Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
A Woman's Place Is Win Other Women Sex Segregation Within Organizations WILLIAM T. BELBY and JAMES N. BARON Sex segregation in the workplace is one of the most visible signs of social inequality. In almost every work setting, it is unusual to see men and women working at the same job. When they do, they typically perform different tasks, with unequal levels of re- sponsibility and authority. Even when job tasks are virtually identical, it is not uncom- mon to find men and women allocated to distinct job classifications within an orga- . . n~zat~on. Even women working full time, year round are paid less than men. While the earnings gap is partly due to unequal access to high wage firms and to unequal pay for compa- rable work, a substantial portion is due to differences in pay scales for job cIassifica- tions filled by men and those Filed by women (Bridges and Berk, 1974; Halaby, 1979; Treiman and Hartmann, 1981~. Sex segre- gation has social-psychological conse- quences as well. For example, groups with limited opportunities for advancement may respond with psychological disengagement from the firm, Towered career aspirations, and an increasingly narrow, instrumental orientation toward work (Kanter, 1977b). In short, sex-segregated workplaces affect 27 us personally. Social structures that gener- ate gender segregation are of great concern to social scientists, and the inequities that segregation engenders are obviously rele- vant to social policy. Yet sociologists know surprisingly little about job segregation by sex. Most of what we have learned concerns segregation among occupations. For exam- ple, we know that equalizing the detailed (census 3-digit) occupational distribution for men versus women would require moving roughly 60 percent of women working out- side the home across occupational cate- gories, and this has changed very little since 1900 (Gross, 1968; Blau and Hendricks, 1978; Williams, 1979; England, 1981a). We are also learning more about how men and women make occupational choices (Bielby, 1978; Marini, 1978, 1980~. Empirical re- search on job segregation across organiza- tional settings, however, is quite sparse. Accordingly, this paper examines sex seg- regation in the workplace, utilizing data de- scribing work arrangements in nearly 400 establishments across a wide range of in- dustrial and institutional settings. We dis- tinguish situations in which employers place men and women in the same job cIassifi-
28 WILLIAM T. BIELBY AND JAMES N. BARON cation from those in which job titles seg- regate the sexes within establishments. Our study could be viewed as a straight- forward job-level disaggregation of findings regarding occupational sex segregation. Those studies acknowledge that considerable seg- regation may exist even within detailed oc- cupational categories. Our measures and methods parallel such studies, and our find- ings confirm their speculations about per- vasive segregation within occupations. However, our aim is not merely to reveal hidden segregation among jobs within firms. Rather, since sex segregation is accom- plished in organizations and is affected by technical, administrative, and social exigen- cies of the workplace, it is important to ex- amine how organizational structures and processes produce sex segregation. Our research does not consider how men's and women's occupational choices, labor force participation, and human capital invest- ments affect the sex composition of the workplace (for contrasting interpretations, see England, 1982, and Polachek, 19791. Nor are we investigating the demand side in the economist's sense of the term, since we have no information on the productivities of dif- ferent classes of workers and the wages em- ployers are willing to offer them (cf. Blau, 19771. Rather, the intersection of labor sup- ply and demand enters into our analysis in- directly, since occupational composition and skill mix of the firm are examined as deter- minants of sex segregation. However, if jobs in most establishments are highly segre- gated by sexeven across firms differing dramatically in their production functions and cost structures then it seems unlikely that marginal adjustments of supply and de- mand account for distinct job assignments of men and women (for a similar view, see Blau and Jusenius, 19761. WHY ARE SOME FIRMS MORE SEGREGATED THAN OTHERS? Diverse explanations of sex segregation have been reviewed thoroughly by others (e.g., Blau and Jusenius, 1976; England, 1981b). Much more has been written, how- ever, about why employers treat men and women differently than about the extent to which they do so. The sparse literature ad- dressing why some firms are more segre- gated than others falls into three categories: institutional accounts, explanations based on tastes for discrimination, and human capital market models. Institutional accounts stress how statisti- cal discrimination in hiring and allocating employees places men and women in dis- tinct career trajectories. Men tend to enter internal labor markets in which they can expect an orderly progression through suc- cessively more attractive jobs, insulated from competition outside the firm. This increases organizational loyalty, decreases costly worker turnover, and allows employers to recoup investments in firm-specific training (Doeringer and Piore, 19711. Women are perceived to have weaker commitments both to specific firms and to paid employment in general and are thus allocated to jobs with low turnover costs and limited opportunity for security and advancement (Bielby and Baron, 19821. Not all firms, however, re- quire specifically trained workers or have internal labor markets. Therefore, if sex biases in allocating workers to job ladders were the only basis for segregating men and women, one would expect less segregation in firms lacking institutionalized employment ar- rangementsparticularly small, labor-in- tensive, entrepreneurial firms in the so-called economic periphery (Averitt, 19681. This is certainly not the only mechanism placing men and women in distinct job cIas- sifications, and perhaps a more reasonable hypothesis is that the process of segregation differs according to an organization's admin- istrative arrangements and location within the economy. For example, small manufac- turing, service, and retail establishments typically rely on an unskilled secondary la- bor market and use simple hierarchy or en- trepreneurial despotism to control workers (Edwards, 19791.
SEX SEGREGATION WITHIN ORGANIZATIONS 29 These enterprises might apply one of sev- eral strategies in allocating men and women to jobs. Highly trained line workers with job- and firm-specific skills typically are not employed in such establishments, nor are highly rationalized personnel and job cIas- sification procedures utilized. Thus, these firms might provide precisely the work con- texts in which men and women who lack credentials for more desirable employment work together within broadly defined job categories. Furthermore, if employers must sacrifice profits in order to discriminate, they must be able to afford the costs of their pol- icies. Marginal firms with weak competitive positions can least afford these costs and have an economic incentive to ignore sex in hiring and allocating workers (Arrow, 19731. In the absence, however, of institution- alized procedures for hiring and allocating workers, male employers in the economic periphery may have more discretion to im- plement tastes for discrimination, which can reflect their own preferences or those of their employees or even their clients. In the most extreme case, patriarchal control strategies would exclude women from the workplace entirely. Such arrangements should be most prevalent in organizational niches that are protected from competitive pressures (e.g., through satellite linkages with larger firms) or where preferences for a segregated work force are so widely held within an industry or area that they have the force of customary law constraining market forces (Doeringer and Piore, 1971:22-271. If labor supply and technical requisites de- ter~ine the distribution of men and women across job categories, then a firm's mix of oc- cupations and skids should largely account for its tendencies to segregate men from women. According to such human capital models, workers expecting intermittent labor force participation (primarily women) choose to en- ter occupations in which job skills do not atro- phy from nonuse (Polachek, 19791. Indeed, if jobs with the highest turnover costs are also those in which skills atrophy most quickly, then extreme segregation can reflect maxi- mizing behavior by both workers and em- ployers. That is, firms wiD assign men and women to the same job titles only under spe- cific, and rare, circumstances: (a) when there is an available labor pool composed of men and women and (b) when employers perceive that the costs of employing men and women roughly are the same. To summarize, certain analysts argue that gender segregation at work is caused by ad- ministrative arrangements for hiring, allocat- ing, and controlling employees. Others em- phasize the impact oftastes or prejudices, while still others claim that sex segregation reflects rational decisions regarding human capital in- vestments on the part of workers and em- ployers. Perhaps because segregation is such a natural attribute of most work situations, little has been written about the conditions under which it does not occur. Our empirical analysis is guided by sev- eral general hypotheses. First, institutional accounts suggest that less segregated firms lack the administrative apparatus to differ- entiate workers by sex and cannot afford the costs of implementing employers' tastes for a segregated work force. Second, neocIass- ical accounts, grounded in notions of tech- nical efficiency, suggest that desegregated organizations do not rely heavily on firm- specifc skills but employ workers in occu- pations that are attractive to both men and women and for which both sexes are eligi- ble. Of course, each of the mechanisms sum- marized above might operate but within specific organizational settings. Conse- quently, we examine the heterogeneity among highly segregated establishments to see if there are alternate strategies by which employers achieve the same result: distinct job assignments for men and women. DATA AND METHODS We analyzed data on work arrangements in hundreds of economic establishments stud- fed in California between 1959 and 1979 by the California Occupational Analysis Field Center ofthe U.S. Employment Service. These
30 WILLIAM T. BIELBY AND [AMES N. BARON data, used primarily in preparing the D'c- t~onary of Occupational Titles (DOI), and procedures for collecting them are described in U. S. Department of Labor (19721. Our unit of analysis is the establishment, the "physical location where business is conducted or where services or industrial operations are per- formed"(U.S. Bureau ofthe Census, 1976:iv). Industrial characteristics serve as contextual Actors, and suborgan~ational information about workers and jobs has been aggregated to char- acterize each enterprise. The majority of the establishments are firms; others are branches, regional divisions, subsidiaries, and produc- tion sites. Since we focused on work sites rather than firms, corporate headquarters of multi- plant organizations are typically not included in our data. Corporations often direct initial desegregation efforts at headquarter mana- gerial and office work (Shaeffer and Lynton, 1979), and progress toward equal employ- ment opportunity (EEO) goals in these areas will not be reflected in our results. The Sample No well-defined sampling frame guides the Employment Service's selection of en- terprises to study, but they try to represent the diversity of activities carried out within any industry (Miller et al., 1980~. The Cal- ifornia Field Center tended to study those industries that are regionally concentrated in the state, so our sample of establishments includes, for example, firms engaged in ag- riculture, aircraft manufacturing, banking, fishing, and motion picture production but not automobile or furniture manufacturing. While our sample provides a reasonable representation of the composition of estab- lishments within industries, the actual in- dustries studied are not fully representative of economic activities in California. Most importantly, manufacturing establishments ~ Reweighting our observations according to pub- lished data on the size distribution of establishments within California's industries has virtually no effect on the distribution of organizational attributes in our sam- ple. are overrepresented in the sample. Major California industries not represented in our sample include construction trades, truck- ing, apparel and general merchandise retail trade (department stores), and insurance carriers. Ike first two industries are male dominated and highly segregated; the latter two employ many women and may be less segregated. While these data do not char- acterize a distinct population, they do re- flect a diversity of work arrangements across a broad range of industrial and organiza- tional contexts. In our view they provide invaluable comparative evidence regarding how administrative, technical, and environ- mental contingencies in organizations affect the structuring of work. The data collected and coded for our proj- ect include 742 observations in over 500 clis- tinct enterprises. About one-fifth of the es- tablishments were visited more than once by Employment Service analysts. The most recent analysis was used for firms with fol- low-up data. Since some of the information used to characterize organizational attri- butes, however, was derived from narrative reports (described below), precedence was given to complete observations that also possessed a contemporaneous narrative re- port, even if a more recent follow-up anal- ysis, lacking a narrative, had occurred. To ensure comparability, analyses re- stricted to the firm's productive component or some other subset of jobs or departments were omitted, since they do not accurately characterize an entire work site. This re- striction reduces the sample of establish- ments to 415. The sex composition of jobs was not reported for 22 of these firms, re- ducing the sample size for analyses reported in this paper to 393.2 Of these, about 26 2 Job composition was not enumerated by sex after 1977-1978, apparently because of increasing resistance from establishments approached by the Employment Service. Unfortunately, this occurred when the Cali- fornia Field Center was studying agricultural estab- lishments; therefore, 7 of the 22 observations lacking information on sex composition are in agriculture.
