Home Submissions Student Writing Exec. Board & Staff Subscription Past Issues Conference Links Search
logo logo

WOMEN CHOOSING DIVERSE WORKPLACES:
A RATIONAL PREFERENCE WITH DISTURBING IMPLICATIONS FOR BOTH OCCUPATIONAL SEGREGATION AND ECONOMIC ANALYSIS OF LAW

Scott A. Moss[*]

Overview

Despite women’s dramatic labor market gains, there remains a striking degree of occupational segregation by gender. Analysts typically blame discrimination or women’s work/family priorities. This Article offers a different explanation.

It is hard for women choosing jobs or occupations to know where they will face discrimination, particularly since recent judicial decisions eliminated certain employer signals that once differentiated fair and discriminatory firms. One way women can effectuate a preference for nondiscriminatory workplaces is by choosing gender-diverse workplaces. Nondiverse workplaces often are not female-friendly, and discrimination may be the reason they are nondiverse. In economic terms, women rationally use level of diversity as a proxy for discrimination, since the latter is harder to observe.

When a preference for diversity is incorporated into standard labor-economic analyses, it generates a bleak prediction: women’s preferences for diversity can yield enduring segregation, even in nondiscriminatory workplaces, and even as more women enter the labor force. Title VII has reduced discrimination and raised female labor force participation, but neither of these successes may reduce segregation.

Normatively, although this Article could be interpreted to support a “conservative” argument to narrow Title VII liability (because a workplace underrepresentation of women may reflect women’s own choices rather than discrimination), it more forcefully supports a contrary “liberal” argument: to redress segregation, we must go beyond just fighting actual discrimination. Advisable measures include changes to the focus

*** Top of Page 2 ***

and burdens of proof in Title VII cases, affirmative action, and education reform.

More theoretically, certain standard economic tenets, such as fixed exogenous preferences and the Coase Theorem, are undercut by key features of these models (e.g., endogenous labor supply preferences, path-dependent market evolution, and multiple equilibria). We cannot assume that such markets, left unregulated, are efficient, so the case for labor market intervention is much stronger than economics typically recognizes.

I. Introduction

One of the more enduring riddles of the American labor market is that even as women have made dramatic labor market gains, there remains a striking degree of occupational segregation by gender. Women’s progress in the labor market has been significant: their labor force participation rates have increased sharply[1] to within almost fifteen percent of men’s.[2] Nearly half the workforce (forty-six percent) is now female,[3] and the male-female wage gap, although still significant, has decreased.[4]

Yet occupational segregation by gender has persisted, to the great detriment of women. Women remain overrepresented in low-paying and underrepresented in high-paying occupations.[5] “Even within narrowly defined occupations, men and women are often segregated across employers,”[6] with some firms including women and others remaining male-dominated, in fields as wide-ranging as restaurants[7] and major law firms.[8]

*** Top of Page 3 ***

Historically, discrimination and harassment served to maintain traditional “patterns of sex segregation in employment that consigned women to lower-status, lower-paying, female-dominated jobs.”[9] Segregation declined when women first entered the labor force in larger numbers, but “the pace slowed” and in the 1990s “fell by only a percentage point or two.”[10] Surprisingly, although the black-white wage gap is wider than the male-female wage gap, “[o]ccupational segregation appears to be less prevalent by race than by gender . . . . [R]acial occupational dissimilarities are smaller and have fallen faster over time than gender-related ones.”[11]

Perhaps worst of all, there has been virtually no decrease in the gender segregation of particular occupations and jobs. Many occupations are less than five percent or more than ninety-five percent female—mainly blue-collar, mechanical, and other stereotypically “male” work and office assistance or caregiving “female” work, respectively.[12] Segregation is present not only among fields, but also at the micro-level; even within the same workplace, women often are clustered into “pink-collar ghettos” of lower-status departments.[13] In white-collar jobs, the “glass ceiling” remains; women are only a small fraction of law firm partners, executives at publicly traded companies, and tenured professors.[14]

Labor economists and legal analysts alike offer dual explanations for the riddle of continued segregation. First, women may be excluded from

*** Top of Page 4 ***

certain occupations[15] by discriminatory employers and by the resulting disincentive for discriminated-against workers to accumulate education and training. Second, women may have “different preferences . . . or different household responsibilities (particularly related to child care).”[16] The latter explanation may reflect either innate gender differences or “premarket discrimination—differential treatment by parents, schools, and society at large that points girls toward lower-paying (including household) pursuits.”[17] The above explanations demonstrate the central debate: whether segregation is due to discrimination (in employment or childhood) or women’s inherent preferences.

Discrimination and work-family conflict certainly play a role,[18] as does early socialization, though not to as large an extent as is commonly assumed, given the evidence that women’s preferences for different types of work change substantially throughout their careers.[19] However, this Article suggests an important additional reason for the labor market’s halting progress toward equality and the surprising degree of segregation that remains.

It is well-established that women prefer nondiscriminatory workplaces. But for a woman facing a real-world choice among jobs or occupations, it is hard to tell which employers discriminate, especially since recent developments in employment discrimination have mandated many of the employment policies, such as family/parental leave and workplace promises of nondiscrimination, that employers once used to “signal” gender inclusiveness.

