Working for the Woman? Female Managers and the Gender Wage Gap Philip N. Cohen University of North Carolina, Chapel Hill Matt L. Huffman 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 “glass ceiling.” We offer the first largescale analysis that turns this question around, asking whether the gender characteristics of managers—specifically, the gender composition and relative status of female managers—affect 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. T overall work experience (England et al. 1994; he past two decades have seen substantial Reskin and Ross 1992). In recent years, a spate increases in the proportion of women in of empirical management. During this time, women’s rep-by Ingenta to : research has addressed women’s Delivered resentation in managerial occupations increased North Carolina University of access to managerial authority (Blum, Fields, and Goodman 1994; Huffman and Cohen These posifrom about one-third to one-half.1 Mon, 12 Nov 2007 18:16:31 2004a; Reskin and McBrier 2000; Smith 2002) tions confer well-documented benefits, includand the “glass ceiling”—an unseen barrier ing improved status, wages, autonomy, and between women and management or highstatus positions (Cotter et al. 2001; Hultin 2003; Wright and Baxter 2000). The authors contributed equally to this manuscript Although the question of access to manageand list their names alphabetically. Direct correrial positions is critical to understanding perspondence to Philip N. Cohen, Department of sistent gender inequality in the labor market, the Sociology, University of North Carolina-Chapel Hill, increase in women’s managerial presence rais155 Hamilton Hall, CB#3210, Chapel Hill, NC es a broader question that is provocative and 27599-3210 (pnc@unc.edu). This research was supinherently sociological: What happens to the ported by a grant from the National Science status of a subordinate group when some of its Foundation (SES-0647265). The authors also thank members attain positions from which they might the ASR editors and anonymous reviewers for their helpful comments and suggestions. In addition, we reduce inequality? We use gender to gain insight benefited from comments by Kenneth Andrews, into this question. Specifically, we ask whether Barbara Entwisle, Roberto Fernandez, Arne the increase in women’s representation in manKalleberg, Joan S. M. Meyers, Calvin Morrill, Ted agement “lifts all boats” by reducing gender Mouw, Judy Ruttenberg, and the participants of the inequality among nonmanagerial workers or Gender, Work, and Family research group at the whether the benefits that accrue to female manUniversity of California, Irvine. Any remaining errors agers are limited only to those women. Clearly, or omissions are solely our own. the actions of managers affect those below them 1 This result is from our unpublished analysis of (Wright 1997). Yet, managers’ role in reprodata from the Current Population Survey (available ducing gender inequality is conspicuously upon request). Interestingly, some of this change understudied, despite its relevance for persistent occurred at the same time that progress for U.S. labor market inequality (Hultin and Szulkin women stalled on many other fronts (Cotter, 1999; Hultin and Szulkin 2003) and for broadHermsen, and Vanneman 2004). AMERICAN SOCIOLOGICAL REVIEW, 2007, VOL. 72 (October:681–704) 682—–AMERICAN SOCIOLOGICAL REVIEW er questions about the status of subordinate by definition, have a certain level of organizagroups. tional authority and thus might be poised to We adopt a unique approach to this queshelp reduce inequality at lower organizational tion. Rather than analyzing managerial workers levels. For female managers to reduce workplace alone (e.g., Huffman and Velasco 1997), or siminequality, however, two assumptions must hold. ply including them with nonmanagerial workFirst, these women must be motivated to act in ers (e.g., Cohen and Huffman 2003a; Jacobs the interests of subordinate women. Second, 1992), we analyze the wages of nonmanagerithey must have the power to influence outcomes al workers as a function of the gender compofor subordinates to affect gender inequality. sition of their managers—the managers in their Gender creates a potential common interest local industries. Further, we extend the existing between female managers and subordinates literature by considering whether the relative based on homophily (Ibarra 1992; McPherson, status of female managers affects the pattern of Smith-Lovin, and Cook 2001) or, in Kanter’s gender inequality for the workers beneath them. (1977) term for the tendency of women to hire If female managers influence gender inequaliother women, “homosocial reproduction” (see ty, the effect may depend on how highly placed also Elliott and Smith 2004; Pfeffer 1983). Sex those female managers are (Denmark 1993). similarity between subordinates and superviAlthough this point may seem prosaic, careful sors increases performance ratings by superviattention to the relative status of female mansors (Roth 2004; Tsui and O’Reilly 1989), and agers allows us to make more nuanced obserwomen’s evaluation of potential female job canvations about the conditions under which labor didates is less subject to pregnancy-related bias market benefits extend to ascriptively similar (Halpert, Wilson, and Hickman 1993). subordinates. Using data from the 2000 U.S. More broadly, women express stronger supCensus, we estimate multilevel wage models port than men do for employer practices aimed with controls at three levels—the individual, the job, and the local industry—to assess the at overcoming gender inequality. The 1996 Delivered by Ingenta to : impact of female managers on gender inequal- of North Carolina Survey asked for the level of General Social University ity. These advances provide new leverage12 Nov 2007 18:16:31 the statement: “Because of past agreement with Mon, on theoretically important yet unanswered quesdiscrimination, employers should make special tions concerning the role of managers in genefforts to hire and promote qualified women.” der stratification, with implications for the Employed women were 1.19 times more likely inequality trajectories of subordinate groups than men to agree (59.5 versus 49.8 percent, more generally. p < .001, N = 1,373). Among managers, the difference was larger, with women 1.32 times MANAGERS AS AGENTS OF more likely than men to agree (53.5 versus 40.4 CHANGE OR COGS IN THE percent, p = .068, N = 193). Not only are women MACHINE? more supportive of efforts toward workplace equality in principle, but manager bias against AGENTS OF CHANGE such efforts among women is also less, sugMany researchers believe that gender inequalgesting that the presence of female managers ity at work results in part from the practices of should (if they have the power) promote gender managers—often assuming that these practices equality.2 are associated with managers’ gender. For example, Cotter and colleagues (1997:715) offer this as one reason why women benefit from occu2 The difference in agreement between nonmanpational integration in the local labor market: agers and managers was not significant for women “As more women in [positions of authority] (60.4 percent versus 53.5 percent agreeing, p = .197, make crucial decisions about salaries, promoN = 753) but substantial for men (51.5 percent vertions, hiring, and firing, gender differences in sus 40.4 percent, p = .047, N = 620). Baunach (2002) earnings should decline” (emphasis added). analyzed the same data set and reports no significant Similarly, Nelson and Bridges (1999) argue that gender difference, but she used a subsample of the scarcity of women in authority positions respondents, with insufficient power to identify the sustains workplace gender inequality. Managers, difference (N = 313). WORKING FOR THE WOMAN—–683 Ely (1995) argues that the presence of women gender inequality, but most rely on narrow samin the upper echelons of organizations reduces ples or settings. An exception is a study of three the persistence of sex as a salient category for U.S. cities, which finds that promotions from all workers, thereby weakening some of the supervisor to manager occur more frequently negative consequences associated with gender under conditions of ascriptive similarity imbalance (e.g., performance pressures, stereo(race/ethnicity and gender) with immediate typed role encapsulation, and exclusion from supervisors (Elliott and Smith 2004). Narrower work-related networks). She demonstrates that studies have shown, for example, that California the demographic composition of those holding state agencies with more female managers powerful positions in organizations can have exhibited less gender segregation in the 1970s substantial effects on all workers, not just those and 1980s (Baron, Mittman, and Newman holding positions at or near the top of organi1991), and savings and loans with women in zational hierarchies. management are more likely to hire women into Supporting this view, some empirical studies managerial roles (Cohen, Broschak, and have shown less inequality where women occuHaveman 1998). Additionally, Carrington and py positions of authority. Hultin and Szulkin Troske (1995) find a strong link between the (2003) offer the most direct test of the wage gender of business owners and the gender comeffect of female managers, using rare employposition of their employees. er–employee linked data from Swedish privateA series of studies investigating higher edusector work establishments. They find a negative cation settings shows that female administrators relationship between the gender wage gap (Kulis 1997) or a female president (Pfeffer, among nonmanagerial workers and the proporDavis-Blake, and Julius 1995) are associated tion of women in managerial positions. This with less gender segregation (see also Konrad relationship, which remains substantial in the and Pfeffer 1991). In the legal profession, law presence of controls for individual attributes, f irms whose corporate clients have many Delivered establishment characteristics, and industry, isby Ingenta to :leadership positions show a greater women in University of present for both white- and blue-collar employ- North Carolina female partners (Beckman and Mon, 12 Nov increase in ees. Importantly, Hultin and Szulkin (2003) dis- 2007 18:16:31 and female decision makers tend Phillips 2005), tinguish between higher-level decision makers to fill more vacancies with women (Gorman (managers) and lower-level decision makers 2005). Finally, prime-time television shows with (supervisors). They find that the effect of the sex female producers, executive producers, and composition of supervisors on wage inequaliwriters have a higher percentage of female major ty is stronger than that of managers.3 Although characters (Glascock 2001; Lauzen and Dozier their analysis shows that the level of authority 1999). is an important consideration, female managers These studies imply that there is less gender at high levels in the hierarchy are not shown to inequality under conditions of greater female have a stronger effect than those at lower levrepresentation (and higher status) in manageels.4 ment.5 This may result from several distinct A number of other studies are also consistent mechanisms, including increased access to orgawith the contention that female managers reduce nizational resources and power, homophily preferences, support for equity efforts, and weaker sex-based biases against female workers. Due to data limitations, this portion of their analysis was only performed on the blue-collar subsample. 4 Although we do not link workers to managers directly, as do Hultin and Szulkin (2003), our study is unique in that it includes controls at three levels— the individual, the job, and the local industry. As such, it accounts for variation in gender inequality across larger social contexts, which purely organization-based analyses do not (Cohen and Huffman 2003b). 