry C Dail L Fields and Jodi S Goodman1994 ldquoOrganization-Level Determinants ofWomen in Managementrdquo Academy ofManagement Journal 37241ndash68

Bobo Lawrence and Franklin D Gilliam Jr 1990ldquoRace Sociopolitical Participation and BlackEmpowermentrdquo American Political ScienceReview 84377ndash93

Bratton Kathleen A and Kerry L Haynie 1999ldquoAgenda Setting and Legislative Success in StateLegislatures The Effects of Gender and RacerdquoJournal of Politics 61658ndash79

Brenner O C Joseph Tomkiewicz and Victoria ESchein 1989 ldquoThe Relationship Between Sex-Role Stereotypes and Requisite ManagementCharacteristics Revisitedrdquo Academy ofManagement Journal 32662ndash69

Butler Daniel M and Ray Christensen 2003ldquoMixing and Matching The Effect on StudentPerformance of Teaching Assistants of the SameGenderrdquo Political Science and Politics 36781ndash86

Carrington William J and Kenneth R Troske 1995ldquoGender Segregation in Small Firmsrdquo Journal ofHuman Resources 30503ndash33

Charles Maria and David B Grusky 2004Occupational Ghettos The Worldwide Segregationof Women and Men Stanford CA StanfordUniversity Press

Cohen Lisa E Joseph P Broschak and Heather AHaveman 1998 ldquoAnd Then There Were More TheEffect of Organizational Sex Composition on theHiring and Promotion of Managersrdquo AmericanSociological Review 63711ndash27

Cohen Philip N and Matt L Huffman 2003aldquoIndividuals Jobs and Labor Markets TheDevaluation of Womenrsquos Work Across US LaborMarketsrdquo American Sociological Review68443ndash63

mdashmdashmdash 2003b ldquoOccupational Segregation and theDevaluation of Womenrsquos Work Across US LaborMarketsrdquo Social Forces 81881ndash908

mdashmdashmdash 2007 ldquoBlack Underrepresentation inManagement Across US Labor Marketsrdquo Annalsof the American Academy of Political and SocialScience 609(January)181ndash99

Correll Shelley J and Stephen Benard 2005ldquoGetting a Job Is There a Motherhood PenaltyrdquoPresented at the Annual Meeting of the AmericanSociological Association August San FranciscoCA

Cotter David A JoAnn DeFiore Joan M HermsenBrenda M Kowalewski and Reeve Vanneman1997 ldquoAll Women Benefit The Macro-LevelEffect of Occupational Integration on GenderEarnings Inequalityrdquo American SociologicalReview 62714ndash34

mdashmdashmdash 1998 ldquoThe Demand for Female LaborrdquoAmerican Journal of Sociology 1031673ndash712

Cotter David A Joan M Hermsen Seth Ovadia andReeve Vanneman 2001 ldquoThe Glass CeilingEffectrdquo Social Forces 80655ndash81

Cotter David A Joan M Hermsen and ReeveVanneman 2004 Gender Inequality at Work NewYork Russell Sage Foundation

Deaux Kay 1985 ldquoSex and Genderrdquo Annual Reviewof Psychology 3649ndash81

Dee Thomas S 2005 ldquoA Teacher Like Me DoesRace Ethnicity or Gender Matterrdquo AmericanEconomic Review 95158ndash65

Denmark Florence L 1993 ldquoWomen Leadershipand Empowermentrdquo Psychology of WomenQuarterly 17343ndash56

DiMaggio Paul J and Walter W Powell 1983 ldquoTheIron Cage Revisited Institutional Isomorphismand Collective Rationality in OrganizationalFieldsrdquo American Sociological Review 48147ndash60

Dobbins Gregory H and Stephanie J Platz 1986ldquoSex Differences in Leadership How Real AreTheyrdquo Academy of Management Review11118ndash27

Downey Douglas B and Shana Pribesh 2004ldquoWhen Race Matters Teachersrsquo Evaluations ofStudentsrsquo Classroom Behaviorrdquo Sociology ofEducation 77267ndash82

Duncan Otis Dudley and Beverly Duncan 1955 ldquoAMethodological Analysis of Segregation IndexesrdquoAmerican Sociological Review 20210ndash17