SEX SEGREGATION WITHIN ORGANIZATIONS 31 percent are plants, franchises, or production sites within some larger entity (though not necessarily corporations); 10 percent are ad- ministrative divisions, regions, or branches of larger companies; 3 percent are subsidi- aries (distinct firms owned by another firm); and 61 percent are independent busi- nesses.3 Over half, 54 percent of these es- tablishments were studied between 1968 and 1971, and 76 percent were visited by the Employment Service between 1965 and 1973. The 393 establishments in our sample em- ploy nearly 47,000 men and over 14,000 women.4 The Documents Lois paper uses two types of data obtained from the records of the California Occupa- tional Analysis Field Center. Staffing Sched- 3 In practice, it sometimes was difficult to determine precisely if establishments studied by the Employment Service were autonomous firms or productive or ad- ministrative units within larger companies. When our materials indicated an owner or president, we assumed the enterprise was autonomous, owner-operated, un- less other information indicated to the contrary. When the top position had such titles as plant superintendent, plant manager, general manager we assumed the en- terprise was a subdivision of a larger firm, unless back- ground information suggested otherwise. Anomalous cases were referred for clarification to the Employment Service analysts who conducted the original studies. Confidentiality restrictions precluded access to estab- lishments' identities, preventing us from resolving such ambiguities directly. 4 The disproportionate share of manufacturing estab- lishments in our sample accounts for the underrepre- sentation of women workers. Nevertheless, the range of industries covered represents nearly every work con- text in which women labor. One important exception: The Employment Service tends to analyze branch plants and to overlook corporate headquarters. Therefore, vir- tually every kind of nonmanual work performed by women is represented in our study, but, unfortunately, we have no instances of such work done at the head- quarter of fices of large corporations. Evidence from the early 1970s suggests that efforts to desegregate non- manual work occurred first in such contexts (Shaeffer and Lynton, 1979~. ules supply, in essence, a complete organi- zational division of labor for the plant or firm in which job titles are analyzed. Face Sheets provide iden~ing information about the es- tablishment and analysis. After assigning the firm to one or more categories of the Standard Industrial Classification (SIC), the Employ- ment Service classifies the enterprise by its primary products and supplies information about any unique or noteworthy characteris- tics of the firm, such as its jobs and processes. Narrative reports prepared for many estab- lishments include information on some or all of the foldowing: history and purpose of es- tablishment; environmental conditions; op- erations and activities (departmentation, workBow, processes or services); personnel policies and practices; recent job restructur- ing, erects of automation on personnel and operations; and optional sections dealing with such topics as the product market and rela- tions with government, the community, or other firms. Operationalization Staking schedules, face sheets, and nar- rative reports were used to measure various environmental, organizational, and techni- cal attributes of establishments as well as the composition of occupations and skills employee! in each enterprise. Operational- izations are summarized in Table 3-1, which also reports descriptive statistics for varia- bles used in our analysis. Organizational scale is measured by the natural logarithm of the number of employees, and positional spe- cialization is measured by the logarithm of median job size. Ike latter measure is com- puted across workers, so a median of 10 in- (licates that one-half of the workers are in establishment job titles with 10 or more in- cumbents as opposed to half the jobs con- taining 10 or more workers. This measure indexes the degree to which establishments "massify" the work force by assigning many workers to the same job title. Consequently, low scores correspond to high levels of spe-
32 WILLIAM T. BIELBY AND JAMES N. BARON TABLE 3-1 (N = 393) Descriptive Statistics and Operationalizations of Organizational Attributes Variable (range) Establishment Scale and Specialization I~og establishment size(.69-8.97) 3.67 Log specialization (~6.49) 1.12 Manufacturing Industry (~1) .63 Standard Mean Deviation N Description Economic Sector Core (0-1) Ambiguous (0-1) Periphery (0-1) Social Organization Log fragmentation ~1.10-1.39) Union or bidding arrangements (~1) Proportion women (.00-1.00) Occupational and Skill Composition Proportion production workers (.0C~1.00) Proportion clerical and sales workers (.0~1.00) Proportion service workers (.00-1.00) Proportion professional, mana- gerial, technical workers (.00-1.00) 1.48 393 Natural log of number of employees in estab- lishment. 1.03 368 Natural log of median job size (see text). 393 One or more of establishment's standard indus- trial classification designations is in the man- ufacturing industry. .38 393 .41 393 .21 393 .14 .25 .32 .56 .20 .08 .16 Average complexity: data 1.92 (0-5.71) Average complexity: people 1.27 (0-8.00) Average technical skills -0.01 (-2.19~.08) Skill specificity (.00-.99) .51 Sex Segregation (12.~100) 93.4 See text. See text. See text. .28 .26 .33 .23 .21 .20 1.09 1.11 1.00 .27 13.8 360 Natural log of ratio of establishment job titles to unique DOT titles. 393 1 = some or all employees unionized or covered by formal bidding arrangements. 393 Proportion of female employees. 380 379 379 379 336 336 379 173 393 Proportion of workers with DOT codes denoting production occupations. Proportion of workers with DOT codes denoting clerical and sales occupations. Proportion of workers with DOT codes denoting service occupations. Proportion of workers with DOT codes denoting professional, managerial, or technical occu- pations. Mean of ratings indicating complexity of work- ers' involvement with data. Mean of ratings indicating complexity of work- ers' involvement with people. Mean of standardized ratings of production workers' technical skill (see text). Proportion of workers in nonentry-level jobs. Index of dissimilarity computed across job titles. 100 = all male, all female, or completely seg- regated. cialization.5 One way to segregate workers is to place each worker in a unique job cat- egory; therefore, we expect greater sex seg- regation in more specialized establish- ments. We Ailed to detect any net differences in relationships between sex segregation 5 This measure correlated - .87 with Gibbs and Pos- ton's (1975) "M4" index of distributive differentiation, describing the evenness of the distribution of workers across positions. Functional (horizontal) differentiation of departments is reflected in our criteria for assigning establishments to economic sectors (see below). Struc- tural differentiation the proliferation of work roles is almost completely determined by organizational scale; the correlation between the number of employees and the number of job titles is .92 when both are measured in a logarithmic metric. Consequently, while such dif- ferentiation may mediate the effects of scale, it is un- likely to affect work arrangements independently.
SEX SEGREGATION WITHIN ORGANIZATIONS 33 across 6 industrial sectors: manufacturing, state, services, social overhead capital, ag- ~ riculture, and trade. We distinguish man- ufacturing from nonmanufacturing, how- ever, to assess indirectly whether physical demands of work account for patterns of sex segregation. We argued above that organizations con- front different incentives to segregate de- pending on their niche within the economy and their size, structure, and technology. These differences in organizational forms and environments capture distinctive locations within the economy what some institu- tionalists and Marxists have called sectors or segments. In certain respects, these are the organizational equivalent of"cIasses"; that is, actors presumed to share certain in- terests and attributes by virtue of their com- mon market positions. Core firms are typ- ically large, differentiated, use automated technologies, produce multiple products, are unionized, and are linked to larger organi- zational entities. Their environments are characterized by interorganizational de- pendence within key industrial sectors, and these establishments tend to be dominant actors in their milieux. In contrast, the eco- nomic periphery is composed of small, un- differentiated enterprises, typically operat- ing in highly competitive markets in industries other than manufacturing. This congruence of organizational form and en- vironment floes not characterize all firms; the 161 establishments allocated to the am- biguous sector include those in small-scale manufacturing and many nonmanufacturing firms in less vulnerable situations with re- gar<1 to their environments. The measures and procedures underlying this sectoral classification scheme are discussed in detail by Baron (1982:Chapter IV). Radical accounts of workplace relations suggest that three aspects of an establish- ment's social organization should be asso- ciated with its level of sex segregation. Frag- mentation is measured by the logged ratio of job titles in the establishment to unique 6-dizit DOT titles assigned by the Employ- ment Service analyst.6 It measures the de- gree to which the organization differentiates its work force administratively beyond what might be expected from a breakdown of de- tailed occupational functions (Braverman, 1974:70-83; Edwards, 19791. Fragmentation is one strategy for segregating male and fe- male workers who perform similar job tasks; that is, separate names are attached to men's and women's work. Other facets of the social organization of workplaces are measured by a dummy variable that denotes the presence of unions or formalized bidcling arrange- ments, covering some or all workers, and the sex ratio, the percentage of workers who are female. Unions and formal bidding ar- rangements are an institutional arrange- ment that may constrain employers' (and male employees') ability to indulge tastes for dis- crimination. Some argue, however, that unions can exacerbate gender inequalities (e.g., Milkman, 1980; Baron and Bielby, 19821. Finally, tokenisma highly unbal- anced sex ratio can facilitate the segre- gation of women or men into one or two separate job titles, while a more balanced work force may be more difficult to segre- gate (Kanter, 1977a). The occupational composition of each en- terprise was computed from the DOT oc- cupational cocles corresponding to each job titIe.7 We measured the distributions of workers across the following broad cate- gories: professional, managerial, and tech- 6 In many firms, not every job title was mapped to a DOT title by the Employment Service analyst. The fragmentation measure was computed only for firms in which both 90 percent of the jobs and 90 percent of the workers were assigned a 6-digit DOT code (see Baron, 1982: Chapter VI). Six-digit DOT codes do not correspond to unique titles in the DOT, but more de- tailed classification is possible only for those jobs as- signed to occupational categories according to the fourth edition classification scheme. 7 Establishments in which less than 75 percent of the workers could be assigned a DOT code were treated as missing data on this outcome.