3 5 Our emphasis on inequality effects differentiates this study from research on gender differences in leadership styles, but the pursuit of that question in the management and psychology literature provides some evidence of a less authoritarian orientation among female leaders, which may benefit female subordinates (for reviews, see Dobbins and Platz 1986; van Engen and Willemsen 2004). 684—–AMERICAN SOCIOLOGICAL REVIEW “authority .|.|. adheres in the office and not in the particular person who performs the official Although there are reasons to believe female role.” In fact, as Charles and Grusky (2004) managers might reduce inequality, the undershow, the increase in managerial integration lying motivation and power assumptions are after the 1970s occurred during a period of debatable. The motivations of female managers growing bureaucratization, which implies limmay be affected by two potential sources of its on the discretionary power of lower-level loyalty or identity: their female peers in subormanagers. Kanter (1977) argues that female dinate class positions and their managerial peers and superiors. Ely (1995) argues that to assume managers in particular occupy weak structural female managers are sympathetic to the women positions. In Ridgeway’s words, they are “handbelow them essentializes gender, while in pracicapped by their lower power and by interactice gender is situationally enacted.6 Class is one tional gender mechanisms” (1997:227). source of distinction that might prevent the Affirmative action programs have been more expression of collective identity among women successful at integrating lower and middle lev(Young 1994). In fact, a selection process may els of management (Brenner, Tomkiewicz, and operate such that female workers are promoted Schein 1989), and to some extent women’s into management partly for their affinity with increasing managerial presence reflects “title the existing hierarchy. The disproportionate proinflation” (see Jacobs 1992)—the reclassificamotion of women who are “team players” may tion of previously nonmanagerial workers as limit the potential for female managers to act managers with little increase in pay or authoragainst inequality. ity. Further, some women share men’s biased Finally, it is possible that due to a baseline views of women’s work (Deaux 1985). For example, women and men in college similarly sectoral segregation, women are typically mandevalue the merit of female job applicants whose agers in workplaces with lower quality jobs. Delivered Ingenta to : resumes reflect motherhood status (Correll and byBoth Shenhav and Haberfeld (1992) and Pfeffer University Benard 2005). With regard to networking, even of North Carolina (1987) find lower earnings and Davis-Blake if female or minority managers areMon, 12 Nov 2007 18:16:31and women in workplaces with likely to for both men pass on job leads or other job-related informamore female managers or administrators. If tion to subordinates, research suggests such female managers are concentrated in organizacontacts may not be systematically beneficial tions with more female workers, then any pos(Huffman and Torres 2002; Mouw 2003). itive effect of female manager attitudes or The managerial power assumption is also behavior may be swamped by negative gender potentially flawed. It is not obvious that manconcentration effects. agers, especially those in bureaucratic organiIn summary, female managers may enhance zations, are able to act autonomously on the the labor market prospects of the women who basis of their own or women’s interests. Instead work below them. Their homophilous preferthey may be compelled to act under the mandates of routinization, efficiency, or profitabilences or affiliations might promote equality, ity—or according to the prejudices of those and they may have less to gain from discrimihigher up the hierarchy. This counterargument nation and therefore be more motivated to help was summarized by Merton (1940:560) as, other women. Additionally, women may be more aware than men of discriminatory practices and less susceptible to cognitive processes leading 6 Research in psychology shows that to the extent to gender bias. Any of these processes may there are gender differences in moral reasoning, as smooth the social and organizational path of advanced by Gilligan (1982), they are contextfemale subordinates. In contrast, bureaucracy, dependent (Ryan, David, and Reynolds 2004) and market pressures, divided loyalties, past disconditioned on, among other factors, socioeconomcrimination, or the mandates of those more ic status. For our purposes, however, we note that powerful may render the ascriptive characterisJaffee and Hyde (2000) find greater gender differtics of managers largely moot with regard to ences in moral reasoning at higher levels of social inequality. class. COGS IN THE MACHINE WORKING FOR THE WOMAN—–685 and Stock 2006). Such competition is one way that corporate practices—which presumably Regardless of how female managers might include those related to gender inequality—are reduce inequality, their impact may depend critadapted by different actors within an organizaically on their status (Denmark 1993). Even tional field. Additionally, employment practices highly motivated female managers working can be adopted by organizations in an effort to together may not be able to influence gender increase legitimacy or to appear in compliance inequality if they are relatively powerless. Both with a changing legal environment (DiMaggio the identity/loyalty issue and the question of and Powell 1983; Tomaskovic-Devey and managerial power highlight the possibility of Stainback 2007). substantial interactions between the represenDiMaggio and Powell’s (1983) theory has tation of women and their relative status. prompted an extensive literature on how to operAlthough disparate studies have investigated ationalize the reference groups or fields for the effect of female representation among manorganizational behavior (Greve 2005; Massini, agers at different levels (e.g., university presiLewin, and Greve 2005; Strang and Soule dents, screenwriters, law firm partners), Hultin 1998). Employing organizations are part of and Szulkin’s (2003) is the only one to directly labor markets that are local (represented by test for the effect across different levels within metropolitan areas), industries that are nationone setting, and their ability to address the issue al (represented by categorization schemes with is limited by their data, which includes only various degrees of detail), and the intersection undifferentiated manager and supervisor cateof the two: local industries. We believe the local gories. In contrast, we use a continuous measindustry—the aggregate of organizations that ure of vertical segregation (explained below) to produce a common product within a common tap the relative status of female managers. local labor market—is an appropriate starting point for our questions. This approach draws LOCAL INDUSTRIES from to : Delivered by Ingentawork on local industrial dominance by University of South and Xu Most of the research in this area is concerned North Carolina (1990), who argue that local Mon, 12 Nov industries are with the direct effects of managers within organ- 2007 18:16:31 sites of intersection between structural forces and individual attainment processizations. Studies have thus been designed to es. These units combine functional commonality draw from managers and workers who are as with social proximity, capturing the interaction closely linked as possible, exemplified by the of these fields. work of Hultin and Szulkin (2003) and Elliott We thus expect local industries to display and Smith (2004). Women in positions of less internal variation in gender-related practices authority, however, may change gender dynamthan either metropolitan labor markets or nationics at various levels of proximity: among immeal industries as a whole. This implies that a diate subordinates, within their organizations in given restaurant, for example, is likely to resemgeneral, across organizations, and across largble other restaurants in its local area more than er social contexts. We know that gender inequalit resembles either restaurants in the entire counity varies systematically across larger social try or all employers in the local labor market. contexts, including metropolitan labor markets Although we cannot fully demonstrate this pat(Cohen and Huffman 2003a; Cotter et al. 1997) tern, we provide a simple illustration based on and national industries (Fields and Wolff 1995; one key variable for establishments: the gender Wharton 1986). This variation is not captured composition of managers.7 when analysis is limited to direct examination Using data from all large U.S. private-sector of organizations (Cohen and Huffman 2003b). firms collected by the U.S. Equal Employment One way the processes reproducing inequalOpportunity Commission (EEOC) in 2002, we ity across contextual levels would be linked is if female managers in one workplace affect proximate organizations. For example, an 7 This is the measure used by Ashenfelter and employer’s decision to hire women creates a competitive advantage relative to those who do Hannan (1986), who examine the pattern of gender not (Ashenfelter and Hannan 1986), especially representation in management for banks across local when women are paid less than men (Neumark markets. MANAGERS’ RELATIVE STATUS 686—–AMERICAN SOCIOLOGICAL REVIEW measured the similarity of establishments on likely to approximate that of her own estabmanagerial gender composition across 275 metlishment, at least compared to the corresponropolitan labor markets, 301 national indusdence obtained at the level of the local labor tries, and 10,131 local-industry cells formed market or national industry. We thus consider by the intersection of labor markets and induslocal-industry managerial composition as a tries. First, we calculated the natural logarithm proxy for establishment management characof the percentage female among “officers and teristics. managers” at each establishment. We then calculated the standard deviation in the natural HYPOTHESES logarithm for each contextual unit—each metPrevious research has shown that the overall ropolitan area, national industry, and local induswage gap largely results from between-job and try. Finally, we averaged the standard deviations between-occupation inequality, whereby femaleacross these contextual units. The means of dominated jobs and occupations pay less than these standard deviations are 1.45 for metrootherwise comparable male-dominated lines of politan areas, 1.22 for national industries, and 1.03 for local industries.8 In terms of the genwork (Cohen and Huffman 2003a; England et der composition of managers, establishments on al. 1994; Huffman and Velasco 1997; Tam average are indeed more similar to others with1997), in part because female-dominated jobs in their local industries than they are to those offer fewer training opportunities (Tomaskovicwithin their entire local labor markets or nationDevey and Skaggs 2002). However, inequality al industries as a whole. This supports the prewithin job and occupational categories also consumption that gender-related organizational tributes to wage inequality. Clearly, manageridynamics generally cluster or reproduce more al composition and relative status could shape tightly within local industries than within these wage inequality through either route, as manlarger units. agers might influence both sorting and the trainOf course, we remain interested inDelivered bying and rewards processes. Rather than make withinIngenta to : organization effects of managerial gender com- of North Carolina predictions without justification from prior theUniversity position on gender inequality. Despite reasons Mon, 12 Nov 2007or research, our models account for both ory 18:16:31 to suspect larger processes, this more direct