Ehrenberg Ronald G Daniel D Goldhaber andDominic J Brewer 1995 ldquoDo Teachersrsquo RaceGender and Ethnicity Matter Evidence from theNational Educational Longitudinal Study of 1988rdquoIndustrial and Labor Relations Review 48547ndash61

Elliott James R and Ryan A Smith 2004 ldquoRaceGender and Workplace Powerrdquo AmericanSociological Review 69365ndash86

Ely Robin J 1995 ldquoThe Power in DemographyWomenrsquos Social Constructions of Gender Identityat Workrdquo Academy of Management Journal38589ndash634

England Paula Melissa S Herbert Barbara S



Kilbourne Lori L Reid and Lori M Megdal1994 ldquoThe Gendered Valuation of Occupationsand Skills Earnings in 1980 Census OccupationsrdquoSocial Forces 7365ndash100

England Paula Joan M Hermsen and David ACotter 2000 ldquoThe Devaluation of Womenrsquos WorkA Comment on Tamrdquo American Journal ofSociology 1051741ndash51

Fields Judith and Edward N Wolff 1995ldquoInterindustry Wage Differentials and the GenderWage Gaprdquo Industrial and Labor Relations Review49105ndash20

Gilligan Carol 1982 In a Different VoicePsychological Theory and Womenrsquos DevelopmentCambridge MA Harvard University Press

Glascock Jack 2001 ldquoGender Roles on Prime-TimeNetwork Television Demographics andBehaviorsrdquo Journal of Broadcasting andElectronic Media 45656ndash69

Goldstein Joshua R and Anne J Morning 2002ldquoBack in the Box The Dilemma of Using Multiple-Race Data for Single-Race Lawsrdquo Pp 119ndash37 inThe New Race Question How the Census CountsMultiracial Individuals edited by J Perlmann andM Waters New York Russell Sage Foundation

Gorman Elizabeth H 2005 ldquoGender StereotypesSame-Gender Preferences and OrganizationalVariation in the Hiring of Women Evidence fromLaw Firmsrdquo American Sociological Review70702ndash28

Greve Henrich R 2005 ldquoInterorganizationalLearning and Heterogeneous Social StructurerdquoOrganization Studies 261025ndash47

Guo Guang and Hongxin Zhao 2000 ldquoMultilevelModeling for Binary Datardquo Annual Review ofSociology 26441ndash62

Halpert Jane A Midge L Wilson and Julia LHickman 1993 ldquoPregnancy as a Source of Biasin Performance Evaluationsrdquo Journal ofOrganizational Behavior 14649ndash63

Huffman Matt L and Philip N Cohen 2004aldquoOccupational Segregation and the Gender Gap inWorkplace Authority National Versus Local LaborMarketsrdquo Sociological Forum 19121ndash47

mdashmdashmdash 2004b ldquoRacial Wage Inequality JobSegregation and Devaluation Across US LaborMarketsrdquo American Journal of Sociology109902ndash36

Huffman Matt L and Lisa Torres 2002 ldquoIs It OnlyWho You Know That Matters Gender and SocialNetworks Among Professional Technical andManagerial Workersrdquo Gender and Society16793ndash813

Huffman Matt L and Steven C Velasco 1997ldquoWhen More Is Less Sex CompositionOrganizations and Earnings in US Firmsrdquo Workand Occupations 24214ndash44

Hultin Mia 2003 ldquoSome Take the Glass EscalatorSome Hit the Glass Ceiling Career Consequences

of Occupational Sex Segregationrdquo Work andOccupations 3030ndash61

Hultin Mia and Ryszard Szulkin 1999 ldquoWages andUnequal Access to Organizational Power AnEmpirical Test of Gender DiscriminationrdquoAdministrative Science Quarterly 44453ndash72

mdashmdashmdash 2003 ldquoMechanisms of Inequality UnequalAccess to Organizational Power and the GenderWage Gaprdquo European Sociological Review19143ndash59

Ibarra Herminia 1992 ldquoHomophily and DifferentialReturns Sex Differences in Network Structureand Access in an Advertising Firmrdquo AdministrativeScience Quarterly 37422ndash47

Jackson Robert M 1998 Destined for EqualityThe Inevitable Rise of Womenrsquos Status CambridgeMA Harvard University Press

Jacobs Jerry A 1992 ldquoWomenrsquos Entry intoManagement Trends in Earnings Authority andValues Among Salaried Managersrdquo AdministrativeScience Quarterly 37282ndash301