One way women can effectuate their preference for nondiscriminatory workplaces is by choosing gender-diverse workplaces—workplaces that feature a mix of men and women.[20] This result may sound obvious, but these criteria (nondiscrimination and gender diversity) are not perfectly overlapping: gender-diverse workplaces may still be affected by gender discrimination, and nondiverse workplaces may not be. Nevertheless, the level of diversity does convey useful information about discrimination. A nondiverse workplace might have developed as a result of gender discrimination, and nondiverse workplaces have certain characteristics inherently disadvantageous to women. Given the history of discrimination and exclusion, the assumption of a connection between gender diversity and better treatment of women is entirely reasonable. In

*** Top of Page 5 ***

economic terms, women rationally use the level of diversity as a proxy for discrimination, which is more difficult to observe. In this light, women’s preference for diversity is “rational discrimination” by employees against employers—a generalization that, though imperfect, is logical given employees’ imperfect information about which employers discriminate.

Worker decisionmaking under conditions of imperfect information is a subject ripe for economic analysis. Yet a preference for diverse workplaces is not incorporated into standard economic analyses of workplace gender disparities. When taken into account, this preference dramatically changes the predictions about the durability of gender disparities. To the extent that women disproportionately choose workplaces already employing a significant number of women, those workplaces become even more inclusive of women. Those without many women will find themselves, over time, lacking female applicants and lagging even further in diversification.

This Article uses several economic models—some quantitative, others qualitative but based on economic theory—that predict that even as more women enter the labor force, the gender gap between diverse and nondiverse occupations (or individual firms) will grow. Just as disturbingly, several models predict that women’s preferences for diversity can lead to enduring segregation even in industries (or individual firms) that do not discriminate. The Title VII era has reduced discrimination and increased female labor force participation—two successes that have been widely viewed as the two-part cure for occupational segregation. This Article sounds a note of pessimism in predicting that neither of these two advances may be sufficient to reduce gender segregation in the labor market. Discrimination does, however, exacerbate the degree and durability of occupational segregation, as this Article’s models illustrate.

The Article then turns to the normative implications of the economic models presented. These models could be interpreted to support a “conservative” argument that a predominantly male workplace might lack women even though it has not discriminated against women and, therefore, that a statistical gender disparity does not necessarily prove discrimination. Further, the ideas presented here could be used to support the argument that it is women’s fault that labor markets remain segregated—a variant of the “lack of interest” argument used by some employers with egregious gender segregation. However, this is a flawed interpretation that ignores the constraints on women’s work choices.

This Article’s observations more forcefully support an entirely contrary, more “liberal” argument: to redress occupational segregation, society must go beyond fighting just what currently constitutes legally actionable discrimination. Since current antidiscrimination laws alone will not abolish segregation given women’s preference for diverse workplaces, further measures to reduce segregation should be taken, including changes

*** Top of Page 6 ***

in the focus and burdens of proof in Title VII cases, affirmative action, and educational reform.

Finally, this Article discusses the theoretical implications of its models for economic analysis of law. Several standard economic assumptions, such as fixed preferences and the efficiency of free market transactions (the Coase Theorem), are undercut by key features of these models. Women’s labor supply preferences are endogenous in that they depend on the level of female labor already being supplied. Labor market evolution thus is path-dependent, hinging on earlier market phenomena. Depending on its history, the labor market can reach multiple equilibrium outcomes, some far more efficient than others. In such situations, there is little basis for assuming that unregulated markets will reach efficient outcomes. The case for a laissez-faire policy of nonintervention in labor markets is therefore weaker, and the case for legal or regulatory intervention stronger, than economic analyses typically recognize.

II. Discrimination as a Deterrent to Workforce Entry

A. The Ability (and Failure) of Most Economic Models To Incorporate Women’s Nonmonetary Preferences

Traditional economics assumes that money drives people to buy or sell goods or services. The basic labor supply graphic model reflects this, showing the direct relationship between wages and workers’ willingness to accept jobs or additional hours.[21] Higher wages certainly are a primary motivation for workers; even people with other job goals will, all else being equal, take the higher-paying job.[22] Wages are not the whole story, of course, and labor economists do not pretend they are: “nonpecuniary factors—such as work environment, risk of injury, personalities of managers, perceptions of fair treatment, and flexibility of work hours—loom larger in employment transactions than they do in markets for commodities.”[23] Indeed, it is a “common misconception[ ] about economics . . . that it slights non-quantifiable costs and benefits.”[24] For example, even bigotry is a nonmonetary preference that economic analyses can consider

*** Top of Page 7 ***

in analyzing labor market decisions; Richard Posner has written that “contact between members of the two races . . . imposes nonpecuniary, but real, costs on those . . . who dislike association with members of the other race.”[25] If economics can incorporate nonmonetary preferences such as distaste for diversity, then it can certainly incorporate preference for diversity.