34 WILLIAM T. BIELBY AND [AMES N. BARON nical; clerical and sales; service domestic, business, personal and production, in- cluding extractive and transportation occu- pations. Levels of informational and inter- personal skill were measured by the mean of ratings indicating the complexity of work- ers' involvements with data and people, re- spectively.8 The measures of technical skills used de- pended on the edition of the DOT used in the analysis of an establishment. Production occupations were classified as skilled, semi- skillecI, or unskilled in the second edition, and we assigned values of 2, 1, and 0 to these categories, respectively. Ratings of workers' involvements with "things" began with the third edition. Average ratings were com- puted for each establishment's labor force using either the measure based on second edition DOT codes or the one based on sub- sequent versions of the DOT. Each estab- lishment's score was then normalized rela- tive to all others sharing the same version of the skill measure. That is, each enter- prise's level of technical complexity is meas- ured relative to other organizations incor- porating the same version of the DOT occupational ciassification.9 Finally, the level of firm-specific skills was assessed indirectly by the proportion of workers in "line" de- 8 This information, when available, was coded from job analyses pertaining to each position in the firm. Otherwise, it was obtained from 6-digit DOT occu- pational codes listed on staffing schedules. Third edi- tion DOT ratings of 7 or 8 for relations with data were recoded to 6 to conform to the fourth edition rating scheme. Mean ratings for each establishment were computed only if: (a) at least half the jobs and workers in line departments could be characterized on the data and people dimensions; (b) at least half of the jobs and workers in other departments could be characterized; and (c) no more than 10 percent of the establishment labor force was missing data on these variables. Finally, scale values were inverted so that large values corre- spond to high levels of involvement with data and peo- ple. 9 The mean level of technical skills was computed from third or fourth edition ratings (from job analyses or DOT codes), subject to the same restrictions de- scribed in footnote 7. Third edition codes of 8 were recoded to 7 to conform to the fourth edition rating partments who were not in entry-level jobs, as indicated on the staffing schedule.~° The index of dissimilarity measures the percentage of workers of one sex that would have to be moved to new job classifications in order to equalize the job distribution of workers by sex (Duncan and Duncan, 19551. It equals zero when the percentage distri- butions of men and women across job cat- egories are identical, and it equals 100 when no men and women work in the same job. ii The Analysis We first describe the distribution of es- tablishments by level of sex segregation. Then we examine the organizational attributes that distinguish propensities to segregate. The relative contribution of social, administra- tive, and technical attributes to patterns of segregation may suggest mechanisms that account for those results, but we do not ex- pect conclusive results from cross-sectional findings. Accordingly, these analyses are supplemented in two ways. First, we ex- amine specific cases for which qualitative information exists on the hiring and alloca- scheme. Mean ratings were computed from second edi- tion codes only if: (a) detailed occupational codes ex- isted for at least 75 percent of the employees in the enterprise and (b) at least 25 percent of the workers were in production jobs. Preference was given to third edition data for establishments that contained jobs an- alyzed in terms of both second and third edition DOT procedures. is Workers in nonproduction departments were ex- cluded because the Employment Service did not al- ways collect information denoting entry-level jobs in those departments. The measure was not computed for establishments with less than 15 percent of their labor force in production-related departments. We also elim- inated observations in which any department had no entry-level workers, or in which certain traditionally entry-level occupations occurred e.g., janitor, re- ceptionist but none was coded as such. Given these restrictions, a measure of skill specificity is available for less than half of our observations. ii We consider all-male and all-female establish- ments to be perfectly segregated and assign a value of 100 to the index of dissimilarity. However, in analyses reported below, these enterprises are considered sep- arately.
SEX SEGREGATION WITHIN ORGANIZATIONS 35 TABLE 3-2 Distribution of Establishments and Workers by Level of Segregation (N = 393) (4) (3) Cumulative (5) (2) Percentage Percentage Median (8) (1) Number of of of Establish- (6) (7) Cumulative Segregation Establish- Establish- Establish- ment Number of Percentage Percentage Index meets meets meets Size Workers of Workers of Workers ~9 10-19 20-29 30-39 40-49 50-59 60-69 70-79 80-89 90-96 9~99.99 100 All male All female TOTAL o 2 2 13 18 27 35 52 202a 21 393 0.0 0.3 0.5 1.0 0.5 1.8 3.3 4.6 6.9 8.9 13.2 51.4 5.3 2.3 100 0.0 0.3 0.8 1.8 2.3 4.1 7.4 12.0 18.8 27.7 40.9 92.3 97.7 100 5 36 30 94.5 64 25.5 11 28 29.5 57 72 195 25 o 30 189 457 51 186 540 3,746 2,281 7,260 26,587 19,250 322 51 60,950 0.0 0.0 0.3 0.7 0.1 0.3 0.9 6.1 3.7 11.9 43.6 31.6 0.5 0.1 100 0.0 0.0 0.4 1.1 1.2 1.5 2.4 8.5 12.2 24.2 67.8 99.4 99.9 100 a Subsequent tables report 201 completely segregated establishments. After completing our analyses, however, we discovered 1 establishment for which the sole integrated job was in fact due to a coding error. tion of women. Second, we examine the subset of organizations for which longitu- dinal information is available in order to team (1) the extent to which patterns of sex seg- regation change over time and (2) the or- ganizational circumstances under which gender segregation increases, decreases, or remains constant. The concluding section addresses implications of our findings for policies aimed at equalizing job experiences and attainments of men and women. RESULTS Descriptive Statistics Table 3-2 presents a remarkable story: Most establishments are either completely seg- regated or nearly so. Less than one-f~fth of the establishments have segregation indices lower than 90, and they employ less than one-eighth of the workers in our sample. Over one-half of the establishments are completely segregated, and over three- quarters of the workers are in organizations having indices between 96 and 100. Indeed, only 10 percent of the nearly 61,000 workers are in establishment job titles that have both men and women assigned to them. Even among the 162 establishments having some men and women in the same job titles, the mean segregation index is 84.1. In short, the workplace is substantially more segre- gate(1 by sex than is suggested by studies that aggregate work force composition across establishments and into 3-digit occupational categories. i2 One way to segregate mate and female workers is to employ either men or women exclusively in an establishment. The 21 es- tablishments without female workers, listed by establishment size in Table 3-3, are al- i2 of course, statistics on the distribution of workers in our sample are not representative of the California labor force, since establishments, not workers, were sampled. Nevertheless, these results show that there are very few work contexts in which men and women are assigned to the same job titles, and results reported below suggest that even the least segregated enter- prises are seldom examples of workplace equity be- tween the sexes.
36 _. o sit o Cal - Ct - o o - Ct U) V' ._ U: 4- o in ·50 ¢ 7 o . Ct . be o ~ . m C,7 O Ct an,, ~ '=t .> 4) ~ O U' S .C} O . o.= ~0. Ct · =5 ~ ~ O ~ _ o _l Ed Cp =0 o ~8 U' - o C) Cal o :~ .> So C) en o Cal o z U' Cal Ct of ,¢ S t~0 0 Ct e S ~ C O C ~ C1 S ~ S 0 ~ S ,,Q ~ C U) ~ ~ C~ ~ P~ ¢ <~¢ << Z~ZZZZZZZ~ ZZZ^^ZZ~ Z Z O ° ° ° ~ ° ~ U. Co ° ~ ~ ~ 0 ~ ~ tS C~ 00 ~ 00 _ Z O O O O O O O O O O O O O O C~ ~ C~ C~ ~ _ O O O O O O ~ O ~ O ~ O O ~ ~ O ~ ~ d4 oo O O ~ O ~ ~ ~ O ~ ~ ~ 0 ~ '7 ~ ~ u: ~r _ _ _ _ _ ~ O O C~ O O O O O O CO O O ~ C~ O O O O O ~ O . . . . . . . . . . . . . . . . . . . . . C~ _ _ _ _ C~ _ C~ _ C~ _ C~ C] ~ _ C~ CO CO O 00 _ _ U: O ~ O O O O ~ 00 0 ¢ 00 U: O ~ ~ O _ O O CO . . . . . . . . . . _ . . . . _ . . . . . _ _ _ _ _ O O _ ~ O O _ O ~ _ _ CD IO ~ ~ ~ O ~ ~ _ _ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ 00 0 _ ~ t- CO c' V V V V ~ V V V V V V V ~ V V o ~ ~ ~ ~ ~ ~ ~ ~ ~ V¢ ~ ~ ¢V ~ ~ ~ ~ _ C~ CC CO S o0 ~ ~ ~ U) U: U) ~ ~ ~ ¢ ~ ~ ~ ¢ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ -¢ ~ .o e e ~ ~ ~ ~ ~ C ~ ~ =- ~ e =^ °~ ~ ' eO
38 most entirely small manufacturing opera- tions in which most workers are in produc- tion jobs. Only 1 of the 9 all-female establishments listed in Table 3-3 has more than 10 employees, and only 2 employ women in manufacturing. The typical all-female es- tablishment provides services and does not utilize the technical skills of blue-collar workers. 13 While extreme sex segregation is perva- sive, hiring policies that exclude one sex entirely are utilized only by establishments engaged in marginal economic activities on a very small scale. Only 5 of the all-male establishments are branches of larger com- panies, and 4 of those cite physical demands of work as the reason for excluding women. It seems that autonomous employers oper- ating small firms need no explicit rationale for excluding female workers; they can uni- laterally exercise their preferences for an all- male work force. The strong association between organi- zational scale and segregation is clocu- mented in the first line of Table 3-4. The table reports mean levels of various or~a- · . - 1 . ·~ nlzatlona~ attributes across 6 categories of sex segregation: moderate (A ~ 75~; high (75 < ~ ~ 901; very high (90 < ~ < 100~; com- plete segregation (A = 1001; exclusive em- ployment of males; and exclusive employ- ment of females. Excluding the 9 establishments employing only women, col- umn 7 reports the proportion of variance in each organizational attribute occurring across categories of segregation. i4 Establishments that exclude men or women are the smallest on average, followed by those that are com- i3 Were me 13 es~bbshmen~ m~ just 1 Mae worker, and these are quite similar to those listed in Table 3- 3. The largest has 22 employees and 9 have 5 workers or less. The 60 firms with just 1 female employee are concentrated in small-scale manufacturing and social overhead capital industries. Only 1 has more than 30 employees. i4 Tow exclusion of males is not only rue but also WILLIAM T. BIELBY AND JAMES N. BARON pletely segregated. Moderately segregated enterprises are larger still, followed by highly and almost perfectly segregated ones (antil- ogs of the means are 4, 9, 27, 30, 67, and 159 employees, respectively). Among the 384 establishments employing men, 30 per- cent of the variance in log size occurs across the 5 segregation categories. This strong association between organi- zational scale and segregation persists in multivariate analyses (see below) and seems to involve the changing mix of employer dis- cretion versus the impersonal rule of bur- eacratic procedures as organizations grow. For example, employers operating on a very small scale may exercise tastes for discrim- ination by excluding women altogether. However, as tasks become increasingly dif- ferentiated and specialized clerical roles are introduced, inexpensive female labor can be utilized in segregated job classifications. However, not all small establishments dif- ferentiate job tasks to the same degree, and some allocate both men and women to broadly defined job classifications. Contin- ued expansion leads to the implementation of rationalized, bureaucratic personnel pro- cedures in nearly all firms. Mechanisms seg- regating the sexes become institutionalizecI, and in large establishments men and women are almost always assigned to separate job families. Other organizational attributes listed in Table 3-4 are moderately associated with segregation levels. Means for each variable listed in Table 34 differ monotonically across categories of moderate, high, and very high segregation. Indeed, the characteristics most strongly associated with segregation core sectoral location, specialization, and frag- mentation are also highly correlated with seems to reflect processes qualitatively different from those that exclude or segregate women. Accordingly, we exclude the 9 all female establishments from the multivariate analyses reported below and from the var- iance explained computations reported in Table 3-4.
