Race, Gender, and Workplace Power

by James R. Elliott
Race, Gender, and Workplace Power
James R. Elliott
American Sociological Review
Start Page: 
End Page: 
Select license: 
Select License

Race, Gender, and Workplace Power

James R. Elliott Ryan A. Smith Tulane University The City University of New York

Survey data support hypotheses regarding dzfferential access to workplace power among women and minorities relative to white men. Specific findings indicate that, relative to white men, all groups encounter increasing inequality at higher levels ofpowel; but only black women seem to experience this form of inequality as a result of direct discrimination. Further analysis indicates that network assistance is more a response to this form of discrimination than an indirect cause. Finally, analysis shows that most groups attain power through homosocial reproduction, but what differs is the opportunity to engage in such reproduction, wherein white men excel. These findings imply that while women and minorities face lower odds than white men of achieving higher levels of workplace powel; the reasons for this disadvantage vary among respective groups and

thus will likely require dzfferent remedies.

ower, defined as "control over resources, people, and things" (Wolf and Fligstein 1979), is an essential aspect of social stratifi- cation (Bendix 1956; Braverman 1974; Dahrendorf 1959). In this study we focus on one dimension of power: authority and control over others in the workplace. Weber ([I9141 1968) conceptualized this dimension of power as a form of "legitimate authority" because it derives from organizational positions that people occu- py rather than from the people themselves. We concentrate specifically on how race, ethnicity, and gender affect the likelihood of attaining successively higher levels of such authority.

This issue merits attention for several reasons. Despite significant progress in the socioeco- nomic status of working women and minorities, data analyses at the city, state, national, and cross-national levels continue to document race and gender inequalities in workplace power (see Smith 2002 for a recent review). Moreover, fed- eral reports indicate that discrimination claims pertaining to promotion have risen steadily in recent years, while discrimination claims per-

Direct correspondence to James Elliott, Tulane University, Sociology Department, 220 Newcomb Hall, New Orleans, LA 701 18 Cjre@tulane.edu). The authors thank Charlie Brody, Beth Rubin, Sanders Korenman, Neil Bennett, and anonymous reviewers for their helpful comments on prior drafts.

taining to hiring have declined (Myerson 1997). These patterns suggest that workplace power constitutes a central battleground in struggles for equalizing opportunities in coming years-a possibility fueled by rising expectations of unprecedented numbers of women and minori- ties now employed in the formal economy.

Another reason for investigating racial and gender inequalities in workplace power is that despite the popularity of the "glass ceiling" metaphor, surprisingly little, direct research exists on related assumptions. For example, numerous quantitative studies document ascrip- tive inequalities in workplace power, but few examine whether these inequalities increase at higher levels of power, and none adequately account for factors that ethnographic research contends are important, namely, the ascriptive contexts in which positions of power are embed- ded and the networks that supposedly play key roles in the allocation of these positions to par- ticular individuals. Conversely, ethnographic research is insightful, but it only comes from a few, select case studies (e.g., Kanter 1977).

plement this article, please see the ASR Web site



Consequently, it remains uncertain just how generalizable the claims from these studies are and the selection biases they might reflect.

Finally, prior quantitative studies on ascrip- tive inequalities in workplace power focus either on race or gender, but not both (for notable exceptions, see McGuire and Reskin 1993; Tomaskovic-Devey 1993). This conventional approach is problematic because it reinforces the erroneous assumption that racial stratification and gender stratification are mutually exclu- sive systems that we can somehow sum to understand differences among non-white-men. By contrast, we view racial, ethnic, and gender stratifications as having fundamentally similar causes, and how these causes overlap to produce different outcomes for different groups remains a decidedly empirical question. In taking this approach, it is important to move beyond exclu- sively black-white comparisons to consider Latinos, not only because Latinos constitute the largest and fastest growing panethnic group in U.S. society, but also because recent research indicates that employers of all ethnoracial back- grounds tend to prefer Latinos over blacks when filling positions in their organizations (Moss and Tilly 2001 ;Wilson 1996).

In the present study we address these short- comings in prior research through an examina- tion of authority attainment among white, black, and Latino men and women. Our objectives are twofold: (1) assess the extent to which inequal- ity in workplace power increases among women and minorities, relative to white men, at higher levels of power; and (2) examine the mecha- nisms of allocation responsible for this form of inequality for each group. In pursuing these objectives, we focus specifically on power posi- tions under the supervision of others. One rea- son for this focus is that workplace power is not limited to the upper echelon of Fortune 500 corporations; it is found in all places of employment. While popular reporting on the glass ceiling often obscures and even trivializes the phenom- enon, it is critical to general understanding because even seemingly mundane jobs are not left to run themselves, as the lack of autonomy in most low-slull positions attests. Instead, people are selected to fill positions of power throughout all levels of the workforce. And mid-level posi- tions not only constitute the most common form of legitimate authority, but they also represent positions wherein vertical and horizontal inter- group competition is likely to be greatest among incumbents of roughly equal credentials.


A popular explanation for ascriptive inequali- ties in workplace power invokes the metaphor of a glass ceiling, which was first popularized in a 1986 Wall Street Journal article that described barriers women often face as they climb corporate ladders. In 1995, the United States Department of Labor issued an official report on this subject, noting that during the interim decade, observers had extended the term two ways: first, to include racial and ethnic minorities in addition to women; and second, to refer to all management and decision-making positions, not just to top-level positions at large corporations (Federal Glass Ceiling Commission 1995:iii). A core idea invoked by the metaphor is that while employers might let women and minorities into low positions of authority, they are much less likely to let them into high positions that involve greater control.

While we do not explicitly test for corporate glass ceilings in this research, we do examine whether women and minorities have an increas- ingly difficult time, relative to white men, accessing jobs with greater organizational power. Recent research along these lines has taken one of two general approaches. The more restrictive approach views this type of inequal- ity in terms of an absolute barrier that blocks women and minorities from higher positions of workplace power because they are women and minorities (Jacobs 1992; Morrison and Glinow 1990; Reskin and McBrier 2000; Reskm and Ross 1992). Taken literally, this perspective implies an invisible barrier below which women and minorities attain a modest degree of work- place power (e.g., supervisory authority) and above which they do not (e.g., managerial con- trol). A less restrictive approach views increas- ing inequality as a form of disadvantage facing women and minorities, relative to white men, which intensifies at higher levels of workplace power.

This second approach to conceptualizing increasing inequality has two important impli- cations for empirical assessment. First, inequal- ity is presumed to occur not in a single, absolute step, but rather over several steps of increasing magnitude relative to white men. Second, and

as such, a declining share of women andfor minorities in positions of higher power offers necessary but insufficient evidence of increas- ing inequality. Instead, sufficient evidence requires decreasing probabilities of advance- ment, relative to white men, at higher levels of power. To illustrate, Table 1 depicts nonsup- portive and supportive evidence for increasing inequality in a simple three-level hierarchy (worker, supervisor, manager). In this table, the "white-black ratio" is the key statistic because it summarizes black men's probability of advancement, relative to white men, at two suc- cessive levels of power: first, from worker to supervisor; and second, from supervisor to man- ager. In the panel labeled "lack of support for increasing inequality," the relative ratio of advancement for black men decreases from 2.5 to 2.0 with movement up the hierarchy. By con- trast, in the panel labeled "support for increas- ing inequality," the relative ratio for black men increases from 2.5 to 4.0. This difference in supportive and nonsupportive evidence for increasing inequality, relative to white men, occurs despite the fact that the share of black men decreases with movement up the hierarchy in both panels (see columns labeled "% black men").

In the present research we focus on this odds- based criterion for assessing inequality and operationalize this focus via the following hypothesis:

Hypothesis I: Women and minorities' odds of advancement decrease, relative to white men, at higher levels of power.

In the most direct test of this hypothesis to date, Baxter and Wright (2000) use a six-level index of workplace power to examine cross-national


differences in gender inequalities and find no evidence in the United States that women's like- lihood of advancement, relative to men, declines at successively higher levels of power (see also Yamagata et al. 1997). Their small sample sizes, however, prevent the authors from showing increasing inequality for racial minorities at higher levels of power, particularly women of color. On this subject, studies by Cotter et al. (2001) and Morgan (1998) are instructive, albeit indirect. Both studies use longitudinal data to examine wage inequality in individual careers but they reach different conclusions about the presence of increasing racial and gender disad- vantage, relative to white men, at later stages. For example, Cotter et al.'s analysis of data from the Panel Survey of Income Dynamics reveals increasing wage inequality for white and black women, relative to white men, but not for black men. From this evidence, the authors conclude that increasing inequality in wages in individ- ual careers is more reflective of gender than racial stratification. Morgan (1998), by con- trast, uses a single-cohort longitudinal design with data from the Survey of Natural and Social Scientists and Engineers and a multi-cohort, cross-sectional design with data from the 1992 Survey of Women and Men Engineers. From these analyses, Morgan concludes that increased pay gaps among men and women at later career stages are more reflective of cohort-replace- ment dynamics-that is, past inequalities work- ing themselves through the system-than increasing gender inequality among men and women as they progress in their careers.