B. Women’s Preference for Nondiscriminatory Workplaces

1. Women’s Preference for Fair Workplace Treatment: The Basic Economics

Viewed in the broadest sense, men and women have the same workplace preferences—tangible items such as compensation and promotion, as well as intangibles such as fair treatment.[26] This does not mean everyone seeks the same specific ends. Gary Becker has noted that even if people pursue “general ends that remain constant,” they pursue more specific goals “as means [to those] more general ends.”[27] Furthermore, even as people pursue the same universal goals, their specific preferences change based on social circumstances.[28]

Given the deep-rooted gender problems of the modern workforce, including discrimination and harassment, it is entirely rational for female job seekers to consider how various workplaces treat women. That specific preference is a means to the more general, universal preference of fair workplace treatment. Basic economic logic supports such a strategy for female workers: discrimination lowers the expected payoff (monetary or otherwise) of employment, so with less discrimination, women “are likely to be more willing to supply labor at any given wage.”[29]

*** Top of Page 8 ***

2. Effects of Discrimination on Women’s Choices To Enter or Remain in the Workforce

As Vicki Schultz has discussed, women often face what could be called the direct effects of discrimination in traditionally male fields: men stereotype women and try to exclude them from the workplace, and female workers, less likely to be treated well, are more likely to quit.[30] She argues that to the extent that women prioritize family or leisure over work, they do so because they perceive little opportunity for career advancement.[31] In traditionally male fields, male hostility (e.g., harassment, sabotage, denial of training) makes women less likely to succeed.[32] Regardless of differing views as to why discrimination occurs, there remains a “broad range of empirical evidence”—studies of the gender wage gap, testing of employer treatment of identical male and female applicants, and widespread cases of discrimination—“that an important part of the explanation for women’s present labor market position is the continued existence of . . . discrimination against women.”[33]

The problem, however, is not limited to women who directly face discrimination; it also includes the indirect effects of discrimination, which heavily influence female labor force participation. In several ways, discrimination discourages women from entering (or remaining in) the workforce, and particularly from seeking traditionally male jobs.

*** Top of Page 9 ***

a. The Human Capital Problem: Expectation of Discrimination as a Disincentive to Women’s Workforce Participation

Expectation of discrimination is a disincentive to obtaining education or training, or otherwise expending time and money, preparing for a career. This is the cycle of diminished human capital that discrimination causes: “[W]hen discrimination lowers a rational actor’s reasonable career expectations, he or she spends less time, effort, and money amassing human capital, as discrimination lowers the rate of return on that capital.”[34] One cause of discrimination is animus-based bias (e.g., misogynist refusals to hire women). Another is stereotypical employer judgments based on group status (e.g., assuming a certain level of competence for all women in certain positions or assuming that all women will become pregnant at some point and leave their jobs). As Robert Cooter has explained, “[i]f employers attribute to each individual the average productivity of members of the group . . . , each individual will have a tendency to under-invest in acquiring productivity-increasing skills.”[35]

Not only does discrimination lower human capital, it also negatively influences the strategic choices of discriminated-against groups. David Charny and G. Mitu Gulati note that such workers have an increased incentive to engage in various otherwise ill-advised workplace strategies: “high-risk” strategies (e.g., taking on a longshot project) to break out of a stereotype and signal “superstar” status, or “low-effort” strategies (e.g., working fewer hours) because the prospect of being judged by a stereotype rather than by performance makes it less worthwhile to work hard.[36] Though perhaps a rational response to discrimination, these strategies can be self-defeating. Even talented workers may fail at high-risk endeavors, and many talented (or at least competent) workers’ careers will stall due to low-effort strategies.[37] As a result, discriminated-against groups will have fewer successful workers, and “[t]he perception of discrimination on the part of the minority workers becomes a self-fulfilling prophecy.”[38]

Women’s lower levels of workforce participation may also be a rational response to discrimination. As Christine Jolls has noted, when a

*** Top of Page 10 ***

married couple feels the need for one person to suspend or limit workforce participation because of family needs such as childcare, and if “because of sex discrimination . . . [women] will have fewer opportunities and earn less than their male counterparts, then it makes perfect financial sense for the female rather than the male . . . to take the leave or depart from her current employer to provide her better-suited partner with greater career opportunities.”[39] Even absent a higher-earning husband,[40] this effect on women remains because the backdrop of discrimination makes it “natural for aspirations and values to develop such that male employees, more so than female employees, view career advancement as central to their identities and life goals.”[41]

b. Expectation of Discrimination in Particular Fields: Disincentive to Women’s Participation in Traditionally Male Jobs

Expectation of discrimination functions as a disincentive not only to workforce participation generally but also to participation in the particular fields with the most discrimination. The problems in such fields include overt discrimination that seeks to exclude women, harmful psychological effects of stereotyping, and human capital disincentives based on this discrimination and stereotyping.

i. Gender Discrimination in Traditionally Male Jobs

Schultz has noted that “[w]omen in female jobs understand that they will be likely to experience harassment if they attempt to cross the gender divide.”[42] She has written at length about the “proven link between hostile work environment harassment and job segregation.”[43] Specifically, there is a “classic pattern of harassment often directed at women who try to claim male-dominated work as their own” by trying to enter traditionally male fields.[44] Such harassment is aimed at “driving women out of

*** Top of Page 11 ***

nontraditional jobs.”[45] In arguing that the focus of antiharassment law should be the negative effects of harassment on women’s workplace equality, not workplace sexuality, Schultz contends that “[w]here sexual misconduct occurs, it is typically part of a broader pattern of harassment designed to reinforce gender difference and to claim work competence and authority as masculine preserves.”[46] In this light, harassment is “both a cause and consequence of larger forms of gender-based stratification of work, such as job segregation by sex and the accompanying wage and status inequalities.”[47]

Women develop negative expectations about specific fields not only from their own experiences but also from the experiences of others. Shawn Pompian has noted that “there is a persuasive argument that members of minority groups . . . share the unifying feature of a similar experience as the targets of discrimination and oppression.”[48] As a result, “members of excluded minority groups . . . may have heard anecdotal evidence of discrimination from friends and acquaintances within their community” because of the “likelihood that communities develop an awareness of which sorts of opportunities are open to them and which are foreclosed because of discrimination.”[49]

Historical perspective also helps to indicate which employment opportunities are available to women. Because “employers have historically relegated women to low-paid, traditionally female jobs,” women may not aspire to traditionally male jobs; “[m]ost people do not aspire to something they have never been permitted to do.”[50] This history of negative treatment depresses women’s expectations of success; consequently, women may “lower[ ] their work aspirations and turn[ ] their energies elsewhere,”[51] to fields traditionally more welcoming to women or nonwork pursuits.