SEX SEGREGATION WITHIN ORGANIZATIONS . TABLE 3-4 Means of Organizational Attributes by Level of Sex Segregation 1~.e ~ I' (1) (2) (3) Log size Log specialization Manufacturing industry Periphery sector Log fragmentation Union/bidding arrangements Proportion women Proportion production workers Proportion clerical & sales workers Proportion service workers Proportion PTMC workers Average complexity: data Average complexity: people Skill specificity (4) (5) All li ~ 75 75 < ~ ~ 90 90 < 1` < 100 ~ = 100 Male (N = 42) (N = 40) (N = 80) . 3.40 1.34 (1. 52)a .45 (.46) .17 (.23) .52 (.50) .31 (.27) .09 (.12) .05 (.09) .44 .44 (.45) .28 (.27) .06 (.07) .22 (.21) 2.24 (2.23) 1.64 (1.58) Average technical skills -.37 (- .34) .43 (.57) 4.20 1.40 (1.12) .68 (.66) .50 (.40) .42 (.47) .08 (.13) .21 (.16) .30 (.23) .42 .55 (.54) .20 (.21) .08 (.08) .17 (.18) 1.88 (1.89) 1.05 (1.14) - .04 (-.09) ( .48 (.52) 07 72 93) .76 71) .78 49) .21 33) .01 18) .33 20) .46 28) .34 .61 56) .17 21) .09 09) .11 14) 73 77) .91 17) .12 02) 52 48) (6) All Female (N = 201) (N = 21) (N = 9) 3.31 2.17 .87 .74 (1.14) (1.64) .62 .67 (.63) (.71) .29 .05 (.37) ( 35) .45 .71 (.41) (.58) .26 .24 (~21) (.05 .10 - .14 (.13) (.00) .22 .14 (.28) (.35 .28 .00 .56 .74 (.56) (.80) .20 .02 ( 19) (- .02) .07 .03 (.08) (.03) .16 .21 (.16) (.19) 1.95 1.66 (1.94) (1.62) 1.36 1.41 (1.28) (1.15) .02 .30 .o5) (.48) .50 .53 (.50) (.47) a Entries in parentheses for columns (1)-(5) are means adjusted for linear association with size. b Entries in parentheses in right-most column are the increments to explained variance due to segregation categories, net of size. c Professional, technical, and managerial. 39 1.47 .39 .22 .00 .22 .88 .06 .00 1.00 .20 .39 .33 .33 2.36 1.50 .96 .31 ~2 Columns (1)-(5) (N = 384) .30 .12 (. 03)b .03 (.02) .21 (.02) .06 (.01) .08 (.02) .16 (.02) .08 (.02) .04 (.04) .05 (.06) .00 (.00) .03 (.01) .02 (.02) .04 (.07) .02 (.04) .01 (.oo) log size, suggesting that the pattern of means in the table may simply reflect concomitants of organizational scale. Therefore, we also control for each attribute's linear relation- ship with log size (coefficients in parenthe- ses). 15 While controlling for size weakens most relationships in Table 3-4, the general pat- tern of coefficients across categories of seg- i5 Specifically, each attribute was regressed on log size and binary variables representing segregation level, and adjusted means were computed at the average value of log size. The value reported in parentheses below the zero-order eta-squared for each attribute "y" is the increase in explained variance when variables denoting segregation group are added to the regression of y on log size.
40 WILLIAM T. BIELBY AND [AMES N. BARON regation is unchanged in many instances. Net of size, segregation is most strongly as- sociated with the level of interpersonal skflIs and the proportion of clerical and sales workers in the enterprise. The former var- iable seems to differentiate moderately in- tegrated establishments engaged in per- sonal services from all others, and the latter differentiates all-male establishments, with virtually no clerical component, from those with a significant fraction of the work force engaged in clerical duties. In sum, work settings in our sample ap- proach complete gender segregation, and descriptive statistics suggest two mecha- nisms that may contribute to patterns of seg- regation: one reflecting the impact of ad- ministrative structures and personnel procedures that vary with organizational scale and the other pertaining to the occupational composition and skid level of establishments that rely primarily on nonmanual tasks typ- ically done by women. The analyses re- ported below examine those mechanisms in greater detail. Multivariate Analyses Stanciarclizec] coefficients for discriminant analyses reported in Table 3-5 indicate the organizational attributes that best differen- tiate establishments according to levels of sex segregation. The first function defines the linear composite of organizational attri- butes that cTiffers most among segregation categories, relative to variation within cat- egories. The second function extracts an ad- ditional dimension differentiating among segregation categories and is uncorrelatec] with the first function. The standardized weights index the relative importance of each attribute in distinguishing among categories of segregation, and the group means locate the segregation categories along each com- posite dimension. The analyses reported i6 Discriminant analysis rests on distributional as- sumptions of within-group multinormality that are clearly in the first two columns inclucle the 21 all- male establishments; accordingly, the vari- ahie measuring organizational sex ratios (proportion women) is excluded from that discriminant analysis. The other results are based on the 363 establishments that em- ploy both men and women. Standardized weights are computed by scaling both the linear composite function and the organi- zational attributes to unit variances. Coef- ficients were computed from a canonical correlation analysis applied to a pairwise- deletion correlation matrix with binary var- iables denoting group membership. Cate- gory means on the discriminant function are metric coefficients for those binary varia- bles. Organizational scale clearly dominates the results. Log size has a standardized loading of.79 on the first Unction and is correlated .89 with the linear composite. Consequently, the group means convey the same message as Table 34: Scale accounts for most of the association between organizational attributes and segregation. The seconc! discriminant function differ- entiates the 42 moderately integrated en- terprises and the 21 all-male ones from other establishments, the overwhelming majority of which have segregation indices between 90 and 100. The function apparently reflects the impact of institutionalized personnel procedures and occupational composition. Formal bidding procedures for job advance- ment, combined with a work force occu- pying specialized, nonentry jobs, appar- entry inhibit the assignment of men and women to the same job classification, in some cases facilitating the exclusion of women employees altogether. This can occur when men and women are assigned to separate not met in our data. We are, however, not making inferences to a larger population, so significance tests are not appropriate for our analyses. We are reporting descriptive statistics about group differences, scaling levels of segregation to maximize the correlation with a linear combination of organizational attributes.
SEX SEGREGATION WITHIN ORGANIZATIONS 4 TABLE 3-5 Discriminant Analyses Relating Level of Sex Segregation to Organizational Attributes Including Establishments With Only Male Workers (N = 384) Standardized Weights 1st 2nd Function Function Variable Excluding Establishments With Only Male Workers (N = 363) Standardized Weights 1st 2nd Function Function Log size Log specialization Core sector Ambiguous sector Log fragmentation Union/bidding arrangements Proportion women Proportion production workers .27 (.02) .15 (.08) - .09 (- .21) -.11 (-.09) - .07 (- .29) .05 (.03) .06 (- .04) Segregation Group Means on Discriminant Function Proportion clerical and sales workers Proportion PTMb workers Average complexity: data Average complexity: people Average technical skills Skill specificity .79 .02 (.89)a (.01) .33 .36 (.48) (-.23) .37 - .03 (.72) (.32) .07 - .23 .40) (-~09) .24 - .15 (.66) (-.04) .00 .53 (.40) (~45) .74 (.87) - .22 (.52) .33 (.72) .07 (- .34) .21 (~59) .14 (.42) .35 (.14) .21 - .14 (.59) (.16) - .36 .16 (-.65) (- 12) - .01 .05 (- .24) (- .18) - .49 - .27 ( - .39) (- .18) .25 - .05 (-.30) (-.34) .03 .12 (.47) (.11) .49 .08 (.35) (.08) .20 .02) .68 .37) .06 (.26) .29 .20) .00 (.09) .32 (.37) .55 .67) .06 (.34) .20 .30) .34 .31) .27 .23) .21 .20) .03 (.35) .23 (.28) Moderate (N = 42) 0.00 0.00 0.00 0.00 High (N = 40) 1.43 2.23 1.36 1.36 Very high (N = 80) 2.19 2.98 2.37 2.71 Complete (N = 201) 0.19 2.64 -0.01 2.98 All male (N = 21) - 1.76 4.52 Canonical correlation .610 .309 .584 .345 a Correlations between composite function and organizational attributes appear in parentheses. b Professional, technical, and managerial.