For criticism of Morgan's analysis and conclu- sions, see Alessio and Andrzejewski (2000), with reply by Morgan (2000).

Table 1. Inequality among Men at Higher Levels of Workplace Power

Men in Power Level Odds of Advancement to Next Level of Power

Power Level White (n) Black (n) % Black White Black White-Black Ratio Lack of Support for Increasing Inequality Manager 10 2 17 Supervisor 20 8 29 Worker 30 30 50 Support for Increasing Inequality

Manager 20 2 9
Supervisor 20 8 29
Worker 30 30 50 .67 .27 2.5:1

Note: Data adapted from Baxter and Wright 2000.

Together, these two studies provide key insights into race, gender, and employment inequalities in the United States, but they also leave us unclear about what to expect in analy- ses of workplace power. Some of this uncer- tainty stems from their mixed results, and some of it stems from the fact that they analyze wages, not workplace power. The last point is impor- tant because racial and gender differences in workplace power are an important source of wage inequality (Halaby 1979; Kluegel 1979; Robinson and Kelley 1979; Smith 1997). Thus, to clarify this issue, it is important not only to examine these differences in workplace power directly, relative to white men, but also to exam- ine the mechanisms of allocation commonly presumed responsible for these differences at higher levels of legitimate authority.


OF ALLOCATION. Observers com- monly point to three broad mechanisms respon- sible for increasing ascriptive inequalities at higher levels of workplace power. One mecha- nism is &ect discrimination, which can take two distinct forms: "taste discrimination" in the form of old-fashioned racism and sexism based on out-group prejudice and antipathy; and "sta- tistical discrimination" in which employers use race and sex as proxies for assessing potential productivity in candidates when they lack other information about the candidates. For exam- ple, if women generally are less likely to put work demands above family demands, then employers might use this easy-to-observe trait (sex) to screen and evaluate managerial candi- dates in favor of men, regardless of the (unob- served) work commitment of individual male and female candidates under review. Researchers typically identify these forms of discrimination using a residual (or "net gap") strategy in which disadvantage, relative to (white) men, is evaluated after statistically con- trolling for job-relevant factors, such as educa- tion, experience, and employment context (e.g., Wright, Baxter, and Birkelund 1995). If relative disadvantage for the minority andlor female group in question increases with movement up the power hierarchy, increasing disadvantage is presumed to exist and to be the result of direct discrimination, although the precise form ("taste" or "statistical") is difficult to differen- tiate.

Another broad mechanism presumed to gen- erate increasing inequalities in workplace power involves a more indirect process, namely the exclusion of women and minorities, relative to white men, from networks that regulate access to information, opportunities, and resources needed to advance in the workplace. Research on this subject generally shows that work-relat- ed networks help workers gain skills, acquire legitimacy, and climb promotional ladders (Bridges and Villemez 1986; Campbell and Rosenfeld 1985; Podolny and Baron 1997) and that these resources are important because most employees' job training and career develop- ment come from informal instruction rather than continuing education and explicit on-the- job training (United States Bureau of Labor Statistics 1996).

Additionally, recent research by McGuire (2002) indicates that exclusion of women and minorities can occur within job-related networks, as well as outside such networks. In her study of over a thousand financial-services employees, McGuire found that even when black and white women held jobs in which they had personal ties to the same types of higher- level employees as white men, they received significantly less work-related help from these ties than similarly situated white men. McGuire concludes that this discrepancy arises because network members are less likely to invest in women than (white) men as a result of cultural beliefs that rank women below men. The impli- cation is that workers, not just employers, use race and gender to rank network members, and this ranking influences the type and amount of assistance available to members of different groups. Although McGuire focused primarily on network assistance within organizations, such assistance can also be useful in gaining positions of power across organizations, by providing informal training and contacts necessary to open doors to other employers. On the basis of this rich body of research, we advance the fol- lowing general hypothesis:

Hypothesis 2: Women and minorities' odds of network assistance decrease, relative to white men, at higher levels of power.

A corollary to this hypothesis is that women and minorities often rely more on education and experience, relative to white men, to "break into" higher levels of power, often having to "out-credential" white-male counterparts to compensate for their relative lack of network assistance. Researchers sometimes characterize these distinct modes of advancement as "spon- sored" and "contest" mobility regimes, respec- tively-the first pertaining to network utilization among white men, the second to skills-based competition among women and minorities (e.g., Mueller, Parcel, and Tanaka 1989). While the "contest" regime might seem fair in its empha- sis on objective, skills-based traits, it can lead to relative disadvantage for women and minori- ties for a couple reasons.

First, given practical limits to educational attainment and experience, it becomes increas- ingly difficult to "out-credential" other workers with movement up workplace power hierar- chies, leaving network assistance still a key fac- tor in determining who will advance and who will not. Second, as women and minorities move up organizational chains of command, their out- group, or "other," status often becomes more evident, leaving them more susceptible to infor- mal processes of exclusion and assessment as symbols of an "other" category rather than as individuals. We examine this corollary set of assumptions via the following hypothesis:

Hypothesis 3:Women and minorities' reliance on education and experience intensifies, relative to white men, at higher levels of power.

Support for this hypothesis would mean that regression coefficients for education and expe- rience in equations predicting workplace power would be statistically significant and increas- ingly large for women and minorities, relative to white men, with upward movement in work- place power.

A third mechanism presumed to generate increasing inequality in workplace power is ascriptive dissimilarity with superiors who over- see higher positions of power. In perhaps the best known discussion of these dynamics, Kanter (1977) contends that with movement up orga- nizational hierarchies, power positions become characterized by increasing uncertainty, inter- dependence, and necessity for rapid, accurate communication about murky matters, such as relations between organizational means and ends and criteria for performance evaluation. These job characteristics, in turn, place a pre- mium on discretion and trust among workers selected to advance up the power hierarchy. One way that higher-level managers try to maximize these traits and impose greater predictability on an otherwise uncertain environment is to main- tain relative social homogeneity among indi- viduals they select to fill positions of organizational power beneath them. The under- lying idea is that communication, discretion, and trust are facilitated by social similarity. Higher-level managers prefer this type of rela- tionship over the strain of dealing with people who are different when higher degrees of legit- imate authority are at stake.2

Kanter refers to this process generally as "homosocial reproduction" because it tends to reproduce the social characteristics of organi- zational power structures over successive gen- erations of workers-an idea that traces back to Wilbert Moore's concept of "bureaucratic kin- ship systems" (Moore 1962). Because white men have historically held the reins of power in

U.S. workplaces, they benefit most from these universal tendencies for in-group favoritism as they move up organizational hierarchies, creat- ing increasing inequality for out-group mem- bers. We examine this mechanism of allocation via the following hypothesis:

Hypothesis4:White men's odds of having self- similar superiors increase, relative to women and minorities, at higher levels of power.

This hypothesis assumes that in most organi- zational contexts, homosocial reproduction operates in a vertical fashion, with superiors selecting individuals like themselves to fill power positions below them rather than in a horizontal fashion, with superiors selecting indi- viduals like themselves to fill power positions alongsidethem. To illustrate, consider a simple

This argument is consistent with recent research in "organizational demography" and "new econom- ic sociology" that claims the following: (1) people tend to make sense of their social worlds by catego- rizing others into in-groups and out-groups; and

(2) this normal information processing occurs large- ly outside conscious control, biasing treatment of others because of race, gender, and other discernible traits of group membership (see Pfeffer 1983; Reskin 2002). See Tsui and O'Reilly (1 989) for specifics on importance of ascriptive similarity for performance evaluation in superior-subordinate dyads.

three-level firm in which managers are As, supervisors are Bs, and workers are Cs. Hypothesis 4 implies the following conditions:

(1) As will tend to fill openings for B with indi- viduals like themselves; (2) likewise, Bs will tend to fill openings for C with individuals like themselves;(3) the first tendency will be greater than the second tendency because more power is at stake; and (4) this process benefits white men more than other groups because white men are the group best positioned to benefit from in- group favoritism at higher levels of power.