Jaffee Sara and Janet Shibley Hyde 2000 ldquoGenderDifferences in Moral Orientation A Meta-Analysisrdquo Psychological Bulletin 126703ndash26

Kanter Rosabeth M 1977 Men and Women of theCorporation New York Basic Books

Konrad Alison M and Jeffery Pfeffer 1991ldquoUnderstanding the Hiring of Women andMinorities in Educational Institutionsrdquo Sociologyof Education 64141ndash57

Kulis Stephen 1997 ldquoGender Segregation AmongCollege and University Employeesrdquo Sociology ofEducation 70151ndash73

Lauzen Martha M and David M Dozier 1999ldquoMaking a Difference in Prime Time Women onScreen and Behind the Scenes in the 1995ndash96Television Seasonrdquo Journal of Broadcasting ampElectronic Media 431ndash19

Lieberson Stanley 1976 ldquoRank-Sum ComparisonBetween Groupsrdquo Sociological Methodology7276ndash91

Mansbridge Jane 1999 ldquoShould Blacks RepresentBlacks and Women Represent Women AContingent lsquoYesrsquordquo Journal of Politics 61628ndash57

Massini S A Y Lewin and Henrich R Greve 2005ldquoInnovators and Imitators OrganizationalReference Groups and Adoption of OrganizationalRoutinesrdquo Research Policy 341550ndash69

McCall Leslie 2001 Complex Inequality GenderClass and Race in the New Economy New YorkRoutledge

McPherson Miller Lynn Smith-Lovin and James MCook 2001 ldquoBirds of a Feather Homophily inSocial Networksrdquo Annual Review of Sociology27415ndash44

Merton Robert K 1940 ldquoBureaucratic Structureand Personalityrdquo Social Forces 18560ndash68

Mouw Ted 2003 ldquoSocial Capital and Finding a Job



Do Contacts Matterrdquo American SociologicalReview 68868ndash98

Nelson Robert L and William P Bridges 1999Legalizing Gender Inequality Courts Marketsand Unequal Pay for Women in America NewYork Cambridge University Press

Neumark David and Wendy A Stock 2006 ldquoTheLabor Market Effects of Sex and RaceDiscrimination Lawsrdquo Economic Inquiry44385ndash419

Pantoja Adrian D and Gary M Segura 2003 ldquoDoesEthnicity Matter Descriptive Representation inLegislatures and Political Alienation AmongLatinosrdquo Social Science Quarterly 84441ndash60

Pfeffer Jeffery 1983 ldquoOrganizational DemographyrdquoResearch in Organizational Behavior 5299ndash357

Pfeffer Jeffery and Alison Davis-Blake 1987 ldquoTheEffect of the Proportion of Women on Salaries TheCase of College Administratorsrdquo AdministrativeScience Quarterly 321ndash24

Pfeffer Jeffery Alison Davis-Blake and Daniel JJulius 1995 ldquoAA Officer Salaries and ManagerialDiversity Efficiency Wages or Statusrdquo IndustrialRelations 3473ndash94

Raudenbush Stephen W and Anthony S Bryk 2002Hierarchical Linear Models Applications andData Analysis Methods 2d ed Thousand OaksCA Sage

Reskin Barbara F 2003 ldquoIncluding Mechanisms inOur Models of Ascriptive Inequalityrdquo AmericanSociological Review 681ndash21

Reskin Barbara F and Debra B McBrier 2000ldquoWhy Not Ascription OrganizationsrsquoEmploymentof Male and Female Managersrdquo AmericanSociological Review 65210ndash33

Reskin Barbara F and Catherine Ross 1992 ldquoJobAuthority and Earnings Among Managers TheContinuing Signif icance of Sexrdquo Work andOccupations 19342ndash65

Ridgeway Cecilia L 1997 ldquoInteraction and theConservation of Gender Inequality ConsideringEmploymentrdquo American Sociological Review62218ndash35

Robinson Corre L Tiffany Taylor DonaldTomaskovic-Devey Catherine Zimmer andMatthew W Irvin Jr 2005 ldquoStudying Race orEthnic and Sex Segregation at the EstablishmentLevel Methodological Issues and SubstantiveOpportunities Using EEO-1 Reportsrdquo Work andOccupations 325ndash38

Roth Louise Marie 2004 ldquoThe Social Psychologyof Tokenism Status and Homophily Processes onWall Streetrdquo Sociological Perspectives47189ndash214