possibilities. We also do not consider more comeffect remains the most plausible pathway by plex interactions (e.g., McCall 2001), such as which female managers influence gendered outthose involving the conditional effects of mancomes, as has been shown in the limited research agers’ race and gender. that uses such linked data. In the absence of such Our analysis concerns the extent to which data on a generalizable scale, however, we take wages for men and women are sensitive to both heart from the results of our EEOC exercise, the representation of women among managers which imply that for a given worker, the gender and the relative status of those female mancomposition of the local industry’s managers is agers. We test two hypotheses, beginning with whether the gender wage gap is smaller in local industries where there are more female man8 Our data are from the 2003 EEOC files, based agers. Specifically, we test: on required filings by all private-sector establishHypothesis 1: There is less gender wage ments with 50 or more employees and smaller firms inequality in local industries with a highif they are federal contractors. The calculations are er proportion of female managers. based on about 170,000 individual establishments with any managerial workers located in identifiable metropolitan areas (as used below). We set logged percent female to 0 where there were less than 1 percent female managers (the results were substantively the same when we used unlogged percentages). The differences in mean standard deviations were highly significant at conventional levels. Details are available from the authors. See Cohen and Huffman (2007) and Robinson and colleagues (2005) for more recent analyses of this data set. This hypothesis addresses the association between the representation of women in managerial positions and wage inequality. If female managers tend to cluster at the bottom of managerial hierarchies, though, their mere representation in management may be insufficient to alter wage inequality. Therefore, we test the interaction between the representation and the relative status of female managers: WORKING FOR THE WOMAN—–687 Hypothesis 2: There is less gender wage inequality in local industries (a) that have more female managers and (b) in which female managers on average hold higher-status positions. To test these hypotheses, one must combine data from several sources and organize them into a multilevel structure. Despite large data sets and sophisticated methods, as is typical in largescale studies of labor market inequality, we cannot offer causal tests of these hypotheses. Rather, we test whether the data are consistent with the patterns predicted by these hypotheses. Following Reskin (2003:14), we believe that “although contextual effects are not themselves mechanisms, they are proxies for mechanisms that vary across settings.” Next, we describe our data collection and manipulation as well as our statistical modeling strategy. industries (representing 155 industries in 79 metropolitan labor markets).9 Additional measures for characteristics of local industries, based on their metropolitan areas, are drawn from the Census 2000 Summary Files. The Dictionary of Occupational Titles (DOT), which includes measures of occupational skills and requirements, is the final source of data (TomaskovicDevey and Skaggs 2002). Appending these measures to occupations in the 2000 Census required converting occupation codes from the 1990 scheme to the new coding scheme employed in 2000.10 9 We exclude workers who were self-employed, in military-specific occupations, or in the armed forces because their wages and promotions are not determined by local managers. We also exclude those with wages outside the range of $1 to $300 per hour; legislators, for whom we have no occupational charDATA, MEASURES, AND MODELS acteristics; and those with no specific metropolitan DATA area identified (mostly rural workers). The age restriction is applied after job cell characteristics are calWe investigate our hypotheses by analyzing culated (so that all workers contribute to job data at three conceptual levels: individual work-by Ingenta to : Delivered characteristics, not only those for whom the outers, jobs, and local industries. We nest individ- North Carolina University of come is analyzed). The job and local industry restricual workers in “jobs,” defined asMon, 12 Nov 2007 18:16:31 the final sample from 3.45 to 1.32 three-digit tions reduce occupation by three-digit industry by metromillion workers, mostly by removing workers from politan area cells (Cohen and Huffman 2003a; smaller labor markets. Notable differences in the final sample include more foreign-born workers, Huffman and Cohen 2004b). Our innovation is more concentration in the Northeast and West regions, that employed respondents are separated into and more highly educated workers. Workers excludmanagerial and nonmanagerial jobs. ed by job and local industry restrictions have logged Nonmanagerial workers are the subjects of our wages .18 lower than those in the final sample, but wage analysis. Each nonmanagerial job is nestthis is reduced to less than .05 when adjusted for ed within a local industry—a three-digit indusobserved individual characteristics. We have no reatry in a metropolitan labor market. Manager son to suspect that our selection criteria introduce syscharacteristics are drawn from the managerial tematic biases with regard to our hypotheses, but we workers in each local industry and used as indecannot rule out that possibility. 10 The DOT database contains information for pendent variables. Our primary data source is the combined nearly 13,000 occupations corresponding to about 500 occupations in the three-digit codes used by the 2000 Census 5- and 1-percent Public Use 1990 Census. We matched 2000 Census occupations Microdata Samples (PUMS). We analyze the to DOT occupations using a crosswalk file from the wages of metropolitan nonmanagerial civilian National Occupational Information Coordinating workers ages 25 to 54 years. We restrict our Committee and calculated mean scores across DOT sample to those who are employed in jobs with occupations for each Census occupation. We used at least 10 people and local industries with at NOICC Master Crosswalk v. 4.3 (revised November least 10 managers in each of at least two man15, 1999), and a f ile titled “DOTCEN00” agerial occupations—this is necessary for cal(“Crosswalk linking the 2000 Census occupations to culating female managers’ relative status (see those from the Dictionary of Occupational Titles”), below). This process yields a sample of approxboth accessed from the National Crosswalk Service imately 1.32 million workers nested in 29,294 Center (http://www.xwalkcenter.org) on March 10, local jobs, which are in turn nested in 1,318 local 2006. 688—–AMERICAN SOCIOLOGICAL REVIEW acteristics of jobs, including percent Black, percent Latino, and percent Asian. We also control LEVEL 1: INDIVIDUALS. In our models, the for the percent part-time employed in the job. dependent variable is the natural logarithm of the From the DOT, we include three measures: stanhourly wage, which is annual earnings divided dard vocational preparation (SVP), general by hours worked. At the individual level, we educational development (GED), and physical include a binary variable for gender (female = strength (STR). SVP measures the amount of 1) and four dummy variables to represent training time needed to learn the techniques respondents’ ethnicity (coded 1 if the respondent and obtain the information necessary for averis Black, Latino, Asian, or other ethnicity).11 age job performance (high values of this scale White is the omitted race/ethnicity. We use represent a longer period of time required to dummy variables to capture differences in eduacquire the skills). SVP can be thought of as a cational attainment (less than high school, some measure of occupation-specific human capital college, bachelor’s degree, or master’s degree or (Tomaskovic-Devey and Skaggs 2002), while above), whether the respondent is married, forGED measures “the typical requirement of the merly married (divorced, widowed, or separatoccupation for schooling that is not vocationally ed), and foreign born. We also include dummy specif ic” (England, Hermsen, and Cotter variables to control for disability status and the 2000:1742). GED is calculated as the mean of presence of children younger than six years in values required for mathematics, language, and the household, as well as whether the responreasoning preparation. Finally, STR is coded dents do not speak English well and whether to reflect strength requirements for each occuthey are currently attending school. Continuous pation, ranging from 1 (sedentary) to 5 (very variables measure potential labor market expeheavy). This variable is intended to capture the rience (age minus years of education minus 5) manual nature of occupations, which features and its square, in addition to number of own chilprominently in the gender division of labor dren in the household. (Charles and Grusky 2004). Delivered by Ingenta to : University of North Carolina LEVEL 2: JOBS. At the job level, Mon, 12 Nov 2007 18:16:31 our indeLEVEL 3: LOCAL INDUSTRIES. At the local pendent variable of primary interest is percent industry level, our key independent variables are female. To account for nonlinear effects of perpercent female and the relative status of female cent female, we include percent female managers. In the absence of a direct measure of squared.12 We measure other demographic chardecision-making authority, we identify managers as those in “management occupations” in the occupational classifications of the federal government (U.S. Census Bureau 2003). This 11 Because of overlapping racial and ethnic idenclassification clearly is a relevant, if imperfect, tification in the 2000 Census, we use a descending measure of authority.13 Percent female among selection to reach mutually exclusive categories, in MEASURES the following order: Latino, Black, Asian, other, White. Latinos are thus coded as such regardless of their responses to the race question, and Whites are those who selected no Latino ethnicity or other race. This conforms to the recommendation of the federal Office of Management and Budget with regard to civil rights enforcement (Goldstein and Morning 2002). 12 Consistent with many findings in the literature (e.g., Cohen and Huffman 2003a; Cotter et al. 1998; England et al. 1994), we found a linear effect on wages of the gender composition of jobs in models with controls at all levels. Cotter and colleagues (2004), however, present evidence from the 2000 Census that calls this simple relationship into question: cubic and 4th-power fits for women and men, respectively, in the bivariate relationship between occupation percent female and median earnings at the national level. Closer examination revealed a 4thpower fit in the bivariate relationship between job gender composition and average wages. With the introduction of controls at the individual level (which also captures women’s lower average individual wages), the best fit was cubic, and with controls at the job level added, the best fit fell to quadratic. This did not change with the addition of local-industry controls. Therefore, we model the job gender composition effect as quadratic. 