CONSIDERATIONS. OUT hypothe- ses reflect common assumptions about increas- ing ascriptive inequality at higher levels of workplace power and the mechanisms that pro- duce it. However, these assumptions remain open to several criticisms. First, thus far empir- ical support for the idea that white men unique- ly benefit from network assistance comes from indirect assessments (e.g., Mueller, Parcel, and Tanaka 1989; Wilson 1997). These assessments typically use regression analysis to show that job-relevant factors, such as education and expe- rience, are more predictive of authority attain- ment by women and minorities than by white men, leaving the latter with greater unexplained variance. Researchers then interpret this greater unexplained variance for white men as evidence of the relative importance of unobserved mech- anisms, including network assistance, for white men's power attainment. This interpretation is a logical but assailable inference in the absence of direct measurement of network assistance.

Second, researchers commonly assume that white men exert control over most, if not all, positions of workplace power, particularly those with greater legitimate authority. While this assumption might be true in relative terms, white men's control over U.S. workplaces is not absolute. Third and relatedly, researchers have never demonstrated empirically tendencies toward homosocial reproduction across groups and organizational contexts. Thus it remains uncertain if the process Kanter describes is uni- versal, or if white men are more likely to engage in this homosocial reproduction than women and minorities. Both points are important polit- ically, as well as sociologically, because they speak to how women and minorities behave when they rise to positions of power, and whether, once in these positions, their behavior is likely to open doors to minority and women's power attainment in the future.

Together these additional considerations raise the possibility that how members of different race and gender groups advance up workplace power hierarchies depends not just on their own race and gender, but also on the race and gen- der of those overseeing the power positions in question-ascriptive similarity with superiors might not be an additive factor but one that conditions how other factors operate. This pos- sibility suggests that a key difference between white men and other groups is that white men, by virtue of being dominant, nearly always rise to power under "similar others," whereas women and minorities generally take two tracks: they advance under white men, or they advance under similar others. Which track women and minori- ties take, in turn, might influence which factor (network assistance or human capital) is likely to be most important for advancement. Extrapolating from white men's experiences under similar others, we might reasonably expect network assistance to be more important for advancement among women and minorities who work under ascriptively similar superiors than those who work under ascriptively dis- similar (e.g., white-male) superiors; whereas, the opposite would be true of education and expe- rience--it would be less important for advance- ment under ascriptively similar superiors than under ascriptively dissimilar superiors. Additional consideration of these potential con- ditional effects of relative superior ascription motivates our final hypothesis:

Hypothesis 5: Women and minorities under self-similar superiors rely more on network assistance and less on human capital to attain higher positions of power than women and minorities under self-dissimilar superiors.


Data for our study come from the Multi-City Survey of Urban Inequality (MCSUI), which is a multistage, stratified, area-probability sample of white and minority respondents in Atlanta, Boston, and Los Angeles conducted during 1992-1994, a time of local and national eco- nomic expansion. The survey was administered through face-to-face interviews that lasted approximately two hours. Race and ethnicity of respondents and interviewers were matched to minimize well-known race-of-interviewer effects (see Johnson, Oliver, and Bobo 1994).3

For our purposes, the advantages of the MCSUI are fourfold. First, its multiethnic sam- ple allows us to examine Latino men and women in addition to whites and blacks-this is a nov- elty in quantitative research on workplace power in the United state^.^ Second, the MCSUI pro- vides data on multiple, successive levels of workplace power and on the race and sex of immediate superiors, in addition to data about human capital and employment context. This information allows us to test for evidence of increasing inequality at higher levels of power (see Table 1) and the extent to which this evi- dence might differ according to the relative and absolute characteristics of superiors involved. Third, the MCSUI provides data about how workers acquired their jobs (e.g., through for- mal searches or with network assistance), which allows us to test assumptions about the impor- tance of such assistance for advancement direct- ly. Finally, the MCSUI draws from a diverse set of metro economies, which, while perhaps not representative of the U.S. labor force as a whole, collectively draws from a wide range of labor market processes and contexts that can influence the distribution of workplace power.

While these features make the MCSUI the best large-scale dataset available for the kind of analyses we wish to conduct, it is not without weaknesses. First, our indicator of workplace power is basic; we measure broad differences

The MCSUI also includes data from Detroit, which we omit because of a lack of information on key labor market variables (e.g., level of workplace power). For limitations of using cross-sectional data to study increasing inequality in workplace power attainment, see Baxter and Wright (2000).

We refer to men of Hispanic descent as "Latino" and women of Hispanic descent as "Latina." The vast majority of these groups in our sample comes from Central America and the Caribbean. Within these subsamples, Mexicans comprise the largest group (39 percent), with most residing in Los Angeles. Puerto Ricans and Dominicans comprise the next largest groups (20 percent and 13percent respec- tively), with most residing in Boston. Sample limi- tations preclude us from analyzing these specific groups directly.

among workers, supervisors, and managers. While these differences are important, they do not allow us to identify individuals' detailed positions on the corporate latter, as prior research on gender inequality has sought to do (e.g., Baxter and Wright 2000). Second, infor- mation on network assistance is self-reported and may reflect group differences in the likeli- hood of such reporting, in addition to hfferences in actual networking effectiveness. Finally, there is no way to determine entirely whether evidence of homosocial reproduction is a product of in- group preference or the result of sex and race segregation that effectively limits candidates to in-group members only.

With these limitations in mind, we select only civilian labor force participants between the ages of 21 and 64 who are not self-employed. We focus on non-self-employed workers because we are interested in power distribution among paid employees, not among individuals who made a qualitative shift from employee to employer. Second, in tests of "homosocial repro- duction" and other allocation processes speci- fied in Hypotheses 2-5, we restrict our sample to workers who report having immediate super- visors, which means that our results cannot be generalized to the very tops of organizational hierarchies where power holders have no supe- riors. We believe that this focus is justified by the fact that such mid-level positions comprise the overwhelming majority of workplace power positions in the United States and represent strata where competition for legitimate author- ity among individuals of different races, eth- nicities, and genders is likely to be most ~ommon.~


In this section we discuss our operational def- initions for key variables, starting with the dependent variable: legitimate authority (i.e., workplace power). All variables are listed in Table 2 with subsample means and standard deviations.

Exclusion of respondents with no immediate superior reduced our sample by 9.3 percent. Further investigation indicates that excluded respondents tended, on average, to exhibit the same odds of power attainment as respondents in our sample.

socioeconomic status, compensation, job com- plexity, education, experience, and ascriptive job context (see Table A1 on the ASR Web site supplement, http://www.asanet,org/ journals/asr/2004/toc039.html).These differ- ences combine with prior research using sim- ilar questions to support the empirical validity of our parsimonious measure of workplace power.

For Hypotheses 2 and 5, we operationalize network assistance as a dummy variable based on the following question: "Did you find your job through friends or relatives, other people, newspaper ads, or some other way?" If the respondent reported using a personal contact, the interviewer collected information about the mode of assistance. Using this information, we define network assistance conservatively as cases in which a job contact talked to the employer on the respondent's behalf, provid- ed a reference, or hired the respondent. We exclude contacts who merely passed along information about the job because this mode of network assistance is considered secondary to workplace power distribution, which empha- sizes processes of sponsorship over mere information flow. Because the MCSUI asked about the use of job contacts only among respondents who reported actively searching for jobs, we also define workers who entered new jobs without an active search as receiving network assistance. The logic here is that, in these cases, job networks brought the employ- er to the respondent, rather than vice versa, and that both scenarios constitute a "strong" form of network assistance (see Granovetter 1995).

For Hypotheses 1, 3, and 5 we operational- ize four indicators of human capital. We meas- ure education as the total number of years of formal schooling. We also include three indi- cators of labor force experience. We measure total work experience as the number of years that a respondent was employed formally since first leaving full-time school. We measure prior job-specific experience as a simple dummy indicator (0 =no; 1 =yes) based on the question, "Did you have any previous experi- ence in this type of job, excluding schooling, before you were hired?" Finally, we measure organizational tenure as the number of years that the respondent reports being employed with his or her respective employer.

For Hypotheses 4 and 5 we operationalize a dummy indicator that is set to 1 if the respon- dent works under an ascriptively similar supe- rior-that is someone of the same racelethnicity and sex-and 0 if the respondent works under an ascriptively dissimilar superi- or. This indicator is based on three nested ques- tions. "Do you have an immediate supervisor on your job to whom you are directly respon- sible?" "What is your immediate supervisor's race or ethnic origin?" "Is your immediate supervisor a man or a woman?" We interpret a value of 1 for this variable as indicating a pat- tern consistent with homosocial reproduction. Our reasoning is that even if a respondent's immediate superior did not have sole respon- sibility for filling the respondent's current position, he or she almost certainly provided meaningful input. In supplemental analyses, we also use information on the race and ethnici- ty of coworkers, which we discuss later in this article.