*** Top of Page 12 ***

ii. Stereotype Threat

When women work in fields that are stereotyped as male, the backdrop of discrimination can interfere with women’s performance as well as their inclination to remain in such jobs. One example of this harm is “stereotype threat”:

[W]hen certain groups of individuals realize that a negative stereotype exists about their group in a given context (e.g., Blacks in academics, females in mathematics), this mere recognition can significantly hinder their performance in that domain. Although the individual does not have to believe the stereotype, he or she will still feel the threat of being evaluated in terms of the stereotype, . . . [a] threat engendered by the fear of confirming the stereotype however false it may be.[52]

Stereotype threat has been shown to affect women in math[53] and “can affect the members of any group about whom a negative stereotype exists.”[54] The first women entering traditionally male fields risk being viewed as “token” females less competent for the job. This tokenism “will increase stress” due to the prospect of “greater scrutiny,”[55] which is why “groups constantly fac[ing] stereotype threat . . . will seek to protect themselves by withdrawing from or ‘disidentifying’ with the affected areas of activity.”[56] The ultimate result is that stereotyping not only can cause group members to underperform but also “can, over the long term, cause . . . reduce[d] participation in activities to which a negative stereotype applies.”[57]

iii. The Human Capital Problem in Particular Fields: General and Specific Human Capital

The “perception of employment discrimination[ ] may act as a barrier to entry,” inhibiting workers from entering those fields in which dis-

*** Top of Page 13 ***

crimination is rampant.[58] This is the human capital problem again, but in a more concentrated form—a disincentive to gaining human capital associated with traditionally male fields. A discriminated-against group will avoid not only specific tests or applications in, but also the educational investments required to enter, fields beset with negative stereotypes of the group.[59] Posner has noted the importance of distinguishing two types of human capital: “general human capital” and “firm-specific human capital.”[60] However, the relative generality or specificity of human capital exists along a spectrum. Certain learning may be applicable to a broad range of jobs (e.g., high school education), only to certain fields (e.g., legal training), or only to a specific job (e.g., experience with a company’s proprietary technology).

The more discrimination there is in a particular field or firm, the less incentive women will have to obtain field-specific or firm-specific human capital, as women will “naturally choose not to invest in education and training that would serve only to prepare them for employment in . . . foreclosed fields.”[61] This is especially true for expensive specific human capital, such as graduate degrees, which are costly and require years of little or no income.[62]

iv. Women Exiting Workplaces That Discriminate

Women are disproportionately likely to exit male-dominated workplaces. Many anecdotal examples exist of women becoming discouraged and leaving traditionally male work.[63] Schultz has gathered considerable

*** Top of Page 14 ***

evidence of this phenomenon, concluding that “women in higher-paying, male-dominated occupations are much less likely to remain . . . than are women in lower-paying female-dominated occupations . . . . Thus, just as employers appear to have begun opening the doors to nontraditional jobs to women, almost as many women have been leaving . . . [as] entering.”[64]

In short, discrimination continues to deprive women of opportunities, and women’s expectations of that discrimination continue this cycle. People typically “adapt their conduct and even their desires to what has been available,”[65] a principle that rebuts the argument that occupational segregation simply results from women’s choices.

As discussed above, discrimination and expectations of discrimination explain a substantial portion of the workplace gender inequities and disparities that continue to exist. The discrimination-plus-expectations story, however, cannot fully explain gender disparities among fields, occupations, or particular employers, because, as the next Part discusses, women often have little information about the extent of discrimination in a given occupation or workplace. Like anyone making an important decision based on incomplete information, women may use, as a proxy for the likelihood of discrimination, other more visible information—specifi-cally, the number of women present in that particular occupation or workplace. As later Parts discuss, such decisionmaking by women makes gender disparities even more durable and self-perpetuating than previously realized.

III. Women’s Preference for Diverse Workplaces: The Economic Rationale

A. Use of Proxies Under Imperfect Information: The Example of “Statistical Discrimination”

Many economic analyses discuss “statistical discrimination”: discrimination not because one dislikes a group but because one thinks its members are more likely to cause certain costs or difficulties, such as employers discriminating against young women (fearing pregnancy leave) or disabled workers (fearing accommodation costs).[66] Better employer in-

*** Top of Page 15 ***

formation (e.g., whether a certain individual with a medical condition will require accommodations) could mitigate employer concerns, but it is often impractical or illegal to obtain such information (e.g., about pregnancy plans) in the hiring process.[67] Even when employers do ask such questions, employees may lie or genuinely may not know certain information in advance (e.g., whether a woman will become pregnant or whether a disability will worsen and then require accommodation). Moreover, employers can invest only so much time and effort in the hiring process; they often “do not have the resources to examine the individual . . . applicants, and instead make judgments based on group affiliations.”[68]