42 WILLIAM T. BIELBY AND [AMES N. BARON - entry classifications and department-spe- cific seniority systems keep women from transferring into male career lines (Shaeffer and Lynton, 19791. Net ofthese tendencies, similar consequences occur when the cler- ical and sales component is small and few workers have complex informational tasks. i7 In other words, establishments in which most workers perform nonmanual tasks tend to have the occupational composition, task requirements, nonunion environment, and unspecialized! job structures that facilitate a modicum of parity in work arrangements by sex. Indeed, 9 of the 16 least segregated firms have no workers in production occu- pations and 2 others have less than 10 per- cent in such roles (see Table 3-6~. In con- trast, most all-male establishments are engaged in manufacturing activities and em- ploy most of their workers in production roles (see Table 3-31. Note that economic sector dominates neither discriminant func- tion. Thus, organizational arrangements as- sociated with sectoral location- not sector per seaffect sex segregation. While the discriminant analyses seem to support institutionalist accounts of sex seg- regation, a more parsimonious explanation may account for the results on the second Unction: It is easier to segregate women when they are a minority of the work force, regardless of administrative arrangements. Relative group size is often a crucial basis . . 17 Since the 4 variables characterizing occupational composition sum to 100 percent, coefficients depend on which of the 4 is omitted. When the proportion of workers in production occupations is omitted, the coef- ficients are -.49, -.13, and -.14 for clerical and sales; service; and professional, technical, and mana- gerial, respectively. Of course, zero-order correlations between the discriminant function and the measures of occupational composition are not affected by alter- native choices for the omitted variable. Percent service correlates - .01 with the second discriminant function. The second function correlates most highly with the percentage of clerical and sales workers ~ - . 65) and the proportion of production workers (.59~. Of solidarity and power (Simmer,  1950; Kanter, 1977a), but our analyses provide only partial support for the relative numbers hy- pothesis. Since sex composition defined membership in 1 of the 5 segregation groups (all-male), it was inappropriate to include that item in the discriminant analysis. Nevertheless, the second discriminant func- tion should be highly correlated with pro- portion women if the measures loading highly on it are simply proxies for organizational sex ratios. The correlation, however, ranges from -.27 and -.32 depending on how missing values are treated, about half the size of the correlations between the second function and occupational composition. Therefore, it appears that technical and ad- ministrative concomitants of production work are more important than relative numbers in differentiating levels of segregation. On the other hand, sex ratios do figure more prominently in the discriminant anal- ysis restricted to the 363 establishments that employ men and women. According to Table 3-5, the first dimension remains dominated by organizational scale, but specialization and sex ratios clearly define the second factor: Segregation increases monotonically as or- ganizations become more specialized and less dependent on female personnel. Unfortu- nately, the 2 cliscriminant analyses are not directly comparable, since the latter differ- entiates among segregation levels condi- tional upon a mixed work force, while the distinction between all-male establishments and others is prominent in the former anal- ysis. In short, there apparently is some strength in relative numbers, but that strength can be offset by countervailing organizational ar- rangements. Union contracts and formal bidding procedures, positional specializa- tion of the work force, reliance on firm-spe- cific skills, and manual job tasks facilitate employer strategies that either keep women out of the establishment completely or con- fine them in segregated job classifications.
SEX SEGREGATION WITHIN ORGANIZATIONS 43 Deviant Cases: Moderately Desegregated Establishments The foregoing analyses discriminate very segregated organizations from segregated ones, but we must not lose sight of the fact that there is very little variance to explain. Less than one-eighth of the establishments in our study are even moderately integrates] (using a generous definition of moderation), and the remaining, highly segregated or- ganizations display virtually every possible configuration of organizational form and en- vironment. The completely segregated es- tablishments include huge bureaucracies that dominate their industrial environments, as well as small entrepreneurial firms at the economic margins. These enterprises are public and private, in manufacturing and nonmanufacturing; some employ women al- most exclusively, while women have only a token presence in others. Indeed, an un- segregated work force is so rare that it is worthwhile examining commonalities among the few aberrant enterprises that do assign men ant] women to the same job cIassifi- cations. The top pane} of Table 3-6 lists organi- zational attributes of the 16 establishments with indices of dissimilarity less than 60. Examining them on a case by case basis re- vealed some regularities not cletected in the statistical analyses. First, we discovered that small enterprises with just one mixed job classification typically have inclices substan- tially less than 100. That is, small establish- ments that employ men and women can ap- pear moderately desegregated, but that desegregation can be nominal. The most ex- treme instances in Table 3-6 are the retail bookstore and pet store, each employing just one woman a sales clerk: Their indices are 37.5 and 50, respectively. Accordingly, sex segregation appears bimodally distributed among small firms. Of the 234 establish- ments with fewer than 50 employees, 173 are completely segregated. But the meclian segregation index among the other 61 is 75. Ten of the 16 establishments in Table 3-6 have relatively low segregation indices sim- ply by virtue of having one job title in which a few men and women are employed. Two real estate firms listed in Table 3-6 have men and women in integrated job titles but segregate them locationally. The third least segregated enterprise in our sample is a real estate management firm whose 23 male and 126 female apartment managers work and live in 149 different buildings, while the managers and officers of the escrow service are dispersed across field offices throughout a large metropolitan area. In both instances, men and women have the same rank and may have similar responsibilities, but within each of the individual workplaces there is perfect or near-perfect segregation. Another source of integrated job cIassifi- cations is the sex-linked practices in dealing with clients, accounting for low segregation indices in 4 other establishments. In 2 res- idential children's camps and an institution providing educational therapy, male coun- selors supervise boys and female counselors supervise girls. Another example is a lan- guage school, the second least segregated establishment in our sample, in which "it is deemed necessary that [students] be taught by both men and women as they are likely to need to communicate with each sex when using their language" (Narrative report 1712, 1970~. Gender role ideologies have histori- cally played an important part in creating and sustaining inequities against women in the teaching professions (Tyack and Strober, 1981~.1~ese educational establishments show that cultural definitions of men's and wom- en's work and responsibilities for social- izing the next generation can demarcate responsibilities by sex even within detailed organizational positions. Integrated work forces are utilized in sev- eral establishments under circumstances that corroborate neoclassical accounts of sex seg- regation (Polachek, 19791. Real estate sales
44 WILLIAM T. BIELBY AND [AMES N. BARON TABLE 3-6 Occupational Attributes of the Least Segregated and Largest Establishments Entry Products or No. Activities Sectora Industryb Total Female Employees Employees Segregation Job Index Titles Sixteen Least-Segregated Establishments 1 Citrus P Agriculture 30 14 12.5 3 2 Language schools A SOC 20 9 23.2 6 3 4 5 6 7 8 9 10 11 12 13 Real estate management A Real estate P Real estate escrow A Real estate Retail books Children's camps Tomatoes Real estate Retail pets Scientific instruments A Educational therapy A A p p p p p SOC SOC SOC SOC Trade Services Agriculture SOC Trade Manufacturing SOC 169 23 105 320 9 10 41 9 3 8 98 137 6 95 128 1 5 23 7 1 .2 57 28.1 33.3 36.3 37.0 37.5 40.0 42.3 50.0 50.0 50.0 53.6 9 5 12 63 5 6 2 2 5 35 14 Plumbing supplies A Manufacturing 11 2 55.6 7 15 Garments A Manufacturing 14 7 57.1 8 16 Children's Jaunts A Services 43 17 57.7 24 Eleven Largest Establishments 17 Mining and quarrying C Manufacturing 825 18 99.5 204 18 Bakery products C Manufacturing 886 268 96.3 60 19 Printing and publishing C Manufacturing 985 236 100.0 148 20 Sugar refining C Manufacturing 1277 82 99.0 337 21 Thoroughbred racing C Services 1464 49 96.1 83 22 Ordnance C Manufacturing 1727 967 99.2 129 23 Banking C SOC 2340 1384 79.2 346 24 Thoroughbred racing C Services 2845 59 99.5 83 25 Airline C SOC 2987 661 90.0 252 26 Telephone C SOC 6874 369 99.2 78 27 Naval shipyards C State, 7825 334 100.0 615 Manufacturing a p = periphery; A = ambiguous; C = core. b S0c = social overhead capital. is a vocation for which reentry costs to em- ployees and turnover costs to employers are minimal. Consequently, one would expect an abundant supply of and demand for qual- ified female workers for these positions. At the same time, real estate sales can be suf- ficiently lucrative especially in Califor- nia to attract males as well. Further, be- cause salespersons work primarily outside the office, there should be fewer costs as- sociated with employee tastes for discrimi- nation. If market forces, however, account for the disproportionate share of real estate figs among the moderately integrated es- tablishments, they cannot account for 4 of the other 5 real estate enterprises that com- pletely segregate men and women in other- wise similar market and organizational cir- cumstances. Similarly, it is not surprising to find in- tegrated work forces engaged in harvesting of fruits and vegetables. There is an ample
SEX SEGREGATION WITHIN ORGANIZATIONS 45 TABLE 3-6 (Continued) Occupational Composition: Percentage in Entry Job Titles Median Clerical No. per DOT Job Size Production Sales PTMC Service Comments Sixteen Least-Segregated Establishments o 8 1 1.0 0.9 s 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 1.1 0.5 1.8 1.4 1.0 1.0 1.0 1.0 1.0 1.0 1.3 0.8 1.3 1.4 1.9 1.4 1.4 2.4 1.3 1.4 3.2 1.3 NA 1.6 2.1 100 o 67.9 0 6.8 0 25.5 0 59.1 8 2.2 0 1.8 0 11.4 100 4.1 0 1.2 0 1.0 62 2.0 0 1.0 57 2.0 86 2.0 9 8.0 61.2 18.0 NA 111.5 102.6 15.0 268.3 111.7 157.5 69.2 79 35 79 75 15 84 o 8 NA 74 77 7 76 22 82 78 o o 94 67 12 17 25 o 5 6 54 13 10 60 6 68 o 92 93 24 78 10 22 70 o 5 o 25 80 18 14 70 0 Picker14M, 14F. 0 Language teachers6M, 6F; department head3M, 2F. o o o 1 o 30 o o 33 o 3 o o 16 Eleven Largest Establishments 14 3 5 11 7 8 30 33 25 NA NA 26 1 8 15 8 3 4 18 2 2 34 NA o o Apt. manager23M, 126F. Salesmen13M, 4F. Escrow manager4M, 24F; escrow officer3M, 32F. Salesman130M, 80F. Sales clerk5M, 1F. Counselor3M, 3F. Worker11M, 12F. Salesman1M, 7F. Salesperson1M, 1F. Instrument maker3M, 1F. Psychologist3M, 2F; therapist13M, 25F; camp counselor4M, 3F. Assembler4M, 1F. Four women in 3 integrated production jobs. CounselorllM, llF. No women placed in production jobs due to nature of work. Women mostly in clerical and bindery classifications. Eleven women in 3 integrated quality control classifica- tions; none in production. Women mostly in clerical classifications. See text. Company policy to place women in other than routine clerical positions. Several integrated officer and man- ager jobs. Women mostly in clerical classifications. Women employed only in clerical, stewardess, and ticket agent classifications. Women employed without restriction except in jobs re- quiring lifting 25 pounds or more. No women assigned to production classifications due to vigorous requirements of various crabcs. c PTM = professional, technical, and managerial occupations. supply of male and female workers who are ill prepared for most other types of em- ployment, and employers bear none of the training costs (l~omas, 19801. Neverthe- less, lhomas's study suggests that the two agricultural establishments in our sample may be atypical. He found that men in lettuce harvesting are typically assigned to higher paic} piece-rate jobs, while women are con- centrated in hourly crews. Unfortunately, we have no evidence of the generalizability of his or our findings, since the sex com- position of jobs was not compiled for the Employment Service's more recent analyses of agricultural work. In sum, no single dimension of desegre- gation emerges from our analysis of"de- viant" organizations. Instead, we found 4 qualitatively di~erent but sometimes over- lapping sets of circumstances that contribute to a desegregated workforce: (1) nominal de- segregation of a single job title in a very