To isolate hypothesized relationships, we include several job-relevant factors identified in prior research as being important covariates of workplace power. One such factor is establishment size, which reflects the vertical and horizontal complexity of the organization in question and the number of power positions likely to be available to respective employees. We operationalize this factor as the natural log of the number of employees that the respon- dent reports working at his or her establish- ment. We also include a dummy indicator for public sector (0 =private sector; 1 =public sec- tor) because prior research indicates that the relative disadvantage that women and minori- ties face in advancing up workplace power hierarchies tends to be lower in public than pri- vate settings, owing to more egalitarian hiring practices and bureaucratic protocols for advancement in the former (Fernandez 1975;

Wilson 1997).

Another factor related to workplace power is time spent at work. We operationalize this variable as the natural log of the average num- ber of hours worked per week. We log this variable to compress higher values because work hours that extend beyond normal full- time status are more likely to be the result of being a manager than a determinant of becom- ing a manager. We also include a four-catego- ry indicator of occupational location. This indicator is based on 1990 Census Occupation Codes and includes the following categories:

professional and technical occupations, which include officially titled managers and supervisors; (2) craft and repair occupations;

service occupations; and (4) clerical and sales occupations (reference ~ategory).~ Consistent with prior research on power attainment (Baxter and Wright 2000; Rosenfeld, van Buren, and Kalleberg 1998; Wright, Baxter, and Birkelund 1995), we use this crude indicator of occupational location to minimize problems of circularity that would result if we used more refined categories or a single, continuous measure of occupational status, such as the socioeconomic index. The problem with these more refined measures of occupational location is that they would be too closely tied conceptually and empirically to our dependent variable of workplace power to be included as compositional controls.

In supplemental analyses, we also include two indicators of family status: currently mar- ried (0 =no; l =yes) and children in the house- hold (0 = no; 1 = yes). Researchers often assume that marriage and parenthood correlate negatively with power attainment among women, who have historically been more like- ly than men to sacrifice employment mobili- ty for domestic responsibilities.

Concern that occupational location is an endogenous variable is minimal here because we use broad categories and because most workplace power is achieved outside officially recognized "manager" and "supervisor" occupations. Still, we reestimated all our models using a 10-category industrial typol- ogy in place of our occupation controls. Results were nearly identical in both cases, except estimates of the effects of educational attainment are slightly lower in models with occupation rather than industry con- trols. Thus, our estimates of educational attainment with occupation controls provide a comparatively conservative estimate of this variable's effect on authority attainment.





Hypothesis 1 states that women and minorities find it increasingly difficult to advance, relative to white men, at higher levels of workplace power. To test this hypothesis we estimate sev- eral, nested multinomial regression equations that predict employment at successive levels of workplace power, focusing specifically on changes between worker-versus-supervisor and supervisor-versus-manager comparisons. We interpret increasing inequality, or disadvantage, specified by Hypothesis 1 as one in which a group's manager-versus-supervisor coefficient is negative, statistically significant, and larger in magnitude than its supervisor-versus-work- er coefficient. This pattern would imply that the group in question finds it more difficult, rel- ative to white men, to advance from supervisor to manager than from worker to supervisor. Results from these analyses appear in Table 3.

Model 1 estimates the "gross gap" in author- ity for each group, with no statistical controls. Comparisons of coefficients in rows la and 1 b offer initial support for the increasing-inequal- ity hypothesis for every group except white women at the .05-level (two-tailed test); a one- tailed test, however, would include white women. As an interpretative example, consid- er black men. Results from Model 1 indicate no statistical difference between black men and white men with respect to being a supervisor versus a worker (exp[-0.027] =0.97). However, the anti-log of -0.792 indicates that black men are only 0.45 times, or about half, as likely as white men to be managers as supervisors. Thus, without statistical controls, we conclude that a pattern of increasing inequality exists for black men, relative to white men, based on the judg- ment that 0.45 differs significantly from 0.97.

A more rigorous test for the presence of increasing inequality includes statistical controls for nondiscriminatory factors associated with workplace power. For this test we fit two addi- tional models. In Model 2, we add human-cap- ital factors (years of education, total work experience, prior job-specific experience, and employer tenure). Comparing results across Models 1 and 2 indicates that these factors explain most of the increasing-inequality effect

Table 3. Multinomial Regression Coefficients for Gross and Net Gaps in Power Attainment

Levels of Power Men Black, B 1

Model 1. Gross Gapa

Supervisor versus worker
Manager versus supervisor

Model X2 = 132.4 (10df) Model 2. Net Gap with Controls for Human Capitalb

Supervisor versus worker
Manager versus supervisor
Model x2= 278.6 (18df)
x2test of model 2 versus model 1 = 146.2 (8df)***

Model 3. Net Gap with Controls for Human Capital and Employment Contexf

a. Supervisor versus worker ,119 (.181)
b. Manager versus supervisor -.311 (.262)
Model x2: 448.3 (30 df) x2test of model 3 versus model 2 = 169.7 (12df)***    

Women Latino, B2 White. B3 Black. B4 Latina. B5

.I21 (.195) -.026 (.180) -.042 (.164) -.371 (.208)

.I45 (.262) -. 178 (.23 1) -.843*** (.238) -.436 (.324) c "g


Note: Data shown for specific group versus white men. Standard errors appear in parentheses. N = 3,480. E
a log[Pr(LevelJ/F'r(Leve1,1)1= a +bi[race-sexi]; bi reported (white men as reference group). !as
log[Pr(Le~el~)/Pr(Level~-~)]= a + b,[race-sex,] +b2[years of education] + b3[total work experience] + b4[prior jobspecific experience] + b5[years with employer]; bi reported 2 u
(white men as reference group). log[Pr(Level,J/Pr(Level,,)] = a +bi[race-sexi] +b2[years of education] + b3[total work experience] + b4[prior jobspecific experience] + b5[years with employer] + ~,b,[employrnent contextj]; indicators of employment context include number of workers in establishment (logged), public sector (0: l), hours worked per week (logged), and occupational location (professional/technical, crawrepair, service, clericaUsales [ref.]); bi reported (white men as reference group). *p < .05; ** p < .01; ***p < ,001 (two-tailed test) 48 5 fi 2

among Latinos and white women, but not among black men, black women, and Latinas (i.e., coefficients for these groups in Row 2b are still negative, statistically significant, and larg- er than coefficients in Row 2a). Next, in addi- tion to these human-capital factors, we add controls for employment context in Model 3 (establishment size, publiclprivate sector, occu- pational location, and hours worked per week). Comparing results across Models 2 and 3 indi- cates that these factors explain most of the increasing-inequality effect among black men and Latinas, but not among black women.

These findings indicate that although each major race-sex group exhibits a pattern of increasing inequality, relative to white men, only black women exhibit this pattern after con- trolling for variation in human capital and employment context, suggesting that they suf- fer more than other groups from direct dis- crimination. Specific calculations from Model 3 indicate that, net of the full set of controls, black women are just as likely as white men to be supervisors as workers (exp[-O.042] =0.96), but they are only 0.43 times as likely as white men to be managers as supervisors (exp[-O.843]). The implication for the remain- ing groups (black men, Latinos, white women, and Latinas) is not that they are free from increasing inequality (relative to white men). Instead, the implication is that this inequality is more attributable to indirect processes affecting human capital attainment and assignment to different employment contexts than it is to direct discrimination. It is worth noting, however, that in Model 3, coefficients for all groups (except Latinos) are in the hypothesized direction.

A potential criticism of these results is that women voluntarily make themselves less avail- able for promotion to save time and energy for fulfilling traditional wife and motherhood roles at home (i.e., increasing inequality for women is attributable to self-removal from higher lev- els of power, not discrimination). To explore this self-removal issue, we estimated a fourth multinomial regression equation that included the full set of controls present in Model 3 plus main-effect and group-specific interaction terms for marriage (yeslno) and presence of children in the household (yeslno). If self-removal is operating, we would expect women's relative gap in power attainment to be larger in com- parisons among married parents than in com- parisons among single nonparents. Results of our supplemental analysis lend little support to this expectation (see Table A2 on the ASR Web site supplement, http://www.asanet.org/ journalslasr/2004/tocO39.html):a chi-squared test indicates no significant improvement in model fit over Model 3; none of the respective family-status interaction terms are statistically significant at the .05-level; an4 appropriate cal- culations reveal that the strongest evidence of increasing inequality among black women, rel- ative to white men, occurs in comparisons among single nonparents (the family status with the least traditional self-removal pressures), not nonrnarried parents (the family status with the most traditional self-removal pressures).