13 We checked the Multi-City Study of Urban Inequality for three large U.S. cities and found that 65 percent of workers in managerial occupations report having the authority to hire and fire others WORKING FOR THE WOMAN—–689 rank occupations according to the level of authority—chief executives and food service managers, for example, are both simply coded as possessing authority. In the absence of a direct measure of authority, we first select occupations in the federal system’s “management occupations” category, then apply a measure based on both skills and training (representing expertise) and earnings (representing recognition and rewards). Our measure of the status of each managerial occupation is the average of two factors: (1) the average of zscores for years of education, SVP, and GED and (2) the z-score for earnings. Managerial occupations are thus ranked according to equal weightings of the skills and training required (education, SVP, and GED) and also average ND = 100 ( MiCFi – FiCMi ) earnings.14 In the resulting authority ranking, natural-science managers are highest and gamMi and Fi equal the proportion of males and ing managers are lowest.15 females, respectively, in managerial occupation We control for other important characteristics i. CFi equals the cumulative proportion of of local industries that could affect gender females in managerial occupations ranked below inequality. Among managers, we include manmanagerial occupation i, and CMi equals the agers as a percentage of all workers. This is analogous cumulative proportion of men. When intended to capture the level of rationalization ND equals zero, men and women are equally or bureaucratization of the local industry, which likely to occupy high-status occupations. NDby Ingenta to : Delivered equals 1 when all women are in higher-status North Carolina University of are traits associated with reduced reliance on Mon, 12 –1 ascriptive characteristics in determining work occupations than men, and when ND equals Nov 2007 18:16:31 positions and rewards (Jackson 1998; Reskin all women are in lower-status positions than 2003). men. We also include characteristics of local labor Calculating ND requires an ordinal ranking markets at this level, drawing data from the of managerial occupations. Unfortunately, PUMS and 2000 Census Summary Files.16 To authority over other workers, although theoretically central to analyses of class dynamics at work (Wright 1997), is not a prominent concern in most occupational research. We know of no 14 Some managerial occupations are specific to one source that reports the relative decision-making industry (e.g., funeral directors), but most occur authority or status of managerial occupations. across at least several industries (e.g., human resource Despite detailed descriptions of thousands of managers and chief executives). To calculate status, occupations, neither the DOT nor the more we combined managers from all industries in all labor markets. recent federal O*Net descriptor system meas15 Because women are heavily represented in some ure authority directly. One approach is to higher-status managerial occupations (e.g., medical dichotomously code occupations from the and health service managers and educational adminCensus as having authority or not, based on the istrators) and poorly represented in some lower-staappearance of the words “manager,” “supervitus positions (e.g., construction managers), our status sor,” or “administration” in the occupation title measure and percent female are only weakly corre(England et al. 1994). This does not, however, lated (r = –.16). Across all local industries, female managers in each local industry is simply obtained from the PUMS files, using the same criteria for selection we use for nonmanagerial workers at level 1. To measure female managers’ status relative to male managers—our proxy for decisionmaking power—we use a measure of vertical segregation, the Index of Net Difference (ND), as described by Lieberson (1976). ND is the difference between the likelihood that a randomly chosen man is employed in a higher-status managerial occupation than a randomly chosen woman and the opposing probability, that a randomly selected woman works in a higher-status managerial occupation than a randomly chosen man. Specifically, ND is given by: (compared to 30 percent of those in other occupations) and 48 percent reported influencing the rate of pay of others (compared to 22 percent of those in nonmanagerial occupations). managers’ relative status is higher where their representation in management is greater (r = .09). 16 In principle, labor market variables could constitute a fourth level of the analysis. However, because our hypotheses do not focus on local labor market dynamics per se, we include these variables at level 690—–AMERICAN SOCIOLOGICAL REVIEW capture other aspects of local gender dynamics centage Black and percentage Latino, and the we control for the overall level of occupationrate of in-migration, measured by the percental gender segregation and the demand for female age of local residents who moved to the metrolabor. Both of these measures are computed politan area in the previous five years. across the 33 major occupational categories Descriptive statistics at each level of the reported in the Summary Files. The segregation analysis appear in Table 1, which shows that men measure uses the index of dissimilarity (Duncan in our sample have average logged wages of 2.84 and Duncan 1955). This measure has a signif($17.13), compared to 2.67 ($14.42) for women, icant association with the gender gap in earnfor a gender wage ratio of .84. Managerial comings across U.S. labor markets, such that those position ranges from 2 to 90 percent, with means with higher levels of segregation have lower of 34 percent for men and 48 percent for women. relative wages for women, regardless of whether Female managers’ relative status has a mean of they work in male- or female-dominated occu–.075 and ranges from –.72 to .79. pations or jobs (Cohen and Huffman 2003b; Cotter et al. 1997). The demand measure reflects ILLUSTRATIVE EXAMPLES the number of women who would be employed Several examples clarify our data structure and if local occupations had the gender composition its applicability to our hypotheses. Table 2 shows observed nationally, divided by the actual numa comparison across four local industries: restauber of employed women. Labor markets with higher levels of demand for female labor also rants and computer systems in the Los Angeles have significantly less gender wage inequality and New York metropolitan areas.18 Our sam(Cotter et al. 1998). Our measure is similar to ple has 1,887 managers in Los Angeles restauthat used by Cotter and colleagues (1998), rants. Of these, food service managers are the although with less occupational detail. These most common (N = 1,450). This local industry two controls are especially important if we are also has 44 nonmanagerial occupations (jobs) to avoid spurious effects whereby wage gaps and bywith 10 workers or more and a total of 10,422 Delivered Ingenta to : women’s managerial representation are both of North Carolina University workers, the plurality of whom are cooks (N = Mon, 12 Nov 2007 18:16:31 more favorable as a result of larger factors 2,928). affecting all women in the local labor market.17 Our analytic strategy works on the assumpTo capture other local economic conditions, tion that the behavior of managers in the local we include the unemployment rate and industrial industry—which is influenced by their gender composition, measured by percentage in mancomposition and the relative status of the women ufacturing among all workers. Region is conamong them—may influence the wages of the trolled with dummy variables for the South, nonmanagerial workers under them. In the West, and Midwest (Northeast is omitted). We restaurant example, we see that there are more also include demographic characteristics: the female managers in Los Angeles (42.4 percent) natural logarithm of population size, the perthan in New York (33.4 percent). The female managers in New York, however, have higher relative status than those in Los Angeles, most3, which allows us to use the HLM software (see ly because they are less concentrated in the lowbelow) for computing and greatly simplifies comstatus food service manager occupation. The putation and interpretation. ND score for New York (.021) is thus higher 17 In our sample, female representation in manthan that for Los Angeles (–.047). Among nonagement is negatively correlated with occupational managerial workers, New York has fewer women segregation (r = –.06, p < .05) but not with demand. (40.7 versus 45.4 percent), and the gender wage Female managers’ relative status, on the other hand, ratio among nonmanagerial workers is higher in is higher in local labor markets with more occupaNew York, where women earn 95 percent of tional segregation (r = .06, p < .05) and lower in those with more female demand (r = –.08, p < .01). This implies that in more integrated markets, and those with more female-dominated occupational structures, women are more likely both to be managers and to be crowded into low-level managerial positions. 18 We chose the restaurant industry because it has the most local cases in our data set and the computer system industry as a male-dominated contrast. WORKING FOR THE WOMAN—–691 Table 1. Descriptive Statistics for Variables Used in the Analysis (by gender) Men Women Min. 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1.00 1.00 1.00 2.33 –.72 –51.72 1.20 .34 .92 0 0 0 0 11.92 .49 .63 3.45 3.66 22.74 Max. 5.70 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 12 1 1 1 48 2304 1 1 1 1 1 6 8.22 5 90.00 .79 45.21 37.72 .48 1.04 1 1 1 1 16.87 39.95 47.66 11.98 39.41 35.62 Individuals —Hourly Wage (natural log) 2.84 2.67 —Less than High School .18 .11 —High School Complete .23 .22 —Some College .27 .32 —B.A. .20 .23 —M.A. or Higher .12 .13 —In School .08 .10 —Work Disability .17 .15 —White .58 .60 —Black .11 .15 —Asian .08 .08 .20 .15 —Latino —Other Race/Ethnicity .02 .02 —Never Married .29 .25 —Married .57 .56 —Formerly Married .14 .18 —Own Children in Household .90 .90 —Children Under 6 in Household .23 .19 —English (not spoken well) .17 .11 —Foreign Born .30 .24 —Potential Experience 18.89 19.23 —Potential Experience Squared 432.73 448.85 Jobs Delivered by Ingenta to : .69 —Proportion Female .28 University of.13 North Carolina —Proportion Part-Time .23 Mon, 12 Nov.11 2007 18:16:31.15 —Proportion Black —Proportion Latino .20 .15 .07 .08 —Proportion Asian —General Educational Development 3.26 3.54 —Specific Vocational Preparation 5.66 5.64 —Strength 2.27 1.82 Local Industries —Manager Percent Female 33.85 47.80 —Female Manager ND –.07 –.08 —ND Percent Female Interaction –2.52 –3.43 —Percent Managers 10.19 9.95 .38 .38 —MA Gender Segregation —MA Female Labor Demand/Supply 1.01 1.01 —Northeast .26 .30 —Midwest .15 .14 —South .26 .27 —West .32 .29 —MA Population (ln) 15.81 15.87 —MA Percent Black 13.20 13.59 —MA Percent Latino 16.70 16.32 —MA Unemployment 5.89 5.89 —MA Percent in Manufacturing 12.42 12.28 —MA Percent In-migration 30.02 29.75 Source: 2000 U.S. Census Public Use Microdata files and other sources (see text). Notes: MAs are (Consolidated) Metropolitan Statistical Areas. ND is the index of net difference, showing women’s 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 < .001, except Asian (n.s.) and own children in household (p < .05). 