46 WILLIAM T. BlELBY AND JAMES N. BARON small firm; (2) spatial segregation across work sites of mate and female workers assigned to the same job classification; (3) sex-linked desegregation of jobs like camp counselor, which employ men and women, but toward different ends; and (4) market desegregation that occurs when a mixed labor pool is avail- able to employers who perceive that train- ing and turnover costs are identical for male and female workers. Moderate desegrega- tion occurs rarely in the work contexts we have examined, and, when it does occur, it is typically of the nominal variety. Large Establishments and Bureaucratic Segregation Almost all large establishments are highly segregated, and most have written rules governing the hiring and allocation of work- ers. Consequently, if bureaucratic control strategies segregate men from women (Ed- wards, 1979), this should be most apparent in the largest establishments. The bottom pane! of Table 3-6 lists attributes of the 11 establishments in our sample employing more than 800 workers. Case materials provide insight into administrative roles and pro- cedures that support sex segregation. All but one of these establishments seg- regate employees almost perfectly by sex, and narrative information available for 7 of them suggests that this total segregation is accomplished largely through bureaucratic rules and procedures. Possibly the most im- portant factor, particularly in manufacturing establishments, is the existence of legal re- strictions on the weight that women may lift. California law specified until 1970 that "no female employee should be requested or permitted to lift any object weighing 50 pounds or over," and regulations enforced by the state's Industrial Welfare Commis- sion further restricted the maximum to 25 pounds. In 1970 a federal court ruled that this law conflicted with Title VII of the Civil Rights Act (Utility Workers' Union of Amer- a vs. Southern California Edwon t69-5434~. References to these restrictions occur re- peatedly in both narrative reports and job analyses. For example, the Naval shipyard in Table 3-6 assigns no women to production jobs due to "vigorous requirements of var- ious crafts" (Narrative report 3, 1965), and the mining enterprise followed the same policy due to the "nature of work" (Narrative report 1800, 1971~. A company providing telephone service to a large metropolitan area states that it was company policy to employ women "without restriction" except in jobs requiring lifting 25 pounds or more, yet the segregation index was 99.2 (Narra- tive report 100, 1965~. The ordnance plant, studied in 1970, em- ployed 555 females as assemblers and 243 males as production workers. Each is an en- try job, and both were mapped to the same detailed DOT occupational category by the Employment Service analysts According to the job analysis, these jobs differ pri- marily in that male "workers" lift 25 to 40 pounds, but female "assemblers" lift 5 to 20 pounds. Weight restrictions are not mentioned ex- plicitly in the narrative for the printing es- tablishment in Table 3-6 (one of the largest in the western U. S. in 1968), but as in nearly all other manufacturing plants, the only pro- duction activities assigned to women are light assembly tasks. Most production tasks in this establishment are done by skilled crafts- men. The union contract establishes pro- cedures for hiring and apprenticeship, and it seems reasonable to conclude that the union plays a substantial role in enforcing sex seg- regation in this plant. A narrative report prepared in 1966 for one of the two race tracks provides evidence of that role in an- other organizational context: Union dis- patching policies explicitly exclude women from the job of parimutuel clerk (Narrative report 15361. No report was prepared for the other track in the same area, but pre- i~ In other words, at the occupational level, entry- level production work has a relatively balanced sex ratio in this firm.
SEX SEGREGATION WITHIN ORGANIZATIONS 47 sumably it falls under the jurisdiction of the same union. Of course, these industries might have been sex-segregated long before they were unionized. That is, unions may be per- petuating gender-based inequalities rather than creating them. Lifting restrictions and union contracts cannot account for all sex segregation in large establishments, because segregation is equally pervasive in large, nonunion estab- lishments outside manufacturing. Indeed, in most firms including all but 1 of the 11 in the bottom pane! of Table 3-6 non- production jobs are as segregates! as pro- duction jobs, if not more so. While this sit- uation primarily reflects distinctions between male managers and female clerical workers, it is also true that the few male clerical work- ers and female managers in our sample of establishments are hardly ever assigned to job classifications with workers of the op- posite sex. One notable exception is the bank studied in 1968, which employs more women than any other establishment in our sample. The bank claimed it had recently initiated a pro- gram to hire and promote women into of- ficer classifications (Narrative report 415, 19681. The staffing schedule supports this claim: In 1968, women comprised 7 of 81 vice presidents, 18 of 108 assistant vice pres- idents, 16 of 49 management trainees, and 29 of 118 operations officers. While females were used exclusively in routine data pro- cessing jobs like keypunch operator, 10 of 23 systems analysts were women. In one sense the bank's efforts are only noteworthy when contrasted! against the uniformly high levels of segregation in other comparable large establishments: Fully equalizing the job distribution by sex would still require reclassifying 80 percent of this bank's female employees. Most managerial and professional positions remained exclu- sively male, while few men were employed in routine clerical duties. An organization's demography, history, technology, and labor supply, however, constrain the degree to which its work force composition can change in a short period of time. This is especially true of large bureaucracies employing many workers in nonentry jobs. As long as sen- iority ant! accrued skills remain important bases for advancement in such contexts, workplace equity cannot happen overnight: The existing stratification regime favoring males essentially guarantees more work- place inequality in the short run, just as reduction of childbearing to below replace- ment levels would not immediately elimi- nate population growth in a society. This particular bank's egalitarian policies toward women seem to have overcome bu- reaucratic inertia because they were imple- mented during a time of extraordinary or- ganizational growth and change. When analyzed in 1968, the bank was described as one of the "largest and fastest growing business concerns in the nation" (Narrative report 415, 19681. Employment increased nearly 50 percent between 1961 and 1968, when the bank was automating its data pro- cessing operation and establishing regional offices and branch banks throughout south- ern California. Growth and technological change appear to be directly responsible for the desegregation of several management, administrative, and data processing job cias- sifications in this firm. Its atypical experi- ence demonstrates that the segregation ob- served in other large establishments is not inevitable. i9 These jobs, however, may have ,9 Other data gathered in California by the Employ- ment Service corroborate this. They studied about 30 other establishments with more than 800 employees not included in our analysis because of incomplete cov- erage of some aspect of their operations. Of these, only 3 were moderately desegregated: a university campus (/\ = 70.5), a unified school district (/\ = 76.5), and an insurance company (/\ = 77.21. Among large bu- reaucracies, assignment of both sexes to the same job titles occurs most often in social overhead capital or- ganizations- firms in health, education, and welfare services, transportation, utilities, finance, insurance, and real estate industries. But the banking and finance industry is not uniformly desegregated. Of the 5 other establishments in our sample engaged in such activi- ties, 3 were completely segregated, 1 was nearly so At\ = 93.3), and 1 was moderately integrated (I\ = 73.7).
48 WILLIAM T. BIELBY AND [AMES N. BARON TABLE 3-7 Descriptive Statistics for Longitudinal Sample (N = 75) Time 1 Time 2 Standard Standard Variable Mean Deviation Mean Deviation Log size 4.26 1.21 4.38 1.21 Log specialization 1.37 0.99 1.48 1.03 Manufacturing industry .79 Core sector .61 Ambiguous sector .32 Periphery sector .07 - Log fragmentation .18 .20 .17 .19 Union/bidding arrangements .37 .39 Proportion women .29 .2S .29 .24 Proportion production workers .64 .30 .62 .29 Proportion clerical and sales workers .16 .17 .18 .18 Proportion service workers .08 .22 .09 .22 Proportion PI Ma workers .10 .12 .11 .1 1 a Professional, technical, and managerial. been integrated precisely because they were new; it may take several years for the sex label of a new line of work to become es- tablished. Longitudinal Analyses: The Permanence of Sex Segregation Seventy-fve of the 393 establishments in our sample were studied more than once. The average interval between visits was about 5 years, with a range of 2 to 12 years. The size composition of the follow-up sample is very consistent with the age and size-spe- cif~c establishment mortality rates reported by Birch (19791. While moderately large, older enterprises were slightly more likely to be revisited, we detected no other sys- tematic biases in the Employment Service's choice of establishments for follow-up anal- yses (for details see Baron, 1982, Chapter VII). Descriptive statistics for the 75 establish- ments are reported in Table 3 7.2° Estab- 20 Several establishments were studied more than twice by the Employment Service. In those instances, we selected the pair for which the interval between anal- yses was closest to 5 years. Consequently, analyses for 16 of the establishments in the longitudinal analyses do not include the one selected for the cross-sectional sample of 393 observations. lishments selected for restudy were slightly larger on average and more concentrated in manufacturing; other differences between descriptive statistics for these observations and those for the entire sample reflect con- comitants of organizational scale (cf. Tables 3-1 and 3-71. In most instances, the temporal data describe changes between the mid-1960s and the early 1970s the period immedi- ately following passage of Title VII of the Civil Rights Act of 1964. Most establish- ments expander} employment between anal- yses; the labor force was stable in just 6 cases and was reduced in 25 establishments. As Table 3-8 shows, neither legislation nor organizational change effected much change in sex segregation in the late 1960s and early 1970s. Two-thirds of the estab- lishments remained all-male, completely segregated, or almost fully segregated (see the 9 cells in the bottom-right corner of the table). Indeed, much of the change in Table 3-8 reflects very small differences in seg- regation indices. Table 3-9 lists characteristics of the 18 establishments for which the segregation in- dex changed by at least 5 points. Only a few attributes differed systematically between the 11 organizations that became less seg- regated, between the 7 that became more segregated, and between the 57 in which
SEX SEGREGATION WITHIN ORGANIZATIONS - TABLE 3-8 Segregation Levels Over Time (N = 75) Segregation: Time 2 Segregation: Time 1 Moderate (lt ~ 75) Very Mod High High All Complete Male Total High (75 < ~ < 90) Very high (90 < ~ < 100) Complete Hi= 100) All male TOTAL 3 2 2 2 4 15 3 4 5 29 2 l 5 10 23 34 3 (7%) (13%) (29%) (47%) (4%) (1%) 9 (12%) 23 (31%) 40 (53%) 2 (3%) 75 (100%) sex segregation remained virtually constant. Unionized establishments and those with formal bidding arrangements tended not to change from their high levels of segregation, nor did those in which women comprised a small minority of the work force. Organi- zations that partially desegregated were typ- ically in small-scale manufacturing or were providers of personal and social services. Most declines in segregation are attrib- utable to a nominal change in the compo- sition of 1 or 2 jobs rather than a major change in personnel practices. For example, 2 male checkers were hired by a large-chain supermarket, a mate operator was hired by a firm providing mobile telephone service, and a woman was hired to work as a lens finisher in a firm manufacturing contact len- ses. The "integrated" title of chip girl in the poker cardroom is completely segregated by shift; apparently, only males work in the early morning hours. 2} A company that man- ufactured silkscreened wall coverings hired 4 female inspectors, but continued to dis- criminate statistically, hiring women only as inspectors, paint mixers, and clericals "due to occasional job requirements of lifting heavy 21 The apparent integration of the chip girl position may simply reflect a typographical error on the 1970 staffing schedule. If so, the establishment was perfectly segregated in 1966 and 1970. reels of paper" in the other production work (Narrative report 1476, 19701. Only a man- ufacturer of kitchen ranges displays a delib- erate effort to desegregate its work force. Only 2 women were employed in produc- tion departments in 1966. As in the case of the bank described earlier, organizational growth appears to have facilitated gender parity. By 1970 the work force of this man- ufacturing firm had more than doubled, and women worked alongsicle men in 14 entry- level jobs in the assembly department, re- clucing the segregation index from 99.5 to 78.8. Nevertheless, as in the bank, employ- ment in several other nonadministrative de- partments remained restricted to males in 1970. The 7 establishments that became more segregated over time were all highly seg- regatecT initially, ant! typically the segrega- tion of only 1 job produced most or all of the change. In several instances, increased segregation took place in the context of or- ganizational retrenchment or consolidation. But in no case was there any evidence that increased segregation resulted from clelib- erate employer strategy to manipulate the sex composition of jobs. Given the high ini- tial levels of segregation, the small increases in the index could easily be attributable to random perturbations in labor supply and demand.