These findings affirm support for our con- clusions regarding black women from Table 2. Moreover, they are consistent with the conclu- sion by Wright et al. (1995) that family status accounts for little of the observed gender gaps in workplace power in the United States. These findings are also consistent with those of Cassirer and Reskin (2000), who found that, net of job-relevant factors, men and women have equal aspirations of promotion, regard- less of family status.

To test Hypotheses 24 we take an "interaction approach." The logic behind this approach runs as follows: To test if specific factors operate dif- ferently for white men than other groups, we first establish how these factors operate for white men by estimating a "main effects" model separately for them. Next, to test the extent to which specific factors vary in their effects between white men and each of the "other" groups, we pool each separate "other" group with white men and estimate a model with appropriate interaction terms. When a coeffi- cient for an interaction term is determined to be statistically significant at the .05 level, we con- clude that the interaction effect under review is statistically different from zero; otherwise, we conclude that the interaction effect occurred by chance. In these analyses, negative and statis- tically significant coefficients at higher levels of workplace power for indicators of network assistance and self-similar superiors would affirm Hypotheses 2 and 4, respectively. By contrast, positive and statistically significant coefficients for human-capital factors at high- er levels of power would affirm Hypothesis 3.

For these and remaining regression analy- ses, we restrict our sample to respondents who changed jobs within five years of the survey, either within the same organization or through a change in employer. We impose this restric- tion because the MCSUI collected data about network assistance only from recent job chang- ers in order to maximize measurement reliabil- ity-a common practice in studies of job networking (see Granovetter 1995). As a result of this restriction, we minimize differences in power attainment that linger from past person- nel practices and maximize differences result- ing from contemporary practices, that is, behavior most likely to be still in operation today.

Results of these regression analyses appear in Table 4. As a point of comparison, we start with the baseline equation for white men. Here results indicate that, net of background factors, the chief variable distinguishing supervisors from workers is employer tenure. Specifically, results imply that for every additional year with an employer, a white man's odds of moving from worker to supervisor increase an average of 5 percent (exp[0.054]). This factor, howev- er, exerts little additional effect on moving from supervisor to manager. Instead the key factor here appears to involve ascriptive similarity with higher-level superiors. Specifically, results indicate that white men are twice as likely to advance from supervisors to managers when these managerial positions are overseen by white men than when they are overseen by ascrip- tively dissimilar superiors (exp[0.706] = 2.03). This finding suggests that, as white men move up workplace power hierarchies, they benefit increasingly from practices of homosocial repro- duction. By contrast, the nonsignificant coeffi- cients for network assistance do not imply that such assistance is unimportant, but rather that it is equally common among white men at all levels of the power hierarchy-a constant can- not explain a variable outcome such as work- place power. Similar nonsignificant findings result for education and experience.

The remaining results in Table 4 test whether these processes of advancement identified for white men differ significantly from those expe- rienced by the other groups in our study. For


these tests, we report coefficients for interaction terms from the respective pooled-equations esti- mated with white men. For example, in the pooled equation for black men, the coefficient

0.001 (p > .05) for "years with employer" refers to the interaction term "years with employer x black man" (with white men as the comparison group). The fact that this coefficient is statisti- cally nonsignificant at the 0.05 level, net of other factors, implies that there is no statistical difference between black men and white men along this dunension of authority attainment: net of other factors, both groups rely approximate- ly equally on organizational tenure to advance from worker to supervisor. Similar conclusions obtain for all other measures in the model. Notably, the statistically nonsignificant coeffi- cients for self-similar superiors suggest that black men and white men rely approximately equally on processes of in-group favoritism to advance from supervisor to manager status (-0.769;~= .22).' In general, results for black men in Table 4 imply that they move up work- place power hierarchies much the same way that white men do: using organizational tenure to advance from worker to supervisor, and using in-group favoritism to advance from supervisor to manager. These findings offer no support for Hypotheses 2-4 and, instead, imply a set of "separate but parallel" processes of authority attainment for black men, relative to white men.

Results for Latinos reveal much the same pattern, with one exception. Organizational tenure plays an even stronger role among Latinos in advancing from worker to supervi- sor than it does among white men. Appropriate calculations from the full set of coefficients (not shown) indicate that, whereas white men receive a 5-percent bonus for each additional year of organizational tenure, Latinos receive an 18-percent bonus. No other factors differ sig-

For example, to estimate the effect of having a self-similar superior among black men, we would sum the coefficient for having a self-similar superior with the coefficient for being a black man (as opposed to a white man) with the coefficient for the interaction of these two factors. This calculation yields a value of -.465 (.757 + -.457 + -.765). This value compares to a value of ,757 for white men. The interaction coefficient of -.765 (SE of ,623) indicates that this difference is statistically insignificant at the .05 level.


Differences from White Men


Differences among White Men Black Men Latinos m


Supervisor vs. Manager vs. Supervisor vs. Manager vs. Supervisor vs. Manager vs. Key Variables and Model Statistics Worker Supervisor Worker Supervisor Worker Supervisor CA

0 Network assistance (0: 1) -.439 (.282) ,381 (.352) .715 (.419) -.632 (.578) .905 (.478) -1.029 (.610) Years of education -.OOl (.061) ,080 (.079) .I31 (.086) .lo8 (.123) .039 (.070) -.039 (.093) Total work experience ,001 (.015) .006 (.ON) .013 (.086) -.001 (.030) -.038 (.023) .020 (.030) Prior job-specific experience (0: 1) .467 (.282) -.I97 (.347) -.023 (.402) -.431 (.538) -.I92 (.408) .600 (.529)

8 Years with employer .054* (.023) ,022 (.025) .00l (.033) -.013 (.039) 113* (.040) -.044 (.051) Ascriptively similar superior (0: 1) -.I74 (.283) .706* (.345) ,232 (.431) -.765 (.623) -.264 (.436) -.852 (.601) Constant -5.608* (2.195) -7.300* (3.128) -3.621* (1.631) -6.247* (2.461) 4.362* (1.744) -8.200* (2.561) Model x2 (df) 93.3 (24) 154.1 (38) 174.9 (38) 3 N 442 829 888

Differences from White Men

White Women Black Women Latinas

Supervisor vs. Manager vs. Supervisor vs. Manager vs. Supervisor vs. Manager vs. Worker Supervisor Worker Supervisor Worker Supervisor

Years of education ,028 (.082) ,057 (.115) .130 (.077) -.013 (.123) .180* (.076) -.061 (.116)
Total work experience -.006 (.022) ,001 (.029) .003 (.019) -.035 (.030) .029 (.022) -.067 (.038)
Prior job-specific experience (0: 1) ,075 (.410) -.003 (.522) .060 (.352) ,045 (.518) -.521 (.443) -.080 (.700)
Years with employer ,026 (.036) -.028 (.041) -.039 (.028) -.021 (.038) -.002 (.042) ,060 (.059)
Ascriptively similar superior (0: 1) ,447 (.396) -1.144* (.521) .I12 (.376) -.586 (.580) N/A N/ A
Constant -3.610* (1.514) -9.246* (2.361) 4.741* (1.559) -8.519* (2.581) -7.507* (1.847) -5.814* (2.714)
Model X2 (df) 174.0 (38) 257.0 (38) 227.5 (36)
N 916 1.186 894

Note: Data shown with standard errors in parentheses. Indicators of employment context include number of workers in the respondent's establishment (logged), public sector (0:1),
hours worked per week (logged), and occupational location @rofessionaVtechnica1, crafthepair, service, clericaUsales [ref.]). Samples include only workers entering new jobs
within five years of the survey.
N/A = not applicable; too few Latina supervisors (n = 7) and managers (n = 0) report having ascriptively similar superiors with which to compare.
*p < .05, two-tailed test.

nificantly from white men, including reliance on

ascriptively similar superiors to advance from

supervisor to manager status. Thus, as with

black men, we find little support for Hypotheses

2-4 among Latinos.