692—–AMERICAN SOCIOLOGICAL REVIEW Table 2. Managerial and Nonmanagerial Occupations in Four Local Industries Los Angeles Percent N Female Wage N New York T-tests Percent Percent Female Wage Female Wage Restaurants and Other Food Services —Largest Managerial Occupations ——Chief Executives 31 22.6 64.30 30 30.0 ——General & Operations Managers 179 30.2 19.30 154 29.9 ——Marketing & Sales Managers 61 44.3 30.97 26 50.0 ——Human Resources Managers 110 48.2 15.33 65 44.6 ——Food Service Managers 1,450 43.7 15.27 1,681 32.9 ——Total (including occupations not shown) 1,887 42.4 17.33 2,014 33.4 ——Gender ND –.047 .021 —Largest Nonmanagerial Occupations ——Cashiers 827 80.9 11.00 701 78.3 ——Waiters & Waitresses 2,529 66.0 12.18 2,824 61.6 ——Food Preparation Workers 444 48.0 10.18 551 42.7 ——Supervisors, Food Prep., & Service Workers 717 49.9 14.41 627 40.8 ——Bartenders 304 35.9 12.86 452 46.7 ——Cooks 2,928 29.5 11.88 2,384 18.8 ——Attendants & Bartender Helpers 401 14.5 10.12 303 27.4 ——Chefs & Head Cooks 482 14.0 14.04 1,215 11.1 ——Dishwashers 255 13.7 9.23 305 11.2 ——Driver/Sales Workers & Truck Drivers 316 13.0 11.98 237 8.0 ——Total (including occupations not shown) 10,422 45.4 12.19 11,025 40.7 Delivered by Ingenta to : 4,487 ———Women 4,730 11.66 University of North Carolina ———Men 5,692 12.65 Mon, 12 Nov 2007 18:16:316,538 ———Gender Wage Gap 92.2 Computer Systems Design and Related Services —Largest Managerial Occupations ——Chief Executives ——General and Operations Managers ——Computer & Information Systems Managers ——Financial Managers ——Marketing & Sales Managers ——Total (including occupations not shown) ——Gender ND —Largest Nonmanagerial Occupations ——Sales Representatives ——Computer Support Specialists ——Computer Programmers ——Computer Software Engineers ——Computer Scientists & Systems Analysts ——Network Systems & Data Comm. Analysts ——Total (including occupations not shown) ———Women ———Men ———Gender Wage Gap 65.06 19.75 21.90 12.41 18.46 19.31 * * * * 11.24 12.44 10.17 14.01 13.62 11.12 9.82 14.19 8.42 12.02 12.24 11.87 12.50 95.0 * * * * * * 89 18.0 45 22.2 115 25.2 24 45.8 84 36.9 589 28.9 –.140 87 92 352 428 296 155 2,104 632 1,472 27.6 31.5 23.9 19.6 20.6 17.4 30.0 55.35 44.13 32.36 32.18 30.69 37.49 108 18.5 57 19.3 216 24.5 44 38.6 132 39.4 893 31.5 –.196 131 128 765 665 617 238 3,586 31.3 25.8 24.2 22.4 20.1 17.2 29.8 1,069 2,517 63.40 48.39 40.28 38.19 40.92 43.06 * * * 37.41 24.12 33.40 34.03 31.10 21.99 28.82 25.33 30.32 83.5 43.24 27.21 34.30 37.92 36.10 33.65 32.94 28.47 34.83 81.7 * * * * * * Source: 2000 U.S. Census Public Use Microdata files. Notes: Full names are Los Angeles-Riverside-Orange County and New York-Northern New Jersey-Long Island. Managerial occupations sorted by status; nonmanagerial occupations sorted by percent female. Wages are means. ND is the index of net difference (see text). * p < .05 (two-tailed). WORKING FOR THE WOMAN—–693 men’s average wages, compared to 92.2 perthe distribution of male and female workers across categories of female manager represencent in Los Angeles. tation, as well as median wages and female The computer systems industry in these two managers’ relative status. The table shows that labor markets is much more male dominated, 31.8 percent of men, but only 6.5 percent of has higher wages than restaurants, and features women, work in local industries with fewer than more striking gender inequality. In this case, 20 percent female managers, typified by the female managers are more common in New construction industry. On the other hand, more York than in Los Angeles, but New York has than one-quarter of women, but only 8 percent more women in lower-status marketing manager of men, work in local industries where 60 perjobs and fewer in higher status positions. Among cent or more of the managers are female, typinonmanagerial workers, the gender wage ratio fied by hospitals. Thus, the gender of workers is lower in New York than in Los Angeles (.82 and managers is clearly related. Note that the versus .84). The similar gender distribution most common situation involves female manaacross jobs suggests this gap mostly reflects gerial representation between 20 and 40 percent inequality within jobs rather than job segrega(e.g., restaurants). This is the group of local tion. industries in which female managers have the Continuing the restaurant example, our data lowest relative status (ND = –.14) and the genset is adequate to analyze 58 local industries, der gap in median wages is greatest, with nonthat is, the restaurant industry in 58 local labor managerial women earning just 75 percent of the markets. Figure 1 shows the relationship male wage. We now investigate these relationbetween manager percent female and the genships on a larger scale, taking into account posder wage ratio among nonmanagerial workers sible confounding factors at the levels of the in these local industries, with the largest metperson, job, and local industry. ropolitan areas highlighted. For illustration, we split the local industries at the median gender STATISTICAL ND score (–.047) and show each half in a sep-by Ingenta to : MODELS Delivered University of arate plot. In the low-ND markets—where North Carolina Mon, 12 Nov Our data have a nested structure, with individfemale managers are in lower-status positions 2007 18:16:31 ual workers nested within jobs, which in turn are relative to male managers—there is no relanested within local industries. We therefore use tionship between percent female and the gender three-level hierarchical linear models wage ratio. In the high-ND markets, however, (Raudenbush and Bryk 2002), which are justilocal industries with more women in managefied both on statistical and substantive grounds. ment exhibit a smaller gender gap. In this indusFirst, hierarchical models avoid downwardly try, the pattern across labor markets suggests that biased estimation of standard errors when data the effect of higher female representation among are nested (Guo and Zhao 2000; Raudenbush managers may be conditional on their attainment and Bryk 2002). Additionally, they provide the of higher-status managerial positions. flexibility to specify cross-level interaction These examples also illustrate an issue we effects, on which our hypotheses are based. For mentioned above—that female managers may example, is the individual gender effect stronger be ghettoized in the same industries where or weaker depending on the gender composition female workers are concentrated, producing a of the job and its local managers? negative relationship between female manageSpecifically, in our individual-level model, rial concentration and average wages for both logged wages are a function of a model interfemale and male workers.19 Table 3 displays cept (average wages for men), a female effect (within-job gender inequality), and controls. Formally, our individual-level model can be 19 In fact, the data show a small positive bivariate expressed as: correlation between female managerial representation and (logged) wages for both men (r = .09) and women (r = .07). A closer inspection, though, shows this is because a cluster of male-dominated blue-collar jobs, such as construction, have relatively low wages and almost no female managers. Yijk = 0jk + 1jk(femaleijk ) mjkamijk + 2jka1ijk + .|.|. + + eijk where Yijk is the logged wage of person i in job j in local industry k, and 0jk is the inter- 694—–AMERICAN SOCIOLOGICAL REVIEW Delivered by Ingenta to : University of North Carolina Mon, 12 Nov 2007 18:16:31 Figure 1. Restaurant Industry Manager Percent Female and the Gender Wage Gap Among Nonmanagerial Workers, by Female Managers’ Relative Status in 58 Labor Markets Notes: Local industries split at median ND (–.047), N = 29 in each figure. Correlations and regression lines weighted by sample size. N’s range from 51 (Santa Barbara, CA) to 2,886 (New York); the largest 10 labor markets are labeled. Wage ratio is women’s median wage as percent of men’s. cept for job j in local industry k. Next, 1jk is the individual-level effect of gender, a mijk denote the M individual-level control variables, and 2jk through mjk are the associat- ed individual-level regression coefficients. Finally, e jk is the level-1 random effect. The coeff icient 1jk is of particular interest because it represents the net within-job wage WORKING FOR THE WOMAN—–695 Table 3. Nonmanagerial Worker Distribution and Wages, by Manager Percent Female in Local Industries Worker Distribution Manager Percent Female ≤ 19 20 to 39 40 to 59 ≥ 60 Total Median ND –.022 –.140 –.074 –.050 –.068 Median Wage Most Common Industry Construction Restaurants K–12 Schools Hospitals Wage Total Men Women Total Men Women Ratio 19.4 25.7 37.7 17.2 100.0 31.8 31.0 29.2 8.0 100.0 6.5 20.3 46.5 26.8 100.0 15.00 16.92 15.63 15.38 15.77 15.38 19.23 17.79 17.00 17.31 12.88 14.42 14.51 14.96 14.42 .84 .75 .82 .88 .83 Source: 2000 U.S. Census Public Use Microdata files. Notes: Worker distribution and wages are measured at the individual level. Manager ND is the median for local industries within each category, weighted by the number of workers in each local industry. Most common industries are those with the most total workers in each category, shown for illustration. ND is the index of net difference (see text). terms mean that the coefficients for the intergap. To simplify interpretation of the results, all variables except gender are centered cepts and job composition are within-localaround their grand means. Thus, 0jk repreindustry effects, just as the intercept and gender sents the average wage for men at the mean coefficient in the individual-level model are of the control variables. within-job effects. Our job-level model estimates both the levelFinally, each level-2 coefficient relating job 1 intercept and the level-1 female effect as a characteristics to level-1 effects on wages can function of job percent female (and its square) be modeled as either a random or a fixed effect and our job-level controls. Each level-1 coeffiacross local industries. In our models, only the Delivered cient relating individual characteristics to wagesby Ingenta to : level-2 intercept and the job proportion female can be modeled as either a fixedUniversity of North Carolina or random Mon,level-1 2007 18:16:31 12 Nov (and its square) coefficients are permitted to effect across jobs. We allow only the vary across local industries. Thus, our level-3 intercept and the coefficient for the female model is: dummy variable to vary across jobs. Thus, our level-2 model is: 00k = 000 + 001(mngr %femalek) + 002(mngr NDk) + 003 0jk = 00k + 01k(job %femalejk) + 2 )+ (mngr %femalek mngr NDk) + 02k(job %female jk 03kX1jk + .|.|. + .|.|. + 00sWsk + u00k 0qkXqjk + r0jk 1jk = 10k + 11k(job %femalejk) + 2 12k(job %female jk) + 13kX1jk + .|.|. + 1qkXqjk + r1jk 01k = 010 + 011(mngr %femalek) + (mngr NDk) + 013(mngr %femalek 012 mngr NDk) + .|.|. + 01sWsk + u01k where 00k is the intercept for the job-level model in local industry k. In turn, 01k and 02k are the effects of job proportion female and its square on 0jk. Likewise, 10k is the job-level intercept for the effect of being female, 1jk, while 11k and 12k are the effects of job proportion female and its square on the level-1 effect of being female (these are cross-level interaction effects). Finally, X1jk through Xqjk denote the Q control variables in each job-level model. These control variables are centered around their grand means. The level-2 error terms are denoted by r0jk and r1jk. These error 02k 022(mngr + 021(mngr %femalek) + NDk) + 023(mngr %femalek mngr NDk) + .|.|. + 02sWsk + u02k 020 = 10k 102(mngr + 101(mngr %femalek) + NDk) + 103(mngr %femalek mngr NDk) + .|.|. + 10sWsk + u10k 100 = = 110 + 111(mngr %femalek) + 112(mngr NDk) + 113(mngr %femalek mngr NDk) + .|.|. + 11sWsk + u11k 11k 696—–AMERICAN SOCIOLOGICAL REVIEW = 120 + 121(mngr %femalek ) + 122(mngr NDk ) + 123(mngr %femalek mngr NDk ) + .|.