Judgments based on group affiliation risk inaccuracy, yet they may still be profit-maximizing. While complete information yields more accurate decisions, it is costly (or impossible) to obtain—a difficulty that induces employers to resort to statistical discrimination.[69] Employers “economize[ ] on the costs of information”[70] by making decisions based on easily visible characteristics, such as gender or age, rather than harder-to-obtain individualized information. If the visible characteristic correlates reasonably well with whatever cost the employer seeks to avoid, it is profit-maximizing to discriminate.[71] The inaccurate decisions that result from statistical discrimination do impose a cost on employers—the loss of a well-qualified member of the disfavored group. However, that cost may not be great enough to deter employers from discriminating.[72]

Statistical discrimination is an example of how, “[i]n the face of incomplete information on individual applicants, employers rely on their reasoning, intuition, and background knowledge to fill in the blanks.”[73]

*** Top of Page 16 ***

The following Parts discuss this same phenomenon at work in employee decisionmaking: in the face of incomplete information on individual employers, applicants rely on their reasoning, intuition, and background knowledge to fill in the blanks.

B. Reverse Statistical Discrimination: Female Employees’ “Discrimination” Against Employers with an Underrepresention of Women

According to economic theory, an employer offering employees something better than other employers do (e.g., better salary, retirement plan, or work environment) tries to market itself to workers by disclosing, or signaling, that trait. Even if firms with gender discrimination problems try to hire as many women at the same wages as nondiscriminatory firms, Charny and Gulati have reasoned that “employees will migrate towards the firms that signal themselves as fairer.”[74] The question becomes how employers can signal fairness and how women can detect such signals. Current antidiscrimination law has made such signaling increasingly diffi-cult.

1. Female Job Applicants’ Information Problem: The Difficulty of Knowing Which Workplaces Discriminate

a. Limited Employee Resources for Researching and Applying for Jobs

As discussed above, employers engage in statistical discrimination because of several limitations on their information about job applicants. Employers lack (1) unlimited resources to interview all applicants, (2) infi-nite time, money, or ability to research the leading candidates, and (3) the ability to foresee whether better candidates will apply.

Job applicants are in much the same position as employers: they must make employment decisions in the face of limited information. They lack (1) unlimited resources to interview with all employers, (2) infinite time, money, or ability to research the leading candidates, and (3) the ability to foresee when they have a job offer whether better offers will arrive.[75]

*** Top of Page 17 ***

In this sense of having to ration finite job search resources, female job applicants are, ironically enough, in the same position as employers who engage in statistical discrimination. Because of their “limited resources to devote to looking for work,” they have to distinguish, or “sort,” prospective employers to focus their job search efforts.[76] However, applicants are likely to have difficulty sorting based on a preference to avoid discrimination.

b. The Difficulty of Spotting Discrimination

Job applicants are largely unknown quantities to employers, and the reverse is also true. When workers first enter an occupation, they “have much less information about the reputations” of employers than they would like.[77] Absent information on how a new workplace treats women, female applicants may fall back on “knowledge of the pervasiveness of stereotypes and discrimination in society at large.”[78] However, this assumption that workplaces treat women in the same manner as society at large does not help workers distinguish among fields or employers. To make such judgments, workers need to know how one field or employer compares to another.

It is not easy for applicants to discern which workplaces discriminate. Anecdotal evidence may exist,[79] but such information is limited and not a comprehensive examination of an entire occupation or labor market. Anecdotes are also of questionable accuracy because they typically spread slowly and do not include the date of occurrence. As a result, anecdotes may be both too slow in circulating to reflect the latest events and too slow to cease circulating when out of date. Workplaces and occupations can change substantially in their inclusion of women,[80] rendering discrimination anecdotes untrustworthy. Thus, anecdotes are an inadequate way to determine which employers discriminate.

*** Top of Page 18 ***

Another problem with such information is that anecdotes rarely reveal whether discrimination was the true reason behind a woman’s failing to receive a particular job—and workplace segregation makes it harder to unearth the true reason. As Judge J. Skelley Wright noted two decades ago, “[t]he days of Bull Connor are largely past; discrimination now works more subtly.”[81] Because antidiscrimination laws by now are well-known, discrimination has gone underground; “‘employers are rarely so cooperative as to include a notation in the personnel file’ that the firing is for a reason forbidden by law.”[82] Absent such a smoking gun, discrimination is most provable when there are directly comparable male and female candidates whom the employer treated differently.[83] But often there is no one directly comparable. “[T]he more segregated the labor force, the more difficult it is to find comparative evidence illustrating discriminatory, disparate treatment.”[84] As discussed elsewhere, many workplaces and occupations are heavily gender-segregated,[85] so an employer faces “the choice . . . between a male and a female candidate . . . less often than one would think.”[86]

It is especially hard to spot discrimination in certain contexts. In high-wage jobs, evaluations of performance, skill, and qualifications are highly subjective, making it difficult to show that an employer passed up a “better” female or minority worker.[87] And in “high-skill or knowledge intensive jobs in which the discretionary, skill-based component of work creates substantial difficulty in monitoring individual workers’ performance,” employers often solve that difficulty by paying “efficiency wages”[88]— above-market wages intended to generate a large applicant pool and make employees more eager.[89] By design, efficiency wages yield a “surplus of

*** Top of Page 19 ***

equally competent” workers,[90] challenging the idea that free markets deter discrimination. With surplus labor available, discrimination is cheap, both because it is more difficult to spot and because another worker easily can replace the discriminated-against worker.[91]