so WILLIAM T. BIELBY AND JAMES N. BARON TABLE 3-9 Attributes of Establishments With Increasing or Decreasing Levels of Sex Segregation Number of Number of Segregation Entry Year Employees Females Index No. Products or Activities T1-T2 Sectora Industryb T1-T2 T1-T2 T1-T2 Establishments With Decreasing Segregation 1 Supermarkets 1961-1967 A Trade 56-49 16-12 100.0-94.6 2 Optical goods 1965-1968 C Manufacturing 42-44 18-21 77.8-69.6 3 Mental health facilities 1966-1970 A SOC 149-145 124-118 96.0-86.4 4 Wall coverings 1965-1970 C Manufacturing 139-142 12-26 96.1-85.8 5 Mobile telephone service 1966-1971 P SOC 15-30 13-21 100.0-88.9 6 Clay products 1964-1971 A Manufacturing 20-35 3-11 100.0-87.5 7 Poker cardrooms 1966-1970 A Services 149-190 97-66 100.0-87.1 8 Optical goods 1965-1968 A Manufacturing 9-10 3-3 100.0-85.7 9 Kitchen ranges 1966-1970 C Manufacturing 198-438 8-32 99.5-78.8 10 Wigs 1965-1971 A Manufacturing 11-11 10-6 90.0-63.3 11 Wall coverings 1965-1970 C Manufacturing 27-54 6-17 95.2-67.9 Establishments With Increasing Segregation 12 Trophies 1961-1971 C Manufacturing 75-58 29-33 86.6-92.0 13 Electronic instruments 1967-1970 C Manufacturing 95-49 20-11 84.0-89.5 14 Pottery 1960-1968 C Manufacturing 309-535 137-294 92.7-99.2 15 Musical instruments 1960-1970 C Manufacturing 164-173 10-4 89.0-95.9 16 Ceramic tiles 1971-1977 A Manufacturing 21-42 9-18 91.7-100.0 17 Securities exchange 1960-1968 A SOC 60-55 42-21 80.1-100.0 18 Garments 1960-1970 A Manufacturing 25-26 16-17 77.8-100.0 a p = periphery; A = ambiguous; C = core. b S0c = social overhead capital. In sum, changes in organizational forms and environments and shifting labor supply and demand had little eject on sex segre- gation in the late 1960s and early 1970s.22 In most of the establishments we examined, the consistently high levels of segregation are probably due to long-standing policies for hiring and allocating workers, perhaps reflecting industrywicle practices predating the establishment itself. Neither a changing 22 Nineteen establishments were visited 3 or more times by the Employment Service. Seven were per- fectly segregated at each visit, and segregation de- creased between analyses in only 1 establishment. In short, there is no evidence of a secular trend in job segregation by sex between 1959 and 1979. political and legal climate nor an influx of women workers in the late 1960s affected definitions of women's work within this sam- ple of work organizations, especially in large, unionized establishments with institution- alized procedures governing hiring and ad- vancement. DISCUSSION In most establishments, few job ciassifi- cations are staked by both men and women. IncleecI, complete segregation was the norm in establishments studied in California by the U. S. Employment Service between 1959 and 1979, and segregation levels were vir- tually constant in these organizations during
SEX SEGREGATION WITHIN ORGANIZATIONS 5 TABLE 3-9 (Continued) If segregation due mostly to one title: T1 T2 Number Number Number of Number of of Males Females of Males Females Comments Entry No. Title 1 2 3 4 5 6 7 8 9 Checker Several Several Several Operator Press operator Chip girl Finisher Several 10 Ventilator 11 Several 12 Several 13 Assembler 14 Packer 15 Assembler 16 Kiln loader 17 Tabulation clerk 18 Establishments With Decreasing Segregation 0 10 2 0 10 1 2 0 3 0 35 16 2 0 1 10 5 3 1 6 Sewing machine operator 1 13 5 3 1 5 3 9 2 18 1 1 3 Establishments With Increasing Segregation 3 o 5 o 8 8 o o o 3 5 Managerial promotion track for box boy but not for checker. Integration of 2 lens grinding jobs. Integration of kitchen helper (initially male) and head nurse (initially female) titles. Integration of 2 inspector titles (initially male). Chip girl title segregated by shifts. Desegregated 14 of 19 jobs in assembly department; 7 of 8 other departments completely segregated. Integration of 1 nonproduction and 3 production jobs; 23 of 28 jobs still segregated in 1970. Discontinued jewelry manufacturing, eliminating 2 integrated jobs. Decline in production between 1967 and 1970. Kiln loader only integrated job in 1971. Data processing operation consolidated as separate subsidiary between 1960 and 1968. Administrative jobs not included in staffing schedule for - 1970. the late 1960s and early 1970s. When bal- anced sex ratios did occur, they almost al- ways reflected just l or 2 integrated job titles within an establishment. Multivariate analyses revealed that or- ganizational scale is strongly associated with levels of sex segregation. Small enterprises are either completely segregated or trivially desegregated; larger enterprises tent! to have almost all workers in segregated jobs. Union contracts and formal bidding procedures, positional specialization, reliance on trained nonentry personnel, and manual job tasks facilitate a division of labor segregated by sex, while women are less extensively seg- regated when they comprise more than a small minority of the work force. Institutional arrangements that prevail in core firms rather than sectoral location per se shape sex segregation, and they are most visible in the largest establishments in our sample. Many had sex-specific policies for allocating workers, oftentimes sanc- tione(1 by collective bargaining agreements. In manufacturing establishments, legal re- strictions on physical demands of women's work typically rationalized complete sex seg- regation among production jobs, even when lifting was only required infrequently. Estab- lishments in the economic core also specialize skills and fragment job tasks more than other enterprises, and these aspects of bureaucracy are not gender-neutral. Of course, most women are not employed
52 WILLIAM T. BIELBY AND [AMES N. BARON in such manufacturing contexts. Less than one-quarter of all women workers are in manufacturing enterprises with 50 or more employees, and one-half the females in non- manufacturing industries work in establish- ments with fewer than 50 employees.23 Even small organizations outside of manufactur- ing, however, are highly segregated. Of the 100 nonmanufacturing establishments in our sample having fewer than 50 employees, 73 were perfectly segregated, and women em- ployed in those establishments are more likely to be segregated through the patriarchical actions of an entrepreneur or supervisor than by institutionalized bureaucratic rules and procedures. In short, there is equifinality in sex segregation of the workplace. Personal and patriarchal control prevails in small es- tablishments, bureaucratic structures seg- regate men and women in large nonmanu- facturing enterprises, and technical control excludes women from many production jobs within the manufacturing sector. Do these findings imply that sex segre- gation is immutable? We detected little change in segregation indices, even among firms analyzed after court decisions in 1971 struck down California's restrictions on women's lifting. 24 In 1968, affirmative action goals and timetables were required of firms holding federal contracts, yet establish- ments in our sample covered by this order 23 Figures are based on the 1967 Technological Ad- vance in an Expanding Economy survey by the Uni- versity of Michigan's Institute for Social Research (Mueller et al., 1969~. Of 766 women working outside the home, 560 were employed in nonmanufacturing industries, and 282 of those were in establishments with fewer than 50 employees. Since then, the female work force has become even less concentrated in man- ufacturing, although there may have been a shift to- ward employment in larger establishments. 24 Of 46 establishments analyzed between 1972 and 1979, 26 were completely segregated. Eleven had seg- regation indices less than 75, but 5 were real estate enterprises studied by the Employment Service in 1973. Levels of segregation remained uniformly high in man- ufacturing establishments studied after 1971. were no less segregated than others (see also Salancik, 19791. Nevertheless, we can point to some ex- traordinary circumstances under which de- segregation did occur. As noted above, one bank and a manufacturing enterprise inte- grated a number of nonclerical jobs during a time of rapid growth or technological change or both, before there was any substantial government pressure to change personnel policies regarding minorities. Short ofthese fortuitous and idiosyncratic circumstances, large and systematic reductions in gender segregation seem unlikely to occur in the absence of fundamental shocks to the social system. For example, during World War II, employers faced extreme labor shortages. California's Industrial Welfare Commission granted 6Q permits that exempted women from its 25-pound restriction on lifting, al- lowing 4,539 women workers to enter pro- duction jobs formerly closed to them.25 Thus, unusual extraorganizational circumstances forced employers to reject long-standing practices based ostensibly on physiological differences between the sexes. Neverthe- less, after the war these women were de- mobilized as rapidly as they had been in- tegrated into the work force. The procedures for obtaining exemptions from the state la- bor cocle remained in effect but were barely utilized, and until 1971 the code rational- ized sex segregation of production jobs in manufacturing. To the extent that our results are gener- alizable, two policy implications of our study are clear. First, sex segregation will un- doubtedly persist if policy makers adopt a laissez faire stance. Neither demographic trends, nor technological changes, nor bu- reaucratic imperatives are natural forces Mat lead to balanced sex ratios within jobs or firms. Second, policy intervention is un- 25 Personal communication, Margaret Miller, Exec- utive Officer, California Industrial Welfare Commis- sion.