|. + 12sWsk + u12k 12k are centered around their grand means. The level-3 error terms are given by u for each of the k local industries. RESULTS Results from the hierarchical linear model appear in Table 4. We show the coefficients only for key variables; complete results are available from the authors. The variance components for several models are presented in the Appendix. Recall that in our models gender is associated with individuals’ wages through three distinct pathways: individual gender, the gender of the co-workers in their jobs, and the gender of the where 000, 010, 020, 100, 110, and 120 are the level-3 intercepts in models of the level-2 coefficients; 001, 011, 021, 101, 111, and 121 are the effects of the gender composition of managers on the level-2 coefficients. Likewise, 002, 021, 022, 102, 112, and 122 denote the effects of the status of female managers (ND) on the level-2 coefficients. Finally, 003, 013, 023, 103, 113, and 123 are manager percent female by manager relative status interaction terms. Coefficients for the S level-3 control variables (W) are denoted by the remaining terms. They Table 4. Coefficients for Three-Level Hierarchical Linear Regressions for Logged Wages On Individual, Job, and Local-Industry Characteristics Level 1 Individual Level 2 Variables Job Variables INTERCEPT Manager Percent Female Index ofDelivered by Ingenta to : Net Difference Manager Percent Female Net Difference University of North Carolina Level 3 Local-Industry Variables Coefficient 2.888*** –.002*** –.056 .002 .124* .001 .324 –.018** –.385*** .002 –.268 .017* –.126*** –.0003 .026 –.002 –.270*** .004* –.456 .023** .327*** –.004* .475 –.023** t-statistic 236.85 –6.43 –1.22 1.66 2.27 .52 1.46 –3.01 –6.00 1.09 –.99 2.55 –7.98 –.68 .40 –1.15 –3.72 2.09 –1.47 2.72 4.34 –2.17 1.46 –2.85 Job Percent Female Mon, 12 Nov 2007 18:16:31 Manager Percent Female Index of Net Difference Manager Percent Female Net Difference Job Percent Female2 Manager Percent Female Index of Net Difference Manager Percent Female FEMALE Manager Percent Female Index of Net Difference Manager Percent Female Job Percent Female Manager Percent Female Index of Net Difference Manager Percent female Job Percent Female2 Manager Percent Female Index of Net Difference Manager Percent Female Net Difference Net Difference Net Difference Net Difference Source: 2000 U.S. 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, on the individual female coefficient, are cross-level interactions showing effects on the wage difference between men and women. * p < .05; ** p < .01; *** p < .001 (two-tailed tests). WORKING FOR THE WOMAN—–697 managers in their local industries. At level 1, the percent female and its interaction with ND on the level-2 variables in the lower half of the gender of individuals is associated with wage diftable. The net result is that female managers are ferences within jobs. At level 2, the gender comassociated with a reduced gender wage gap position of jobs is associated with the average especially when those female managers hold wages of men and women across those jobs relatively high-status positions. We will illustrate within local industries. This effect may be difthe results graphically rather than walk through ferent for women and men and may be nonlina summary of many interactions and nonlinear ear. At level 3, the gender characteristics of effects. managers are associated with overall wage levFigure 2 shows predicted wages (the y-axis) els, as well as with the job- and individual-level for male and female workers in jobs of typical effects. The variables at each level are labeled in gender composition (30 percent female for men, three columns in Table 4. Because the number 70 percent female for women), working under of interactions complicates the interpretation of manager compositions ranging from 0 to 80 the results, we only briefly discuss the coeffipercent female (the x-axis), with relative gencients before moving to a plot of predicted wages der status one standard deviation above and for men and women under plausible scenarios. below the mean (one line for each). The preFor men, the predicted wage at the mean of dictions show men’s wages are consistently all control variables is the model intercept: 2.888 lower where the percentage female among manin logged dollars or $17.96 per hour. This reflects agers is higher (consistent with the suggestion an all-male job with no female managers and a that women managers are concentrated in less net difference of zero because those variables are highly-paid industries). Among female workers, not centered. The individual-level gender effect wages fall as a function of female management is –.126, which means that women in the same where those managers are of low status, but job are predicted to earn 2.888 – .126 = 2.762, wages rise where female managers are of high or $15.83, for a gender wage ratio of .88. To status. For extend the example simply, if 50 percent of theby Ingenta to : our focus on the gender wage gap, Delivered managers in that local industry are women, those North result is clear. Net of controls at all three levUniversity of the Carolina els, the higher Mon, manmanagers are of equal status as the male12 Nov 2007 18:16:31 presence of female managers in local industries is associated with a reduced agers (ND = 0), and the job remains all-male, gender wage gap among nonmanagerial workthen the predicted wage for men would be the ers only where those managers hold relatively model intercept plus the manager percent female high-status positions. In the figure, the gender effect: 2.888 + (–.002 50) = 2.788, or $16.25. wage ratio is constant at .81, where female manTo calculate the predicted wage for women in agers’ relative status is one standard deviation such a job, we start with men’s wage, then add below the mean, but narrows from .76 to .99 the individual gender effect as modified by the when that status is one standard deviation above manager composition of 50 percent female: the mean and manager composition rises to 80 2.788 + [–.126 + (–.0003 50)] = 2.747, or percent, closing the gender gap.21 $14.11, for a gender wage ratio of .87. The results support both of our hypotheses. This simple example is not realistic, howevFirst, on average (i.e., midway between the high er, because it reflects men and women working and low lines in Figure 2) the gender gap is in an all-male job with equal status between smaller in local industries where a high promale and female managers. When we assess the gender gap at different levels of job and managerial gender composition, and female men’s wages start slightly upward but then turn managers’ relative status, a more conclusive sharply downward after 16 percent female (the inflecresult emerges. Specifically, managerial comtion point is [–1 .124]/[2 –.385] = .16). The nonposition and relative status largely work on the linearity is much less pronounced for women, gender gap through the effects of job composiresulting in a negative net effect of female job comtion.20 Note the significant effects of manager position. The baseline effect of job composition is such that, as female representation at the job level rises, 20 21 Such extreme cases are rare but do exist. In the sample, 90 percent of employees work under managerial pools that are between 12 and 67 percent female. 698—–AMERICAN SOCIOLOGICAL REVIEW Figure 2. Predicted Wages for Men and Women, by Manager Percent Female and Relative Status portion of the managers are women. This effect research design and choice of control variables is small, though, and results primarily from a help to minimize their effects. Endogeneity Delivered Ingenta to : negative effect on average (men’s) wages; net of bycould confound our results to the extent that University of North Carolina female managers’ relative status, women’s wages unmeasured variables are driving both the genMon, 12 Nov 2007 18:16:31 are largely unaffected by manager percent der wage gap and the representation of women female. The results also strongly support our in management. However, we include controls second hypothesis, which posits an interaction for key indicators of gender inequality, includbetween the representation and relative status of ing gender segregation and demand for female female managers. Specifically, we hypothesize labor in the local labor market, as well as a less gender wage inequality in local industries proxy for the level of rationalization or bureauwith (1) more female managers and (2) female managers averaging higher-status positions. cratization of the local industry, which is likeThis is consistent with our data. The gender ly to reduce the reliance on ascriptive wage gap is smaller under female managers, characteristics in determining work-related and this effect is much stronger when those rewards. Differences across local industries in female managers are of relatively high status. endogenous factors that drive both the wage Thus, although female managers may boost gap and female representation in management women’s wages relative to men, the represenare likely to be captured by these controls. tation of low-status female managers alone does Controls at the job and individual levels play not affect the gender wage gap. It is represensimilar roles. In addition, because the job-level tation in upper-status managerial positions that is associated with the gender wage gap. intercepts of our models are permitted to vary across local industries, and the individual-level intercepts vary across jobs, our analysis accounts ENDOGENEITY for unobserved factors influencing average Although the statistical relationship is strong, wages across local industries. In essence, then, potential sources of error or bias exist in our our results reflect within-job and within-localanalysis. In particular, endogeneity and the industry effects. This improves confidence in omission of potentially relevant control variour results. ables may be sources of bias, although our WORKING FOR THE WOMAN—–699 limited. Despite two decades of movement toward managerial integration, we still do not An additional problem could result from our have conclusive evidence that the entry of some definition of managerial authority. Identifying women into managerial positions has brought managers using occupational classifications is material benefits to the majority who remain perhaps more straightforward than identifying below. those who are not managerial. Some people To address this question, we offer the first identified as nonmanagerial in our sample (e.g., large-scale analysis using nationally represendoctors) may wield considerable authority, in tative data on workers tied to managers and some cases, more authority than those counted including theoretically relevant variables at the as managers (e.g., administrative service manlevel of the individual, job, and local industry. agers in doctors’ offices). This represents the We consider the relative status of female manlong-standing problem of intermediate class agers as well as their numeric representation in locations that preoccupies some analysts of models of the gender wage gap. Specifically, we class (e.g., Wright 1997). Excluding selftest two hypotheses: (1) female workers earn employed workers presumably helps reduce the more when their local industries include more number of such cases, but many workers in prowomen among the managerial ranks and (2) fessional occupations remain who have ambigusuch representation is more beneficial when ous authority relations. the relative status of female managers is highTo verify that our results are not unduly influer. enced by such ambiguity, we reestimated our Although we cannot draw causal conclusions regression models but excluded from the nonfrom our data, our results are consistent with the managerial sample those in “business and finanargument that female managers do matter. In the cial operations” or “professional and related” models, the representation of women in manoccupations. This reduced the worker sample by agement reduces the wage gap. The interaction more than one-third (39 percent of women and between : Delivered by Ingenta to managerial gender composition and 29 percent of men). In fact, three of the largest female managers’ relative status, however, University of North Carolina occupations in our original sample were excludMon, 12reg- 2007 18:16:31 relationship is much stronger in Nov shows that the ed: elementary and middle school teachers, local industries where female managers hold relistered nurses, and accountants and auditors atively high-status positions. The addition of (these three occupations alone accounted for women at the low end of the managerial hieralmost 10 percent of the original sample). The archy may have weak effects on the gender largest remaining nonmanagerial occupations wage gap, but the potential effects of high-stawere unambiguously nonmanagerial: secretus female managers are much more positive. taries, truck drivers, and customer service repThis finding highlights—in a new way—the resentatives together