In cases of discrimination against women based on pregnancy or parental responsibilities, examining employers’ treatment of comparable men is difficult. There is obviously no man directly comparable, and likening pregnancy to a different medical condition is an imperfect comparison.[92] In cases of discrimination against mothers based on parental responsibilities, a comparison could be made to men with similar responsibilities, but men on average still perform far fewer childcare tasks than do women.[93] In sum, job applicants’ knowledge of which employers discriminate is likely to be very limited. Anecdotal knowledge is of questionable accuracy, and it is difficult to know whether a particular termination or failure to hire was based on discrimination.

c. The Lack of Effective Signals of Employer Nondiscrimination: An Unintended Consequence of Judicial Micromanagement of the Workplace

The flip side of discrimination being easy for employers to hide is that nondiscrimination is hard for employers to signal. Employers do not admit discriminating, and “good” employer policies tend to be uninformative recitations of basic law—a peculiar result of increasing legal formalization of equal employment standards. One way for an employer to signal gender fairness would be to publicize its enlightened policies on discrimination, employee leave, and affirmative action. Virtually all major employers have antidiscrimination policies, and while not all have

*** Top of Page 20 ***

maternity leave options or affirmative action measures, those that do often disclose them in job postings.[94]

Increased legal scrutiny of such employment policies, however, has made them a far less useful signal than they once were. Recent Supreme Court decisions have induced all major firms, including those that discriminate, to adopt equal employment opportunity (“EEO”) policies, essentially creating a judicially imposed “minimum” nondiscrimination policy. Recent decisions and legislation have also set “maximum” elements of EEO policies: the Family and Medical Leave Act (“FMLA”) provides a standard employee leave policy under federal law,[95] and courts have sounded a cautious tone regarding affirmative action,[96] imposing restrictions likely to limit employers’ freedom to maintain or expand affirmative action policies. These trends in externally mandated antidiscrimination measures have made it more difficult for employees to infer anything meaningful about a workplace from its policies.

i. Judicially Set “Minimum” EEO Policies: Case Law Allowing Employers To Use EEO Policies as Affirmative Defenses to Discrimination/Harassment Liability

Over the years since the enactment of Title VII,[97] major employers have adopted EEO policies, which typically include: (1) promises of nondiscriminatory treatment, (2) affirmative action policies, (3) discrimina-tion complaint procedures, and (4) employee sensitivity/antidiscrimina-tion training.[98] Increasingly, the Supreme Court has made employers’ Title VII liability turn on the quality of their EEO policies.

*** Top of Page 21 ***

In Faragher v. City of Boca Raton[99] and Burlington Industries v. Ellerth,[100] the Court placed EEO policies at the heart of the employer’s defense in hostile work environment cases. The Court held that an employer is strictly liable for sexual harassment by supervisors unless the employer proves “(a) that [it] exercised reasonable care to prevent and correct promptly any sexually harassing behavior, and (b) that the plaintiff employee unreasonably failed to take advantage of any preventive or corrective opportunities provided by the employer or to avoid harm otherwise.”[101]

Immediately after Faragher and Ellerth, lower courts made clear that an employer goes a long way towards meeting its burden, delineated in the first part of the Faragher test, by simply promulgating an antiharassment policy and publicizing it to all employees (absent evidence of a specific need for other prevention or correction measures).[102] Antidiscrimination training for managers and employees can also help an employer meet its burden, especially to the extent that the training directly evidences implementation of an antiharassment/EEO policy.[103]

The year after Faragher and Ellerth, the Court in Kolstad v. American Dental Ass’n[104] created a similar affirmative defense to punitive damages liability applicable to both discrimination and harassment claims. Kolstad “insulates an employer from punitive damages liability if it has made ‘good faith efforts to enforce an antidiscrimination policy.’ This defense requires an employer to establish both that it had an antidis-

*** Top of Page 22 ***

crimination policy and made [a] good faith effort to enforce it.”[105] Kolstad echoed the Faragher/Ellerth pronouncement that company EEO policies carry great weight in Title VII cases: “The purposes underlying Title VII are similarly advanced where employers are encouraged to adopt antidiscrimination policies and to educate their personnel on Title VII’s prohibitions.”[106]

Thus, the Court has allowed employers to use EEO policies to defuse the significant threat of liability in hostile work environment cases and punitive damages in all kinds of discrimination cases. The Court thereby has increased the incentive for employers to adopt the antidiscrimination policies, trainings, and grievance procedures that are standard fare in corporate EEO policies.[107]

While genuine antidiscrimination efforts certainly benefit employees, Susan Sturm has noted “the risk that employers will adopt legalistic, sham, or symbolic internal processes that leave underlying patterns of bias unchanged.”[108] For example, “[m]any of the internal dispute resolution mechanisms developed by employers . . . consist of boilerplate from the most recent decisions of the court or the reproduction of EEOC guidelines.”[109] Virtually all law firms, for instance, publish identical-sounding EEO policies, with standard promises of equal treatment and assertions of nondiscriminatory decisionmaking.[110] This seems to be enough to protect employers from liability, as “[s]ome courts have deferred to an employer’s procedures, regardless of their actual effective-

*** Top of Page 23 ***

ness in eliminating . . . discrimination. . . . [This] uncritical acceptance of internal dispute resolution processes legitimates purely formalistic solutions . . . .”[111]