Ryan Michelle K Barbara David and Katherine JReynolds 2004 ldquoWho Cares The Effect of

Gender and Context on the Self and MoralReasoningrdquo Psychology of Women Quarterly28246ndash55

Shenhav Yehouda and Yitchak Haberfeld 1992ldquoOrganizational Demography and InequalityrdquoSocial Forces 71123ndash43

Smith Ryan A 2002 ldquoRace Gender and Authorityin the Workplace Theory and Researchrdquo AnnualReview of Sociology 28509ndash42

South Scott J and Weiman Xu 1990 ldquoLocalIndustrial Dominance and Earnings AttainmentrdquoAmerican Sociological Review 55591ndash99

Strang David and Sarah A Soule 1998 ldquoDiffusionin Organizations and Social Movements FromHybrid Corn to Poison Pillsrdquo Annual Review ofSociology 24265ndash90

Tam Tony 1997 ldquoSex Segregation and OccupationalGender Inequality in the United StatesDevaluation or Specialized Trainingrdquo AmericanJournal of Sociology 1021652ndash92

Tomaskovic-Devey Donald and Sheryl Skaggs 2002ldquoSex Segregation Labor Process Organizationand Gender Earnings Inequalityrdquo AmericanJournal of Sociology 108102ndash28

Tomaskovic-Devey Donald and Kevin Stainback2007 ldquoDiscrimination and Desegregation EqualOpportunity Progress in US Private SectorWorkplaces Since the Civil Rights Actrdquo Annals ofthe American Academy of Political and SocialScience 60949ndash84

Tsui Anne S and Charles A OrsquoReilly III 1989ldquoBeyond Simple Demographic Effects TheImportance of Relational Demography in Superior-Subordinate Dyadsrdquo The Academy of ManagementJournal 32402ndash23

US Census Bureau 2003 Census 2000 Public UseMicrosample Technical DocumentationWashington DC US Census Bureau

van Engen Marloes L and Tineke M Willemsen2004 ldquoSex and Leadership Styles A Meta-Analysis of Research Published in the 1990srdquoPsychological Reports 943ndash18

Wharton Amy S 1986 ldquoIndustry Structure andGender Segregation in Blue-Collar OccupationsrdquoSocial Forces 641025ndash31

Wright Erik O 1997 Class Counts ComparativeStudies in Class Analysis Cambridge UKCambridge University Press

Wright Erik O and Janeen Baxter 2000 ldquoThe GlassCeiling Hypothesis A Comparative Study of theUnited States Sweden and Australiardquo Genderand Society 14275ndash94

Young Iris M 1994 ldquoGender as Seriality ThinkingAbout Women as a Social Collectiverdquo Signs19713ndash38



er questions about the status of subordinategroups

We adopt a unique approach to this ques-tion Rather than analyzing managerial workersalone (eg Huffman and Velasco 1997) or sim-ply including them with nonmanagerial work-ers (eg Cohen and Huffman 2003a Jacobs1992) we analyze the wages of nonmanageri-al workers as a function of the gender compo-sition of their managersmdashthe managers in theirlocal industries Further we extend the existingliterature by considering whether the relativestatus of female managers affects the pattern ofgender inequality for the workers beneath themIf female managers influence gender inequali-ty the effect may depend on how highly placedthose female managers are (Denmark 1993)Although this point may seem prosaic carefulattention to the relative status of female man-agers allows us to make more nuanced obser-vations about the conditions under which labormarket benefits extend to ascriptively similarsubordinates Using data from the 2000 USCensus we estimate multilevel wage modelswith controls at three levelsmdashthe individualthe job and the local industrymdashto assess theimpact of female managers on gender inequal-ity These advances provide new leverage ontheoretically important yet unanswered ques-tions concerning the role of managers in gen-der stratification with implications for theinequality trajectories of subordinate groupsmore generally

MANAGERS AS AGENTS OFCHANGE OR COGS IN THEMACHINE

AGENTS OF CHANGE

Many researchers believe that gender inequal-ity at work results in part from the practices ofmanagersmdashoften assuming that these practicesare associated with managersrsquogender For exam-ple Cotter and colleagues (1997715) offer thisas one reason why women benefit from occu-pational integration in the local labor marketldquoAs more women in [positions of authority]make crucial decisions about salaries promo-tions hiring and firing gender differences inearnings should declinerdquo (emphasis added)Similarly Nelson and Bridges (1999) argue thatthe scarcity of women in authority positionssustains workplace gender inequality Managers

by definition have a certain level of organiza-tional authority and thus might be poised tohelp reduce inequality at lower organizationallevels For female managers to reduce workplaceinequality however two assumptions must holdFirst these women must be motivated to act inthe interests of subordinate women Secondthey must have the power to influence outcomesfor subordinates to affect gender inequality