Results for women depict different scenarios. For white women, findings indicate that advancement from worker to supervisor occurs much the same way as it does for white men. However, advancement from supervisor to man- ager occurs much less often under ascriptively similar superiors. Calculations from the full set of coefficients (not shown) indicate that white women are three times more likely than white men to break into managerial positions under ascriptively dissimilar superiors, 86 percent of whom, in our sample, are white men. This high prevalence of out-group status, relative to (most- ly white-male) superiors, may help to explain why, in the face of the weakest statistical evi- dence for increasing inequality of any group in our study, many observers still insist that such inequality exists for white women: if white women are increasingly likely to work under white men as they advance up workplace power hierarchies, there may be both real and per- ceived obstacles to fiuther advancement to unsu- pervised, top-level positions. These results affirm Hypothesis 4 for white women but not Hypotheses 2 and 3.

Results for Latinas are similar to those for white women, except ascriptive similarity with superiors appears even less effective, relative to white men, for assuming higher positions of power. In fact, the likelihood of Latina super- visors and managers having ascriptively simi- lar superiors is so low that the effects of this variable on power attainment cannot be reli- ably calculated. (In our sample, 0 of the 20 Latinas with manager status report an ascrip- tively similar superior, and only 7 of the 51 Latinas with supervisory status report an ascriptively similar superior.) The implication is that Latinas almost always break into power posi- tions under dissimilar superiors, which likely limits their odds of further advancement, if practices of homosocial reproduction are oper- ating. As with white women, these results affirm Hypothesis 4 but not Hypotheses 2 and 3.

The final and perhaps most surprising set of results occur among black women. Contrary to Hypothesis2, results indicate that network assis- tance is increasingly effective among black

women, relative to white men, for moving into higher positions of power. Appropriate calcu- lations from the full set of coefficients (not shown) indicate that, whereas white men rely on network assistance almost equally at all levels of workplace power, the odds of black women advancing from workers to supervisors increase 39 percent when they receive network assis- tance, and the odds of black women advancing from supervisors to managers increase 500 per- cent when they receive network assistance. These findings suggest that instrumental net- work assistance can be an important response to discrimination, rather than simply an indirect cause. Further investigation of the data reveal that black women most often rely on black men to assist them in attaining managerial positions. Three-quarters of the time these men are fnends or relatives and nearly two-thirds of the time they also work for the employer in question. In our sample, such assistance occurs most com- monly among registered nurses, sales repre- sentatives, and secretaries in predominantly black work settings.

Overall, then, results in Table 4 offer no sup- port for Hypotheses 2 and 3 and occasional support for Hypothesis 4, regarding patterns of homosocial reproduction. Notably, this support for Hypothesis 4 is countered by the finding that, net of other factors, black men, Latinos, and black women receive roughly the same relative benefit from homosocial reproduction in advancing up the workplace power hierarchy as white men.

A potential criticism of our test of homosocial reproduction (Hypothesis 4) in Table 4 is that it relies on a measure of ascriptive similarity with superiors rather than a measure of ascrip- tive similarity with coworkers.*This criticism builds on an alternative interpretation of homosocial reproduction that understands power holders as reserving power positions alongside, rather than under, themselves for in-group members. To test this alternative inter- pretation, we constructed a dummy indicator for

We thank an anonymous reviewer for raising this point.

working with, rather than under, ascriptively similar coworkers. We constructed this indica- tor from two sources of information. First, we identified the racial majority, if one existed, of coworkers from the MCSUI question, "What is the race and ethnicity of most of the employ-

ees doing the kind of work you do at the place where you work?" Next, lacking similar infor- mation about the gender of coworkers, we used metropolitan-level data from the 1990 5% Public Use Micro Samples (PUMS) to identi- fy the locally dominant sex for each of the roughly 500 detailed Census Occupation Codes in each metropolitan area. We then appended this information to the MCSUI and created an alternative indicator of homosocial reproduc- tion that takes a value of 1 if a respondent works with mostly coethnic coworkers in an occupation that, locally, consists of mostly same-sex workers; otherwise, the indicator takes a value of 0. We then reestimated the equations reported in Table 4, substituting this horizontal indicator of homosocial reproduction for our original vertical indicator of having an ascriptively similar superior.

Results of this supplemental analysis (see Table A3 on the ASR Web site supplement, http://www.asanet.org/journals/asr/2004/tocO3 . html) indicate only one substantive change from findings reported in Table 4. Importantly, this change is that white men do not benefit from patterns of (horizontal) homosocial repro- duction as they move up workplace power hier- archies, net of other factors; this finding is consistent with similar race-only analyses in prior research (Smith and Elliott 2002). Relatedly, white and Latina women no longer differ significantly from white men along this (horizontal) dimension of homosocial repro- duction. One implication of these findings is that when generalizing about power attainment across a wide array of work settings, homoso- cial reproduction is perhaps better conceptual- ized in terms of ascriptively similar superiors regulating access to power positions beneath them, rather than in terms of ascriptively similar superiors regulating access to power positions alongside them. The opposite conceptualization might be more valid empirically when gener- alizing about top-level positions in large cor- porations of the type Kanter (1977) studied.

Another consideration with respect to homosocial reproduction is that it actually derives from the product of two distinct rates:

(1) the relative opportunity to practice homoso- cial reproduction (i.e., how often group mem- bers are in positions to fill power positions beneath themselves); and (2) the rate of homosocial reproduction among group mem- bers given the opportunity (i.e., how often group members select in-group members to fill the power positions they oversee). To examine these two rates and their product, we use our origi- nal (vertical) indicator of homosocial repro- duction to examine how often different levels of power are overseen by respective groups and how often these groups appear to select other in-group members to fill positions of power immediately below them. Results appear in Table 5.

Column 1 of Table 5 provides information about the relative opportunity to practice homosocial reproduction. Unsurprisingly, results indicate that white men have the great- est opportunity to practice homosocial repro- duction, and this opportunity increases at higher levels of power. For example, results show that 59 percent of manager positions in our sample are overseen by white men compared with only 4 1 percent of supervisor positions and 39 per- cent of worker positions. Column 2 shows that after controlling for this opportunity structure, women and minorities actually appear to prac- tice homosocial reproduction in positions of power at higher rates than white men. For example, results indicate that 28 percent of white-male superiors select other white men to fill supervisor positions immediately below them. This rate of homosocial reproduction contrasts with the rate of 36 percent among white women; approximately 50 percent among black men, Latinos, and Latinas; and 65 percent among black women. Rates of homosocial reproduction in manager positions converge by comparison, indicating that, after controlling for relative opportunity, all groups are rough- ly the same in their tendency to select similar people to fill manager positions immediately beneath themselves.

Overall, these findings help to refine our understanding of homosocial reproduction. Table 5 indicates that patterns consistent with homosocial reproduction are common among all race-sex groups. Yet, only white men have sufficient opportunity to engage in these prac- tices with relative frequency, and this frequency

Table 5. Opportunity and Rate of Homosocial Reproduction Decomposed by Group and Level of Power

Gender and Race of Positions that Superiors Fill Superiors that Oversee Positions that Superiors Fill with In- through Homosocial Respondents' Work Respondentsa (%) Group Membersb (%) ReproductionC(%)

Workers (n = 2,480)

Men White 39.4 22.9 (n = 978) 9.2 Black 8.9 38.9 (n = 221) 3.5

Latino 12.5 47.3 (n = 309) 5.9

Women White 24.0 30.3 (n = 596) 7.3 Black 10.6 71.9 (n = 263) 7.6 Latina 4.6 61.1 (n= 113) 2.8

Total 100.0 -36.3 Supervisors (n = 470)

Men White 41.3 28.4 (n = 194) 11.7 Black 9.4 47.7 (n=44) 4.5 Latino 9.6 46.7 (n = 45) 4.5

Women White 26.4 36.3 (n = 124) 9.6 Black 10.4 65.3 (n = 49) 6.8 Latina 3.0 50.0 (n = 14) 1.5

Total 100.1 -38.6 Managers (n = 3 15)

Men White 59.3 45.5 (n = 188) 27.0 Black 8.2 38.5 (n = 26) 3.2 Latino 6.2 57.9 (n = 19) 3.6

Women White 18.9 45.0 (n = 60) 8.5 Black 7.3 50.0 (n = 22) 3.7 Latina 0.0 NIA (n = 0) 0.0

Total 99.9 -46.0

Note: Sample includes all eligible respondents, not just workers entering new jobs within five years of the survey.

a Relative opportunity for homosocial reproduction. Rate of homosocial reproduction given the opportunity. The percent is an estimation. Data in this column show aggregate rate of homosocial reproduction.

increases with upward movement within work- archies is conditioned by whether such place power hierarchies. In other words, in- advancement occurs under ascriptively similar group favoritism may be universal, but or dissimilar superiors. The underlying idea is opportunities to practice it are not. As a result, that homosocial reproduction influences the column 3 of Table 5 indicates that 27 percent relative importance of network assistance and of all manager positions in our sample were human capital in moving up organizational filled in a manner consistent with white-male chains of command. To test this hypothesis homosocial reproduction-over three times we estimate a multinomial regression equa- the rate for white women and over eight times tion to predict the likelihood of employment at the rate for respective minority groups. successive levels of workplace power for each group of non-white-men. The independent variables are the same as those in Model 3 of



Table 3, with the addition of interaction terms



for having an ascriptively similar superior Hypothesis 5 asserts that how women and (yeslno) by network assistance and the four minorities advance up workplace power hier- indicators of human capital. If Hypothesis 5 is correct, we would expect coefficients for "ascriptively similar superior x network assis- tance" to be positive and statistically signifi- cant, whereas we would expect coefficients for corresponding interaction terms with edu- cation, total work experience, prior job-specific experience, and organizational tenure to be negative and statistically significant. Because Latinas very rarely gain positions of power under ascriptively similar superiors, we do not include them in this analysis.