In light of this case law, an employer’s tough-sounding EEO policy might signal deep concern for equality, but it might also be simply a tactical decision, adopted on advice of counsel and designed to protect the employer from liability. In fact, the more discrimination at a firm, the more lawsuits it can expect to face, and thus the greater the “return on investment” for paying pricey lawyers to draft an EEO policy satisfying Kolstad and Faragher/Ellerth. The worst discriminators therefore have the greatest incentive to adopt the best-sounding EEO policies. Given the likelihood that the most nondiscriminatory and the most discriminatory firms will have similar-sounding EEO policies, such policies can no longer serve as useful signals of employers’ concern for equality.

ii. Judicially Set “Maximum” EEO Policies: Judicial Proscription of Workplace Affirmative Action

In direct contrast to the minimum EEO policies set by decisions such as Kolstad and Faragher/Ellerth, the limitations courts have imposed on affirmative action[112] can be considered a maximum limit on employers’ EEO policies and efforts. The Supreme Court has narrowed the circumstances in which affirmative action is permissible,[113] and attorneys for

*** Top of Page 24 ***

employers are advising their clients to exercise caution in adopting any affirmative action policies and to review carefully any current programs.[114]

Before the Supreme Court began to restrict affirmative action, employers could send particularly clear signals of concern for equality by engaging in aggressive affirmative action. Affirmative action could increase female and minority hiring by drawing more women or minorities from the applicant pool, but it could also have dynamic effects. Affirmative action can induce more women and minorities to apply because it is “a valuable means of signaling to potential applicants” the employer’s “high value on diversity.”[115] Thus, as voluntary affirmative action becomes rarer due to the Supreme Court’s restrictions, another formerly useful signal to job applicants concerned about gender discrimination is vanishing.

The irony here is that the judicial declarations of minimum and maximum EEO policies are purportedly conservative policies, in the sense of limiting employers’ liability for inequality and responsibility to redress it. Yet these precedents heavily regulate and distort employers’ decisions about how to run their workplaces. Where there once existed a wide range of antidiscrimination and affirmative action policies, there is now a judicially imposed set of workplace rules. While serving the conservative goal of minimizing employer liability, these decisions fly in the face of the older conservative exhortation that “[c]ourts are generally less competent than employers to restructure business practices, and unless mandated to do so by Congress they should not attempt it.”[116] Indeed, when courts reject employment discrimination claims, they regularly invoke just these sorts of old bromides, such as the admonition that courts not serve as “super-personnel departments,” monitoring employment decisionmaking.[117] When constructing liability shields for employers, how-

*** Top of Page 25 ***

ever, many courts remain aggressive micromanagers unconcerned about judicial intrusion into workplace policy.

iii. Legislation Setting a Standard Package of Accommodations

Oddly, employers’ abilities to signal their fairness to job applicants are diminished not only by judicial decisions limiting employer responsibility for workplace inequality, but also by legislative enactments increasing employer responsibility for workplace inequality. The FMLA, for example, mandates up to twelve weeks of unpaid leave for an employee who has a “serious medical condition” or is caring for a relative with such a condition.[118] The FMLA covers both men and women but was expressly conceived as an effort to force employers to accommodate women in the workplace[119] by allowing them to return to jobs after taking time off for pregnancy, caregiving, and homemaking responsibilities that women, even when they also work outside the home, continue to undertake more often than men.[120]

The FMLA, however, may not have raised the floor of employee leave benefits for some employees, and it may have set an unofficial ceiling. Michael Selmi has noted that before the FMLA, employee leave benefits were common but quite varied; federal survey data show that “a majority of large employers . . . provided some form of leave before the passage of the FMLA . . . to take time off around the birth of a child.”[121] The most common forms were paid sick leave, unpaid maternity leave, and disability plan coverage for pregnancy.[122] Some firms offered benefits well above the FMLA minimum: employees working in medium- to large-size companies for at least five years received an average of nine weeks of fully paid leave.[123] In addition, some states mandated leave benefits more generous than the FMLA’s, including wage replacement during leave.[124]

*** Top of Page 26 ***

Thus, before the FMLA, a firm’s generous family/pregnancy leave provisions could serve as a useful signal of genuine efforts to facilitate the entry and retention of female employees. The FMLA’s passage, however, raised the less sensitive employers’ leave policies up to a statutory minimum. As a result, such firms’ inclination to offer stingy family/preg-nancy leave is no longer readily apparent, eliminating another potential signal to workers concerned about gender equity.

While the FMLA increased family/medical leave rights in some workplaces, it

may also have preempted the development of more friendly policies in the private sector. Prior to the . . . FMLA, there appeared to be a modest trend by private employers to provide increasingly generous paid leave policies, a trend that now seems to have stalled as the federal legislation has become the ceiling of benefits.[125]

In fact, the enactment of the FMLA has resulted in family leave reform vanishing from legislatures’ and unions’ agendas.[126] Except for California, which in 2002 passed a law mandating paid leave, no state has passed significant legislation since the FMLA’s enactment, and the relevant potential lobbies have been “likely to move on to other issues.”[127]

To the extent that it has preempted or slowed any trend by firms, unions, or states to provide more generous leave, the FMLA has in effect imposed an unofficial ceiling limiting what many employers might come to provide. By creating a mandatory floor and an unofficial ceiling on employee leave benefits, the FMLA has diminished the range of variation among firms’ leave policies, thereby lessening such policies’ value as signaling mechanisms.

iv. Employer Signals Reduced to “Cheap Talk”

As discussed above, employers’ EEO, affirmative action, and family/medical leave policies no longer serve as effective signals of employer fairness. The resulting information problem for employees is similar to a long-noted information problem for employers: because employee talent and effort is hard to discern in advance, employers end up skeptical of the “large number of noisy and false signals”[128] from employees. The result is a “pooling equilibrium”:[129] even though employees vary, employ-

*** Top of Page 27 ***

ers assume each new employee to have about the same talent and work ethic as the average worker in the labor pool.[130]

Similarly, employer policies have degenerated from useful employer signaling to a form of “cheap talk”—boasts that are “profitable because of the low odds of being caught and the immense gain in respectability (and legality)” of claiming fairness.[131] Employee assessments of employers’ fairness have, as a result, degenerated into a pooling equilibrium, with employees assuming that one employer has about the same odds of discriminating as another.