Gender creates a potential common interestbetween female managers and subordinatesbased on homophily (Ibarra 1992 McPhersonSmith-Lovin and Cook 2001) or in Kanterrsquos(1977) term for the tendency of women to hireother women ldquohomosocial reproductionrdquo (seealso Elliott and Smith 2004 Pfeffer 1983) Sexsimilarity between subordinates and supervi-sors increases performance ratings by supervi-sors (Roth 2004 Tsui and OrsquoReilly 1989) andwomenrsquos evaluation of potential female job can-didates is less subject to pregnancy-related bias(Halpert Wilson and Hickman 1993)

More broadly women express stronger sup-port than men do for employer practices aimedat overcoming gender inequality The 1996General Social Survey asked for the level ofagreement with the statement ldquoBecause of pastdiscrimination employers should make specialefforts to hire and promote qualified womenrdquoEmployed women were 119 times more likelythan men to agree (595 versus 498 percentp lt 001 N = 1373) Among managers thedifference was larger with women 132 timesmore likely than men to agree (535 versus 404percent p = 068 N = 193) Not only are womenmore supportive of efforts toward workplaceequality in principle but manager bias againstsuch efforts among women is also less sug-gesting that the presence of female managersshould (if they have the power) promote genderequality2


2 The difference in agreement between nonman-agers and managers was not significant for women(604 percent versus 535 percent agreeing p = 197N = 753) but substantial for men (515 percent ver-sus 404 percent p = 047 N = 620) Baunach (2002)analyzed the same data set and reports no significantgender difference but she used a subsample ofrespondents with insufficient power to identify thedifference (N = 313)













COGS IN THE MACHINE












LOCAL INDUSTRIES













HYPOTHESES











DATA








MEASURES









































Men Women Min Max


















STATISTICAL MODELS














































RESULTS






























ENDOGENEITY




























REFERENCES







APPENDIX

VARIANCE COMPONENTS

























































































































COGS IN THE MACHINE












LOCAL INDUSTRIES













HYPOTHESES











DATA








MEASURES









































Men Women Min Max


















STATISTICAL MODELS














































RESULTS






























ENDOGENEITY




























REFERENCES







APPENDIX

VARIANCE COMPONENTS
















































































































LOCAL INDUSTRIES













HYPOTHESES











DATA








MEASURES









































Men Women Min Max


















STATISTICAL MODELS














































RESULTS






























ENDOGENEITY




























REFERENCES







APPENDIX

VARIANCE COMPONENTS

















































































































HYPOTHESES











DATA








MEASURES









































Men Women Min Max


















STATISTICAL MODELS














































RESULTS






























ENDOGENEITY




























REFERENCES







APPENDIX

VARIANCE COMPONENTS

















































































































DATA








MEASURES









































Men Women Min Max


















STATISTICAL MODELS














































RESULTS






























ENDOGENEITY




























REFERENCES







APPENDIX

VARIANCE COMPONENTS














































































































MEASURES









































Men Women Min Max


















STATISTICAL MODELS














































RESULTS






























ENDOGENEITY




























REFERENCES







APPENDIX

VARIANCE COMPONENTS











































































































































Men Women Min Max


















STATISTICAL MODELS














































RESULTS






























ENDOGENEITY




























REFERENCES







APPENDIX

VARIANCE COMPONENTS



























































































































STATISTICAL MODELS














































RESULTS






























ENDOGENEITY




























REFERENCES







APPENDIX

VARIANCE COMPONENTS



















































































































































RESULTS






























ENDOGENEITY




























REFERENCES







APPENDIX

VARIANCE COMPONENTS


























































































































ENDOGENEITY




























REFERENCES







APPENDIX

VARIANCE COMPONENTS




































































































































REFERENCES







APPENDIX

VARIANCE COMPONENTS














































































































working for the woman? female managers and the gender wage …pnc/asr07.pdf · narrow the gender...

Documents