Results of this test offer little support for Hypothesis 5 (see Table A4 on the ASR Web site supplement, http:llwww.asanet.org/ journalslasr/2004/toc039.html).For black men, Latinos, and black women, none of the coefficients for respective interaction terms reach statistical significance at the .05 level. These nonsignificant findings imply that net- work assistance and human capital are equal- ly predictive of authority attainment under ascriptively similar superiors as under ascrip- tively dissimilar superiors. Results for white women, by contrast, produce several statisti- cally significant coefficients for relevant interaction terms. The finding most consistent with Hypothesis 5 indicates that for white women, organizational tenure matters less for advancing from supervisor to manager under other white women than it does under ascrip- tively dissimilar superiors. The other statisti- cally significant interaction term involves total work experience, but its effects run con- trary to Hypothesis 5. To illustrate, we solve the equation for white women for increasing years of work experience, setting all other factors equal to subsample means for white women. We then plot the estimated odds of employment at successive levels of power in Figure 1.

Results reveal countervailing effects of work experience for white women's advancement. First, the top panel of Figure 1 indicates that white women with relatively little work expe- rience are much more likely to advance from worker to supervisor under ascriptively simi- lar superiors than under ascriptively dissimi- lar superiors. This conditional difference appears to last until white women gain between 15 and 20 years of work experience, at which time the relative odds of advancing from work- er to supervisor become roughly equal (but low) under both types of superiors. By contrast

and contrary to Hypothesis 5, the bottom panel of Figure 1 indicates that work experience matters less for advancing from supervisor to manager status under ascriptively dissimilar superiors than under ascriptively similar supe- riors.

While unexpected, these findings suggest that white women tend to take one of two tracks up workplace power hierarchies: (1) they enter into supervisory positions relative- ly early in their careers under other white women and then, as they accrue experience, slowly increase their odds of advancing from supervisor to manager; or (2) they enter into managerial positions under white men rela- tively early in their careers. Because white men oversee more managerial positions than white women (see Table 5), the second route is currently more common for white women's managerial attainment. These findings sug- gest a very different picture than that implied by Hypothesis 5. Instead of experience being more important for advancement from super- visor to manager status under white men, expe- rience is less important.

One possible explanation for this finding is that white men tend to view experience among white women less in terms of productive capac- ity and more in terms of fading desirability, leading white men to favor younger, less expe- rienced white women over older, more expe- rienced white women for manager positions, all else equal. Another possible explanation is that times have changed in recent years, such that white women entering the labor force now do not have to prove themselves to white-male superiors to the same extent as they had to in the past in order to advance into managerial positions. Either way, results yield little over- all support for Hypothesis 5. The main find- ings imply that human capital and network assistance are equally determinant of advance- ment among black men, black women, and Latinos, regardless of ascriptive similarity with superiors. Among white women, however, ascriptive similarity with superiors improves odds of attaining supervisory status with little work experience, but this benefit does not then open doors to managerial power; youth under white men does.

Supervisor v. Worker Status

1 5 10 15 20 Years of Total Work Experience

Under Ascriptively Similar Superiors Under Ascriptively Dissimilar Superiors

Manager v. Supervisor Status


1 5 10 15 20

Total Years of Work Experience

Under Ascriptively Similar Superiors Under Ascriptively Dissimilar Superiors

Source: Estimated multinomial regression equation predicting successive levels of workplace power among white women (see ASR website, Table A4). Control variables are set at subsample means for white women.

Figure 1. White Women's Estimated Odds of Employment at Successive Levels of Power by Total Years of Work Experience

inequality among recent job entrantslchangers.


In this investigation, we focused specifically This research had two broad goals. First, we on hypothesized differences in human capital wanted to determine if there is empirical evi- and observable, interpersonal mechanisms that dence of increasing inequality in workplace differentiate white men from other groups power for a wider array of women and minori- depending on their particular combination of ties than previously examined in sociological ascriptive traits. One of these mechanisms was research. Second we wished to study the mech- effective network assistance; the other mecha- anisms that help create and sustain this form of nism was superiors' preferences for similar oth- ers. Certainly these are not the only mecha- nisms that determine who attains power, and these mechanisms are surely mediated by orga- nizational policies and behavior that require further investigation.

Still, our findings indicate that, with reason- able statistical confidence, men and women of various races and ethnicities experience increas- ing inequality in workplace power, relative to white men, but they experience it to different degrees and via different mechanisms. Among Latinos and white women, increasing inequal- ity appears to result largely from human-capi- tal deficiencies relative to white men (specifically education among Latinos and work experience among white women). Therefore, one policy recommendation might be to improve human capital among these groups, in hopes that greater similarity in education and experience will bring greater similarity in workplace power attainment. However, there are at least two rea- sons to be skeptical of this plan's success over the long term.

First, white men currently do not need to dis- criminate against Latinos to ensure Latinos' noncompetitiveness because educational dif- ferences yield effectively the same result. If the human-capital difference between white men and Latinos decline, and competition intensifies, white men might close ranks against Latinos through other mechanisms. Second, white women appear to fare worse, not better, under white-male superiors as they gain work expe- rience-the opposite of what we might expect. Because white men oversee the majority of managerial positions in U.S. workplaces, this pattern can offset human-capital improvements among white women. Both scenarios, of course, ultimately depend on organizational practices that convert human capital into equal opportu- nity and suggest that merely equalizing human capital credentials will be insufficient, by itself, to remove patterns of increasing disadvantage in tomorrow's workplaces.

With respect to networking, the strongest empirical results run counter to expectation. Black women, not white men, appear most like- ly to rely on instrumental network assistance to attain positions of power. This pattern could reflect several dynamics. First, networking can serve as an important response, as well as cause, of direct discrimination, as research on immi- grant adaptation and ethnic economies sug- gests. Second, people who face multiple oppres- sions, such as black women due to their race and gender, might be uniquely conscious of net- work assistance when they receive it, making them more likely than other groups, including white men, to report better information on net- work assistance in surveys. Third, because net- working is typically more an intra-organizational process than an extra-organizational process when positions of power are at stake, our use of a random sample of employees across many workplaces might understate the relative impor- tance of network assistance for white mens' authority attainment. A larger random sample within organizations might yield results more consistent with traditional conceptualizations of "old boy" networks, since this type of data would permit better analysis of the differential effectiveness of networks within internal labor pools.

Finally, with respect to preferences for sim- ilar others, there are strong findings to indicate that most superiors, regardless of their race and sex, tend to fill power positions they oversee with ascriptively similar others, that is, they appear to engage in what Kanter called "homosocial reproduction." Findings also show that because there are more white men at high- er levels of workplace power than members of other groups, white men have greater opportu- nity to exercise this self-similar preference an4 in the process, reproduce their advantage over successive generations of employees. What remains to be determined is the extent to which these patterns reflect a priori segregation of races and sexes across establishments and jobs as opposed to "real time" preferences of supe- riors for similar others. In extreme cases, forces of segregation are in operation long before employers make hiring decisions regarding posi- tions of power. This can leave superiors with few nonsimilar candidates from which to choose, rendering their ascriptive preferences moot. In other cases, the circle of eligible candidates might be quite diverse, rendering the prefer- ences of those in charge more salient for under- standing group inequalities in power attainment.

This consideration points to another area where more and better information about intra- organizational dynamics could be useful in determining the specific subprocesses at work in producing increasing ascriptive inequality in higher positions of workplace power. In addition to this new (and costly) data collection strate- gy, future research on this form of inequality might look to comparative ethnographies that examine how different groups perceive and adjust to the unique sets of obstacles they appear to face, relative to white men, as they move up workplace power hierarchies. As we await these insights, we should resist the conclusion that all people who are not white men face the same hurdles to attaining higher levels of workplace power. Evidence here suggests that a one-size- fits-all explanation hides more than it reveals and that more research is still needed to pinpoint the precise mechanisms that convert different combinations of ascriptive characteristics into inequalities in workplace power.