compose 13 percent of the significance of the “glass ceiling.” If our findreduced sample. Results from this alternative ings hold, not only are qualified women blocked specification (not shown, but available from from upper-level managerial positions and the authors upon request) are substantively idendenied the benefits of those jobs, but their tical to those reported above. All the central absence has ripple effects that shape workplace coefficients are of the same or larger magnitude outcomes for nonmanagerial women as well. To and remain statistically significant. the extent that women continue to cluster at the low end of managerial hierarchies, our findDISCUSSION AND CONCLUSIONS ings may temper the optimism generated by the rapid increase in the proportion of women in After analyzing a survey of managers’ attitudes management in the last several decades. We are nearly two decades ago, Brenner and colleagues also given pause by the finding that, with all the (1989:668) concluded: “Unlike her male councontrols in the model, wages for men are lower terpart, today’s female manager would be in local industries with more female managers. expected to treat men and women equally in This suggests that female managers remain conselection, promotion, and placement decisions.” centrated in workplace settings with lower Unfortunately, as they noted, women in manwages across the board, in ways that we cannot agement mostly held lower- and middle-management positions, where their impact was capture with the variables used here. On the ALTERNATIVE SPECIFICATION 700—–AMERICAN SOCIOLOGICAL REVIEW other hand, our findings imply that inroads with our examination of labor markets, these studies confront possible confounding effects made by women into upper-status managerial across multiple levels of social interaction. positions will “lift all boats” by also boosting Researchers in these disparate fields should the wages of women employed in nonmanagelearn from each other. rial occupations. All women may benefit from Further theorizing and data collection efforts the desegregation of managerial occupations, undoubtedly will help to overcome the weakeven those who do not themselves attain such nesses in our case. For example, we are not able positions—which compliments Cotter and colto link workers and managers in their actual leagues’ (1997) finding that all women benefit work settings. Just as workplace processes genfrom occupational desegregation. erating inequality vary across organizational In addition, our results are consistent with settings (Baron and Newman 1990; Huffman Jacobs’s (1992) notion of “title inflation”—the and Velasco 1997), female managers’ ability to reclassification of previously nonmanagerial alter wage setting practices also may vary workers as managers with little change in markedly across workplace contexts. On the authority or wages—in that the mere represenother hand, in our analysis the presence and tation of women in management has limited status of female managers is measured at the effects on the gender wage gap. Importantly, our level of local industries, which may be the relanalysis shows that simply looking at the perevant organizational field within which employcentage of females among managers is not ers make crucial gender-related decisions. This enough if one is interested in the effects of manpermits us to extend existing research on the agerial access. In this case, including the relagender wage gap to incorporate managerial tive status of female managers is necessary— characteristics using high-quality, large-scale this assertion garners strong support from our Census data that links population-based samples multivariate results. of workers to managers. We hope this innovaOur results are intrinsically sociological and these broadly provocative. We hope they will entice bytion, and to : results, will generate increased Delivered Ingenta attention to the others to investigate not only the role of man- of North Carolina role managers play in producUniversity ing and sustaining labor market inequality, and agers, but also analogous cases of subordinate Mon, 12 Nov 2007 18:16:31 by extension to the potential influence of subgroup members in positions of authority. For ordinate group members who attain positions of example, in studies of political alienation and authority. participation, research shows that the presence of Black (Bobo and Gilliam 1990) and Latino Philip N. Cohen is an associate professor of sociol(Pantoja and Segura 2003) elected leaders ogy at the University of North Carolina at Chapel Hill increases political empowerment among memand a faculty fellow at the Carolina Population bers of those groups. Female and minority politCenter. His research concerns social inequality in ical leaders appear to be more responsive to the labor markets and families. With Matt Huffman, he concerns of their ascriptively similar conis currently investigating the underrepresentation of women and minorities in management and its constituents (Bratton and Haynie 1999; Mansbridge sequences. Another area of his research focuses on 1999), but the long-term implications of such family structure and inequality within and between integration remain to be seen. In education, stufamilies across social contexts, including household dents receive more positive evaluations and living arrangements, women’s employment, and feedback from teachers who match their unpaid labor. race/ethnicity and gender, but the evidence that Matt L. Huffman is an associate professor of socithis results in better educational outcomes is limology and Co-Director of Graduate Studies at the ited and mixed (Butler and Christensen 2003; University of California, Irvine. His research focusDee 2005; Downey and Pribesh 2004; es on race and gender inequality in work organiza22 As Ehrenberg, Goldhaber, and Brewer 1995). tions and across labor markets. Among other things, this work examines the wage effects of segregation, 22 Ehrenberg and colleagues (1995:548) note: “The relationships between supervisors and employees is analogous, in important respects, to that between teachers and students.” They conclude, “Research addressing these issues should be high on the priority list of those concerned with the progress of women and minorities in the labor market” (p. 560). WORKING FOR THE WOMAN—–701 gender and racial differences in access to managerial jobs, and the consequences of changing managerial composition for inequality among workers. APPENDIX VARIANCE COMPONENTS The hierarchical linear models permit us to decompose the total variance in wages into three component parts: (1) that which occurs between individuals and the mean wages in their jobs, (2) that which occurs between mean wages in each job and the mean wages for all jobs in their local industries, and (3) that which occurs between the mean wages for local industries. Results from the unconditional model (which includes no independent variables at any level) in Table A show that overall, 72.3 percent of wage variation among nonmanagerial workers occurs within jobs as we have defined them; 14.8 percent occurs between jobs but within local industries; and 12.9 percent occurs across local industries. This construction differs from previous threelevel wage decompositions (e.g., Cohen and Huffman 2003a) in its use of local industries as the third level rather than metropolitan labor markets. Not surprisingly, variation across local industries accounts for a larger share of the total variation than does metropolitan-area labor markets. The differences between average wages for the restaurant industry versus the computer systems industry within a given labor market would be expected to be larger than differences in average wages between New York and Los Angeles overall. The variance components also show that our variables do much more to explain variance between jobs and local industries (variance components are reduced by about 80 percent each from the unconditional model to the final model) than they do to explain variation between individuals, which is only reduced 8 percent in the full model. Finally, Table A shows that about three-fourths of the variance in the effect of being female occurs across jobs but within industries. In the final model, 42 percent of that within-local-industry variance is explained, as the variance component is reduced from .0104 to .0068. Table A.—Variance Components for Three-Level Hierarchical Linear Regressions Model (UM) Female Dummy Full Model Delivered by Ingenta to : University of North Carolina Mon, 12 Nov 2007 18:16:31 UM + Unconditional Variance Percent Variance Percent Variance Percent Component of Total Component of Total Component of Total Intercept —Individual .3459 72.3 —Job .0708 14.8 —Local Industry .0620 12.9 —Total 100.0 Female —Job —Local Industry —Total Job Percent Female (on intercept) —Local Industry Job Percent Female Squared (on intercept) —Local Industry Job Percent Female (on individual female effect) —Local Industry Job Percent Female Squared (on individual female effect) —Local Industry .3404 .0718 .0623 71.7 15.1 13.1 100.0 75.1 24.9 100.0 .3191 .0146 .0118 92.3 4.2 3.4 100.0 56.9 43.1 100.0 100.0 100.0 100.0 100.0 .0104 .0035 .0068 .0051 .0776 .1031 .1055 .1051 Note: All variance components have a two-tailed p-value of less than .05. REFERENCES Ashenfelter, Orley and Timothy Hannan. 1986. “Sex Discrimination and Product Market Competition: The Case of the Banking Industry.” Quarterly Journal of Economics 101:149–74. Baron, James N., Brian S. Mittman, and Andrew E. Newman. 1991. “Targets of Opportunity: 702—–AMERICAN SOCIOLOGICAL REVIEW Organizational and Environmental Determinants ———. 2007. “Black Underrepresentation in of Gender Integration Within the California Civil Management Across U.S. Labor Markets.” Annals Service, 1979–1985.” American Journal of of the American Academy of Political and Social Sociology 96:1362–401. Science 609(January):181–99. Baron, James N. and Andrew E. Newman. 1990. Correll, Shelley J. and Stephen Benard. 2005. “For What It’s Worth: Organizations, Occupations, “Getting a Job: Is There a Motherhood Penalty?” and the Value of Work Done by Women and Presented at the Annual Meeting of the American Nonwhites.” American Sociological Review Sociological Association, August, San Francisco, 55:155–75. CA. Baunach, Dawn M. 2002. “Progress, Opportunity, and Cotter, David A., JoAnn DeFiore, Joan M. Hermsen, Backlash: Explaining Attitudes Toward GenderBrenda M. Kowalewski, and Reeve Vanneman. Based Affirmative Action.” Sociological Focus 1997. “All Women Benefit: The Macro-Level 35:345–62. Effect of Occupational Integration on Gender Beckman, Christine M. and Damon J. Phillips. 2005. Earnings Inequality.” American Sociological “Interorganizational Determinants of Promotion: Review 62:714–34. Client Leadership and the Attainment of Women ———. 1998. “The Demand for Female Labor.” Attorneys.” American Sociological Review American Journal of Sociology 103:1673–712. 70:678–701. Cotter, David A., Joan M. Hermsen, Seth Ovadia, and Blum, Terry C., Dail L. Fields, and Jodi S. Goodman. Reeve Vanneman. 2001. “The Glass Ceiling 1994. “Organization-Level Determinants of Effect.” Social Forces 80:655–81. Women in Management.” Academy of Cotter, David A., Joan M. Hermsen, and Reeve Management Journal 37:241–68. Vanneman. 2004. Gender Inequality at Work. New Bobo, Lawrence and Franklin D. Gilliam Jr. 1990. York: Russell Sage Foundation. “Race, Sociopolitical Participation, and Black Deaux, Kay. 1985. “Sex and Gender.” Annual Review Empowerment.” American Political Science of Psychology 36:49–81. Review 84:377–93. Dee, Thomas S. 2005. “A Teacher Like Me: Does Bratton, Kathleen A. and Kerry L. Haynie. 1999. Race, Ethnicity, or Gender Matter?” American “Agenda Setting and Legislative Success in State Economic Review 95:158–65. Legislatures: The Effects of Gender and Race.” byDenmark,to : Delivered Ingenta Florence L. 1993. “Women, Leadership, Journal of Politics 61:658–79. and Empowerment.” Psychology of Women University of North Carolina Brenner, O. C., Joseph Tomkiewicz, and Mon, 12 Nov 2007 18:16:31 Victoria E. Quarterly 17:343–56. Schein. 