SEX SEGREGATION WITHIN ORGANIZATIONS . 53 likely to make matters worse most estab- lishments are about as segregated as they can possibly be. While there may be polit- ical or economic incentives for focusing on certain kinds of work contexts, policy efforts could justifiably be directed at almost any area of economic activity, since almost all establishments are equally segregated. Our findings also suggest some strategies for attacking sex segregation. First, policies that segregate men and women are most visible in large organizations. They are doc- umented in written contracts, rules, job def- initions, and procedures; they do not exist solely in subjective tastes of employers, em- ployees, and clients. Large firms are often subject to public scrutiny, and their greater dependence on government contracts and regulations makes them susceptible to pol- icy interventions (Salancik, 19791. They are also more likely to have slack resources with which to absorb the costs of social change. In short, it seems reasonable first to redress visible and easily identifiable mechanisms of sex segregation in organizations that are vulnerable to outside pressures, even if only a fraction of all women work in such estab- lishments. Second, interventions seem more likely to succeed in organizations that already have a sizable female work force. Sex ratios affect the balance of power among organizational constituents (Kanter, 1977a), and our results show that, as women comprise a larger per- centage of an organization's labor force, em- ployers seem less likely to segregate them. Changing the sex composition of jobs will require modifying organizations' rules for advancement through internal labor mar- kets. Such changes are easier to accomplish when female workers command firm-spe- cific experience. In short, segregation seems likely to persist in the absence of severe external pressures on the organization. Fur- thermore, the technical and political viabil- ity of efforts to abate the sexual division of labor depends on the existence of a constit- uency inside. The presence of a visible con- tingent of minority employees within an or- ganization even at the lowest ranks may facilitate efforts to desegregate work. Efforts to impose workplace equity in the absence of such a constituency might ac- tually backfire. "Tokens" or "solos" can so- lidify resistance by male workers and de- moralize those who should champion and benefit from equal employment opportunity programs (Kanter, 1977a; NortheraPc and Martin, 1981~. This underscores once again the limits of laissez faire approaches, since nearly all of the natural desegregation we observed in our sample involved tokenism. Our recommendations are based on statis- tical associations rather than examinations of specific policy interventions. Nevertheless, recent surveys of attempts to reduce sex seg- regation do document the effectiveness of the organizational strategies we have suggested (Shaeffer and Lynton, 1979; O'FarreD and Harlan, 19821. One conclusion cannot be dis- puted: Doing nothing guarantees persistent sex segregation in all areas of economic activ- ity. Although not the focus of this study, we repeatedly encountered instances of sex seg- regation of jobs leading to gender-specific pro- motion lines an orderly progression through jobs of successively greater authority and re- sponsibility for men and dead-end careers for women. The pervasive sex segregation across organizational and institutional contexts that our study has documented almost certainly accounts for the substantial sex differences re- vealed by individual-level analyses of work inequalities (e.g., Baron and Bielby, 1982; Bielby and Baron, 1982; Wolf and Fligstein, 19791. The degree and persistence of sex seg- regation leave us somewhat pessimistic about prospects for rapid or extensive change. On the other hand, by documenting the impact of organizational structures and dynamics on the sexual division of labor ant! by un- derscoring the pervasiveness of gender seg- regation we hope to facilitate more in- formed research efforts and policy inter- ventions in the future.
54 _ WILLIAM T. BIELBY AND [AMES N. BARON ACKNOWLEDGMENTS The authors were supported in part by a grant from the National Science Foundation (SES 79-24905~. We gratefully acknowledge the cooperation and assistance of the Cali- fornia Occupational Analysis Field Center. Our thanks go to Isabelle M. Aliain, Hallie Anderson, and Kelly Moreno who assisted in this research, and to Trivia Miller who processed many versions of this manuscript. We are indebted to our many colleagues who have influenced our thinking on sex segregation and have commented on earlier versions of this report, especially to Donald I. Treiman and Karen O. Mason for their detailed suggestions. REFERENCES Arrow, Kenneth 1973 "The theory of discrimination," Pp. 3-33 in Or- ley Ashenfelter and Albert Bees (eds.), Dis- crimination in Labor Markets. Princeton, N.J.: Princeton University Press. Averitt, Robert 1968 The Dual Economy. New York: Horton. Baron, James 1982 Economic Segmentation and the Organization of Work. Unpublished Ph.D. dissertation, University of California, Santa Barbara. Baron, James, and William Bielby 1982 "Workers and machines: dimensions and de- terminants of technical relations in the work- place." American Sociological Review 47:175- 188. Bielby, Denise 1978 "Career sex-atypicality and career involvement of college educated women: baseline evidence from the 1960's." Sociology of Education 51 Qanuary):7-28. Bielby, William, and James Baron 1982 "Organizations, technology, and worker attach- ment to the firm." In Donald J. Treiman and Robert V. Robinson feds.), Research in Social Stratification and Mobility, Vol. II. Greenwich, Conn.: JAI Press. Birch, David 1979 The Job Generation Process. Cambridge, Mass.: MIT Program on Neighborhood and Regional Change. Blau, Francine 1977 Equal Pay in the Office. Lexington, Mass.: Lexington Books. Blau, Francine, and Wallace Hendricks 1978 "Occupational segregation by sex: trends and prospects." Journal of Human Resources 14 (2~:197-210. Blau, Francine, and Carol Jusenius 1976 "Economists' approaches to sex segregation in the labor market: an appraisal." Pp. 181-99 in Martha Blaxall and Barbara Reagan (eds.), Women and the Workplace: The Implications of Occupational Segregation. Chicago: Uni- versity of Chicago Press. Braverman, Harry 1974 Labor and Monopoly Capital. New York: Monthly Review Press. Bridges, William, and Richard Berk 1974 "Determinants of white-collar income: an eval- uation of equal pay for equal work." Social Sci- ence Research 3:211-34. Doeringer, Peter, and Michael Piore 1971 Internal Labor Markets and Manpower Anal- ysis. Lexington, Mass.: D. C. Heath. Duncan, Otis, and Beverly Duncan 1955 "A methodological analysis of segregation in- dexes." American Sociological Review 20:210- 17. Edwards, Richard 1979 Contested Terrain. The Transformation of the Workplace in America. New York: Basic Books. England, Paula 1981a "Assessing trends in occupational sex segre- gation, 1900-1976." Pp. 273-95 in Ivar Berg (ed.), Sociological Perspectives on Labor Mar- kets. New York: Academic Press. 1981b "Explanations of occupational sex segregation: an interdisciplinary review." Unpublished manuscript, University of Texas at Dallas (Jan- uary). 1982 "The failure of human capital theory to explain occupational sex segregation." Journal of Hu- man Resources 17:358-70. Gibbs, Jack, and Dudley Poston, Jr. 1975 "The division of labor: conceptualization and related measures." Social Forces 53:468-76. Gross, Edward 1968 "Plus ca change . . . ? The sexual structure of occupations over time." Social Problems 16:198- 208. Halaby, Charles 1979 "Job-specific sex differences in organizational reward attainment: wage discrimination vs. rank segregation." Social Forces 58 (September): 108-127. Kanter, Rosabeth 1977a "Some effects of proportions on group life: skewed sex ratios and responses to token women." American Journal of Sociology 82:965- 90. 1977b Men and Women of the Colporation. New York: Harper & Row. Marini, Margaret 1980 "Sex differences in the process of occupational attainment: a closer look." Social Science Re- search 9:307-61.
SEX SEGREGATION WITHIN ORGANIZATIONS 55 1978 "Sex differences in the determination of ado- lescent aspirations: a review of research." Sex Roles 4:723-53. Milkman, Ruth 1980 "Organizing the sexual division of labor: his- torical perspectives on women's work and the American labor movement." Socialist Review 49:95-150. Miller, Ann R., Donald J. Treiman, Pamela S. Cain, and Patricia A. Boos 1980 Work, Jobs, and Occupations: A Critical Re- view of the Dictionary of Occupational Titles. Washington D.C.: National Academy Press. Mueller, Eva, Judith Hybels, Jay Schmiedeskamp, John Sonquist, and Charles Staelin 1969 Technological Advance in an Expanding Econ- omy. Ann Arbor, Mich.: Institute for Social Research, University of Michigan. Northcraft, Gregory, and Joanne Martin 1981 "Double jeopardy: why some women and mi- norities~object to affirmative action programs." Research paper 599, Graduate School of Busi- ness, Stanford University. O'Farrell, Brigid, and Sharon Harlan 1982 "Job integration strategies: today's programs and tomorrow's needs." Paper presented at the National Research Council's Workshop on Job Segregation by Sex, May, Washington, D.C. Polachek, Solomon 1979 "Occupational segregation among women: the- ory, evidence, and a prognosis." Pp. 137-57 in Cynthia B. Lloyd, Emily S. Andrews, and Cur- tis L. Gilroy feds.), Women in the Labor Mar- ket. New York: Columbia University Press. Salancik, Gerald 1979 "Interorganizational dependence and respon- siveness to affirmative action: the case of women and defense contractors." Academy of Man- agement Journal 22:375-94. Shaeffer, Ruth Gilbert, and Edith F. Lynton 1979 Corporate Experiences in Improving Women's Job Opportunities. New York: The Conference Board. Simmel, George 1950 "Quantitative aspects of the group." Pp. 87- 177 in Kurt H. Wolff (ed.), The Sociology of George Simmel. New York: Free Press. Orig- inally published in 1923. Thomas, Robert 1980 "The social organization of industrial agricul- ture." Paper presented at the 1980 meeting of the American Sociological Association, New York. Treiman, Donald J., and Heidi I. Hartmann feds.) 1981 Women, Work and Wages: Equal Pay for Jobs of Equal Value. Washington, D.C.: National Academy Press. Tyack, David, and Myra Strober 1981 "Jobs and gender: a history of the structuring of educational employment by sex." Pp. 131- 52 in Patricia Schmuck and W. W. Charters feds.), Educational Policy and Management: Sex Differentials. New York: Academic Press. U.S. Bureau of the Census 1976 County Business Patterns, United States Sum- mary. Washington, D.C.: U.S. Government Printing Office. U.S. Department of Labor 1972 Handbook for Analyzing Jobs. Manpower Administration. Washington, D.C.: U.S. Gov- ernment Printing Office. Williams, Gregory 1979 "1he changing U.S. labor force and occupa- ffonal differentiation by sex." Demography 16:73-88. Wolf, Wendy C., and Neil D. Fligstein 1979 "Sex and authority in the workplace: the causes of sexual inequality." American Sociological Review 44:232-52.