James R Elliott is an Assistant Professor of Sociology at Tulane University, where he conducts research on urban development and ascriptive inequalities in the labor market. In addition to research on workplace powel; he is currently examining the social and eco- nomic organization of immigrant dispersal from gate- way cities to new destinations throughout the US. urban hierarchy.

Ryan A. Smith is an Associate Professor at the School of Public Afairs, City University of New York and a recent Scholar in Residence at the KE.B. DuBois Institute of Harvard University. He has published internationally recognized research on race andgen- der stratification in workplace authority and serves as an organizational change consultant to public and private organizations.


Alessio, John C. and Julie Andrzejewski. 2000. "Comment: Unveiling the Hidden Glass Ceiling: An Analysis of the Cohort Effect Claim." American Sociological Review 65:3 1 1-5.

Baxter, Janeen and Erik 0.Wright. 2000. "The Glass Ceiling Hypothesis: A Comparative Study of the United States, Sweden, and Australia" Gender and Society 14:275-94.

Bendix, Reinhard. 1956. Work and Authority in Industry. New York: Harper and Row. Braverman, Harry. 1974. Labor and Monopoly Capital. New York: Monthly Review Press.

Bridges, William P. and Wayne J. Villemez. 1986. "Informal Hiring and Income in the Labor Market." American Sociological Review 5 1:574-82.

Campbell, Karen E. and Rachel A. Rosenfeld. 1985. "Job Search and Job Mobility: Sex and Race Differences." Research in the Sociology of Work 3:147-74.

Cassirer, Naomi and Barbara Reskin. 2000. "High Hopes: Organizational Position, Employment Experiences, and Women's and Men's Promotion Aspirations." Work and Occupations 27:43863.

Cotter, David A., Joan M. Hermsen, Seth Ovadia, and Reeve Vanneman. 200 1. "The Glass Ceiling Effect." Social Forces 80:655-82.

Dahrendorf, Ralf. 1959. Class and Class Conflict in Industrial Society. Stanford, CA: Stanford University Press.

Federal Glass Ceiling Commission. 1995. Good for Business: Making Full Use of the Nation 5 Human Capital. Washington, DC: U.S. Government Printing Office.

Fernandez, John l? 1975. Black Managers in White Corporations.New York: John Wiley and Sons.

Granovetter, Mark. 1995. "Afterword 1994: Reconsiderations and a New Agenda." Pp.139-82 in Getting a Job: A Study of Contacts and Careers, 2d edition. Chicago, IL: University of Chicago Press.

Halaby, Charles N. 1979. "Job-Specific Sex Differencesin Organizational Reward Attainment: Wage Discrimination vs. Rank Segregation." Social Forces 58: 108-27.

Jacobs, Jerry. 1992. "Women's Entry into Management: Trends in Earnings, Authority and Values among Salaried Managers." Administrative Science Quarterly 37: 282-301.

Johnson, James H., Melvin Oliver, and Lawrence Bobo. 1994. "Unraveling the Paradox of Deepening Urban Inequality: Theoretical Underpinnings and Research Design of a Multi- City Study." Urban Geography 15:77-89.

Kanter, Rosabeth M. 1977. Men and Women of the Corporation.New York: Basic Books.

Kluegel, James. 1979. "The Causes and Cost of Racial Exclusion from Job Authority." American Sociological Review 43:285-301.

McGuire, Gail M. 2002. "Gender, Race, and the Shadow Structure: A Study of Informal Networks and Inequality in a Work Organization." Gender & Society 16:303-22.

McGuire, Gail M. and Barbara F. Reskin. 1993. "Authority Hierarchies at Work: The Impacts of Race and Sex." Gender and Society 7:487-506.

Morgan, Laurie A. 1998. "Glass-Ceiling Effect or Cohort Effect? A Longitudinal Study of the Gender Earnings Gap for Engineers, 1982 to 1989." American Sociological Review 63:479-93. . 2000. "Reply to Alessio and Andrzejewski: Is Engineering Hostile to Women? An Analysis of Data From the 1993 National Survey of College Graduates." American Sociological Review 65:316-21.

Moore, Wilbert. 1962. The Conduct of the Corporation.New York: Random House Vintage.

Momson, Ann M. and Mary Ann Von Glinow. 1990. "Women and Minorities in Management." American Psychologist 45:20&8.

Moss, Philip and Chris Tilly. 200 1. Stories Employers

Tell: Race, Skill, and Hiring in America. New York: Russell Sage.

Mueller, Charles W., Toby L. Parcel, and Kazuko Tanaka. 1989. "Particularism in Authority Outcomes of Black and White Supervisors." Social Science Research 18: 1-20.

Myerson, Allen R. 1997. "As Federal Bias Cases Drop Workers Take Up the Fight." New York Times January 12, Section 1, Page 1, Column 3.

Pfeffer, Jeffrey. 1983. "Organizational Demography." Pp. 299-357 in Research in Organizational Behavior,edited by Lany L. Cummings and Barry

M. Staw. New York: JAI.

Podolny, Joel M. and James N. Baron. 1997. "Resources and Relationships: Social Networks and Mobility in the Workplace." American Sociological Review 62:673-93.

Reskin, Barbara F. 2002. "Rethinking Employment Discrimination and Its Remedies." Pp. 21844 in The New Economic Sociology: Developments in an Emerging Field, edited by Mauro F. Guillen, Randall Collins, Paula England, and Marshall Meyer. New York: Russell Sage Foundation.

Reskin, Barbara F. and Catherine Ross. 1992. "Authority and Earnings Among Managers: The Continuing Significance of Sex." Work and Occupations 19:34245.

Reskin, Barbara F. and Debra B. McBrier. 2000. "Why Not Ascription? Organizations' Employment of Male and Female Managers." American Sociological Review 65:210-33.

Robinson, Robert V. and Jonathan Kelley. 1979. "Class As Conceived by Marx and Dahrendorf." American Sociological Review 41 :209-34.

Rosenfeld, Rachel A., Mark E. Van Buren, and Arne Kalleberg. 1998. "Gender Differences in Supervisory Authority: Variation among Advanced Industrialized Democracies." Social Science Research 27:23-49.

Smith, Ryan A. 1997. "Race, Job Authority, and Income: A Cross-Temporal Study of Changes in the Socioeconomic Status of Black and White Men, 1972-1994." Social Problems 44:701-19.

. 2002. "Race, Gender and Authority in the Workplace: Theory and Research." Annual Review of Sociology 28:509-42.

Smith, Ryan A. and James R. Elliott. 2002. "Does Ethnic Concentration Influence Employees' Access to Authority? An Examination of Contemporary Urban Labor Markets." Social Forces 8 1 :255-79.

Tomaskovic-Devey, Donald. 1993. Gender and Racial Inequality at Work: The Sources and Consequences of Job Segregation. Ithaca, NY: ILR Press.

Tsui, Anne and Charles A. O'Reilly. 1989. "Beyond Simple Demographic Effects: The Importance of Relational Demography in Superior-Subordinate Dyads." Academy of Management Journal 32:402-23.

United States Bureau of Labor Statistics. 1996. News Release: BLS Reports on theAmount of Formal and Informal Training Received by Employees. Washington, D.C.: U.S. Department of Labor.

Weber, Max. [I9141 1968. Economy and Society: An Outline of Interpretative Sociology. Translated and edited by Guenther Roth and Claus Wittich, 3 volumes. New York: Bedminster.

Wilson, George. 1997. "Pathways to Power: Racial Differences in the Determinants of Job Authority." Social Problems 44:38-54.

Wilson, William Julius. 1996. When Work Disappears: The World of the New Urban Poor. New York: Knopf.

Wolf, Wendy C. and Neil Fligstein. 1979. "Sexual Stratification: Differences in Power in the Work Setting." Social Forces 58:94-107.

Wright, Erik O., Janeen Baxter, and Gunn E. Birkelund. 1995. "The Gender Gap in Workplace Authority: A Cross-National Study." American Sociological Review 60:407-35.

Yamagata, Hisashi, Kuang S.Yeh, Shelby Stewman, and Hiroko Dodge. 1997. "Sex Segregation and Glass Ceilings: A Comparative Static Model of Women's Career Opportunities in the Federal Government over a Quarter Century." American Journal of Sociology 103:566432.

  • Recommend Us