1989. “The Relationship Between SexDiMaggio, Paul J. and Walter W. Powell. 1983. “The Role Stereotypes and Requisite Management Iron Cage Revisited: Institutional Isomorphism Characteristics Revisited.” Academy of and Collective Rationality in Organizational Management Journal 32:662–69. Fields.” American Sociological Review 48:147–60. Butler, Daniel M. and Ray Christensen. 2003. Dobbins, Gregory H. and Stephanie J. Platz. 1986. “Mixing and Matching: The Effect on Student “Sex Differences in Leadership: How Real Are Performance of Teaching Assistants of the Same They?” Academy of Management Review Gender.” Political Science and Politics 36:781–86. 11:118–27. Carrington, William J. and Kenneth R. Troske. 1995. Downey, Douglas B. and Shana Pribesh. 2004. “Gender Segregation in Small Firms.” Journal of “When Race Matters: Teachers’ Evaluations of Human Resources 30:503–33. Students’ Classroom Behavior.” Sociology of Charles, Maria and David B. Grusky. 2004. Education 77:267–82. Occupational Ghettos: The Worldwide Segregation Duncan, Otis Dudley and Beverly Duncan. 1955. “A of Women and Men. Stanford, CA: Stanford Methodological Analysis of Segregation Indexes.” University Press. American Sociological Review 20:210–17. Cohen, Lisa E., Joseph P. Broschak, and Heather A. Ehrenberg, Ronald G., Daniel D. Goldhaber, and Haveman. 1998. “And Then There Were More? The Dominic J. Brewer. 1995. “Do Teachers’ Race, Effect of Organizational Sex Composition on the Gender, and Ethnicity Matter? Evidence from the Hiring and Promotion of Managers.” American National Educational Longitudinal Study of 1988.” Sociological Review 63:711–27. Industrial and Labor Relations Review 48:547–61. Cohen, Philip N. and Matt L. Huffman. 2003a. Elliott, James R. and Ryan A. Smith. 2004. “Race, “Individuals, Jobs, and Labor Markets: The Gender, and Workplace Power.” American Devaluation of Women’s Work Across U.S. Labor Sociological Review 69:365–86. Markets.” American Sociological Review Ely, Robin J. 1995. “The Power in Demography: 68:443–63. Women’s Social Constructions of Gender Identity ———. 2003b. “Occupational Segregation and the at Work.” Academy of Management Journal Devaluation of Women’s Work Across U.S. Labor 38:589–634. Markets.” Social Forces 81:881–908. England, Paula, Melissa S. Herbert, Barbara S. WORKING FOR THE WOMAN—–703 Kilbourne, Lori L. Reid, and Lori M. Megdal. of Occupational Sex Segregation.” Work and 1994. “The Gendered Valuation of Occupations Occupations 30:30–61. and Skills: Earnings in 1980 Census Occupations.” Hultin, Mia and Ryszard Szulkin. 1999. “Wages and Social Forces 73:65–100. Unequal Access to Organizational Power: An England, Paula, Joan M. Hermsen, and David A. Empirical Test of Gender Discrimination.” Cotter. 2000. “The Devaluation of Women’s Work: Administrative Science Quarterly 44:453–72. A Comment on Tam.” American Journal of ———. 2003. “Mechanisms of Inequality: Unequal Sociology 105:1741–51. Access to Organizational Power and the Gender Fields, Judith and Edward N. Wolff. 1995. Wage Gap.” European Sociological Review “Interindustry Wage Differentials and the Gender 19:143–59. Wage Gap.” Industrial and Labor Relations Review Ibarra, Herminia. 1992. “Homophily and Differential 49:105–20. Returns: Sex Differences in Network Structure Gilligan, Carol. 1982. In a Different Voice: and Access in an Advertising Firm.” Administrative Psychological Theory and Women’s Development. Science Quarterly 37:422–47. Cambridge, MA: Harvard University Press. Jackson, Robert M. 1998. Destined for Equality: Glascock, Jack. 2001. “Gender Roles on Prime-Time The Inevitable Rise of Women’s Status. Cambridge, Network Television: Demographics and MA: Harvard University Press. Behaviors.” Journal of Broadcasting and Jacobs, Jerry A. 1992. “Women’s Entry into Electronic Media 45:656–69. Management: Trends in Earnings, Authority and Goldstein, Joshua R. and Anne J. Morning. 2002. Values Among Salaried Managers.” Administrative “Back in the Box: The Dilemma of Using MultipleScience Quarterly 37:282–301. Race Data for Single-Race Laws.” Pp. 119–37 in Jaffee, Sara and Janet Shibley Hyde. 2000. “Gender The New Race Question: How the Census Counts Differences in Moral Orientation: A MetaMultiracial Individuals, edited by J. Perlmann and Analysis.” Psychological Bulletin 126:703–26. M. Waters. New York: Russell Sage Foundation. Kanter, Rosabeth M. 1977. Men and Women of the Gorman, Elizabeth H. 2005. “Gender Stereotypes, Corporation. New York: Basic Books. Same-Gender Preferences, and Organizational Konrad, Alison M. and Jeffery Pfeffer. 1991. Variation in the Hiring of Women: Evidence from “Understanding the Hiring of Women and Law Firms.” American Sociological Reviewby Ingenta to : Delivered Minorities in Educational Institutions.” Sociology 70:702–28. University of North Carolina 64:141–57. of Education Greve, Henrich R. 2005. “Interorganizational 2007 18:16:31 Mon, 12 Nov Kulis, Stephen. 1997. “Gender Segregation Among Learning and Heterogeneous Social Structure.” College and University Employees.” Sociology of Organization Studies 26:1025–47. Education 70:151–73. Guo, Guang and Hongxin Zhao. 2000. “Multilevel Lauzen, Martha M. and David M. Dozier. 1999. Modeling for Binary Data.” Annual Review of “Making a Difference in Prime Time: Women on Sociology 26:441–62. Screen and Behind the Scenes in the 1995–96 Halpert, Jane A., Midge L. Wilson, and Julia L. Television Season.” Journal of Broadcasting & Hickman. 1993. “Pregnancy as a Source of Bias Electronic Media 43:1–19. in Performance Evaluations.” Journal of Lieberson, Stanley. 1976. “Rank-Sum Comparison Organizational Behavior 14:649–63. Between Groups.” Sociological Methodology Huffman, Matt L. and Philip N. Cohen. 2004a. 7:276–91. “Occupational Segregation and the Gender Gap in Mansbridge, Jane. 1999. “Should Blacks Represent Workplace Authority: National Versus Local Labor Blacks and Women Represent Women? A Markets.” Sociological Forum 19:121–47. Contingent ‘Yes.’” Journal of Politics 61:628–57. ———. 2004b. “Racial Wage Inequality: Job Massini S., A. Y. Lewin, and Henrich R. Greve. 2005. Segregation and Devaluation Across U.S. Labor “Innovators and Imitators: Organizational Markets.” American Journal of Sociology Reference Groups and Adoption of Organizational 109:902–36. Routines.” Research Policy 34:1550–69. Huffman, Matt L. and Lisa Torres. 2002. “Is It Only McCall, Leslie. 2001. Complex Inequality: Gender, Who You Know That Matters? Gender and Social Class and Race in the New Economy. New York: Networks Among Professional, Technical, and Routledge. Managerial Workers.” Gender and Society McPherson, Miller, Lynn Smith-Lovin, and James M. 16:793–813. Cook. 2001. “Birds of a Feather: Homophily in Huffman, Matt L. and Steven C. Velasco. 1997. Social Networks.” Annual Review of Sociology “When More Is Less: Sex Composition, 27:415–44. Organizations, and Earnings in U.S. Firms.” Work Merton, Robert K. 1940. “Bureaucratic Structure and Occupations 24:214–44. and Personality.” Social Forces 18:560–68. Hultin, Mia. 2003. “Some Take the Glass Escalator, Mouw, Ted. 2003. “Social Capital and Finding a Job: Some Hit the Glass Ceiling? Career Consequences 704—–AMERICAN SOCIOLOGICAL REVIEW Gender and Context on the Self and Moral Do Contacts Matter?” American Sociological Review 68:868–98. Reasoning.” Psychology of Women Quarterly Nelson, Robert L. and William P. Bridges. 1999. 28:246–55. Legalizing Gender Inequality: Courts, Markets, Shenhav, Yehouda and Yitchak Haberfeld. 1992. and Unequal Pay for Women in America. New “Organizational Demography and Inequality.” York: Cambridge University Press. Social Forces 71:123–43. Neumark, David and Wendy A. Stock. 2006. “The Smith, Ryan A. 2002. “Race, Gender, and Authority Labor Market Effects of Sex and Race in the Workplace: Theory and Research.” Annual Discrimination Laws.” Economic Inquiry Review of Sociology 28:509–42. 44:385–419. South, Scott J. and Weiman Xu. 1990. “Local Pantoja, Adrian D. and Gary M. Segura. 2003. “Does Industrial Dominance and Earnings Attainment.” Ethnicity Matter? Descriptive Representation in American Sociological Review 55:591–99. Legislatures and Political Alienation Among Strang, David and Sarah A. Soule. 1998. “Diffusion Latinos.” Social Science Quarterly 84:441–60. in Organizations and Social Movements: From Pfeffer, Jeffery. 1983. “Organizational Demography.” Hybrid Corn to Poison Pills.” Annual Review of Research in Organizational Behavior 5:299–357. Sociology 24:265–90. Pfeffer, Jeffery and Alison Davis-Blake. 1987. “The Tam, Tony. 1997. “Sex Segregation and Occupational Effect of the Proportion of Women on Salaries: The Gender Inequality in the United States: Case of College Administrators.” Administrative Devaluation or Specialized Training?” American Science Quarterly 32:1–24. Journal of Sociology 102:1652–92. Pfeffer, Jeffery, Alison Davis-Blake, and Daniel J. Tomaskovic-Devey, Donald and Sheryl Skaggs. 2002. Julius. 1995. “AA Officer Salaries and Managerial “Sex Segregation, Labor Process Organization, Diversity: Efficiency Wages or Status?” Industrial and Gender Earnings Inequality.” American Relations 34:73–94. Journal of Sociology 108:102–28. Raudenbush, Stephen W. and Anthony S. Bryk. 2002. Tomaskovic-Devey, Donald and Kevin Stainback. Hierarchical Linear Models: Applications and 2007. “Discrimination and Desegregation: Equal Data Analysis Methods. 2d ed. Thousand Oaks, Opportunity Progress in U.S. Private Sector CA: Sage. Workplaces Since the Civil Rights Act.” Annals of Reskin, Barbara F. 2003. “Including Mechanisms in by Ingenta to : Delivered the American Academy of Political and Social Our Models of Ascriptive Inequality.” American of North Carolina University Science 609:49–84. Sociological Review 68:1–21. Mon, 12 Nov 2007 18:16:31and Charles A. O’Reilly III. 1989. Tsui, Anne S. Reskin, Barbara F. and Debra B. McBrier. 2000. “Beyond Simple Demographic Effects: The “Why Not Ascription? Organizations’ Employment Importance of Relational Demography in Superiorof Male and Female Managers.” American Subordinate Dyads.” The Academy of Management Sociological Review 65:210–33. Journal 32:402–23. Reskin, Barbara F. and Catherine Ross. 1992. “Job, U.S. Census Bureau. 2003. Census 2000, Public Use Authority, and Earnings Among Managers: The Microsample, Technical Documentation. Continuing Signif icance of Sex.” Work and Washington, DC: U.S. Census Bureau. Occupations 19:342–65. van Engen, Marloes L. and Tineke M. Willemsen. Ridgeway, Cecilia L. 1997. “Interaction and the 2004. “Sex and Leadership Styles: A MetaConservation of Gender Inequality: Considering Analysis of Research Published in the 1990s.” Employment.” American Sociological Review Psychological Reports 94:3–18. 62:218–35. Wharton, Amy S. 1986. “Industry Structure and Robinson, Cor re L., Tiffany Taylor, Donald Gender Segregation in Blue-Collar Occupations.” Tomaskovic-Devey, Catherine Zimmer, and Social Forces 64:1025–31. Matthew W. Irvin Jr. 2005. “Studying Race or Wright, Erik O. 1997. Class Counts: Comparative Ethnic and Sex Segregation at the Establishment Studies in Class Analysis. Cambridge, UK: Level: Methodological Issues and Substantive Cambridge University Press. Opportunities Using EEO-1 Reports.” Work and Wright, Erik O. and Janeen Baxter. 2000. “The Glass Occupations 32:5–38. Ceiling Hypothesis: A Comparative Study of the Roth, Louise Marie. 2004. “The Social Psychology United States, Sweden, and Australia.” Gender of Tokenism: Status and Homophily Processes on and Society 14:275–94. Wall Street.” Sociological Perspectives Young, Iris M. 1994. “Gender as Seriality: Thinking 47:189–214. About Women as a Social Collective.” Signs Ryan, Michelle K., Barbara David, and Katherine J. Reynolds. 2004. “Who Cares? The Effect of 19:713–38.