2. The Solution to Women’s Information Problem: Using the Percentage of Women in a Workplace as a Proxy for Employers’ Treatment of Women

As previously discussed, in recent years both overt discrimination and variation among employers’ EEO, affirmative action, and employee leave policies have decreased, reducing their signaling value to employees. Thus, women must utilize other criteria to predict which occupations or particular employers are more likely to treat female employees well.

One such criterion that female job applicants can employ is the percentage of a workforce that is female[132]—a figure that can serve as a proxy for the presence of discrimination. In several ways, the number of women in a workplace correlates to the likelihood of gender discrimination. First, and most obviously, discriminatory hiring and promotion practices may be the reason there are few women in a given workplace. Second, hostile work environments are far more likely in workplaces with fewer women. And third, even absent conscious discrimination or hostility, workplaces with fewer women are more likely to feature gendered norms favoring men and to advantage men in other subtle but important ways.

*** Top of Page 28 ***

a. The Rationality of the Assumption That Discrimination May Have Caused the Underrepresentation of Women

Female job applicants may see gender disparities among workplaces, but they may not know why such disparities exist. One reason could be that the firms with fewer women discriminated against women in hiring and promotions. This commonsense point is reflected in the case law viewing significant underrepresentations of women as prima facie evidence of discrimination.[133] Another possibility is that because women still bear disproportionate childcare responsibilities,[134] they opt against certain work requiring extremely long hours.[135] Joan Williams argues that failing to accommodate childcare responsibilities may be a form of discrimination;[136] yet even if it does not constitute judicially recognized “discrimination,” it is still a form of inhospitability to women.

Another possible cause of workplace gender disparities is unequal attrition, even absent any hiring discrimination. If an employer hires equal numbers of men and women, but women have higher attrition rates, that workforce will become male-dominated over time. Hostile work environments and other intangible discriminatory acts, such as discriminatory evaluations, are one reason for higher female attrition; in male-dominated workplaces, hostility to female workers may subvert women’s performance or otherwise induce women to leave.[137]

Gender disparities also may arise from random chance or nondiscriminatory causes, but applicants may not know whether a specific underrepresentation resulted from discrimination. Yet because discrimination is a possible cause, it is rational, given their uncertainty, for women to avoid workplaces with underrepresentations of women.

b. The Reverse of the Basic Assumption That Discrimination Causes Disparities: Disparities Cause Discrimination

When a gender disparity exists, women may rationally assume that discrimination caused the disparity. Yet the opposite is true as well: a lack of women in the workplace can cause, or at least facilitate, discrimination. In predominantly male workplaces, stereotyping women as incapable of performing “men’s work” is more frequent, hostile work environments are more common, and subjective, informal decisionmaking favoring men over women is more likely. To the extent that gender dis-

*** Top of Page 29 ***

parities cause discrimination, avoiding male-dominated workplaces is a rational strategy for female job applicants.

i. The Ease of Stereotyping in Traditionally Male Jobs and the “Lack of Interest” Defense

When states prohibited women from practicing law, it was easy for the Supreme Court to reject a Fourteenth Amendment challenge to the Illinois ban on women attorneys because “[t]he paramount destiny and mission of woman are to fulfil the noble and benign offices of wife and mother. This is the law of the Creator. And the rules of civil society must be adapted to the general constitution of things, and cannot be based upon exceptional cases.”[138] The Illinois law and other similar state laws prevented any women from becoming counterexamples and thus from disproving that stereotype. Many have viewed stereotyping exactly this way—as a result of inaccurate information that lessens in diverse settings, where those holding inaccurate stereotypes are exposed to counterexamples.[139]

Economists tend to view discriminatory preferences—distaste for other groups—as a distinct phenomenon from stereotyping, and they assume tastes are fixed, not the product of current conditions.[140] An alternative view of discriminatory preferences is that they are a product of living in a nondiverse environment and can change as the environment changes. Noneconomic analyses long have noted that desegregation decreases prejudice.[141] Even economic analysis is capable of analyzing bigotry as an “information market failure[ ].” Under this view, people’s perceptions are skewed because they receive and selectively process biased information.[142] These biased perceptions can then produce and reinforce bigoted

*** Top of Page 30 ***

views.[143] Thus, a lack of diversity may reinforce not only inaccurate stereotypes but pure bigotry as well.

One example of disparities facilitating stereotypes is the success of the “lack of interest” defense used by employers charged with discrimination. Schultz has collected examples of courts adopting stereotypes that (in the words of one federal judge) “certain work . . . is not attractive to females. This is a fact of life that an Act of Congress cannot overcome.”[144] Numerous courts have accepted employers’ arguments that women are underrepresented in higher-paying work not because of discrimination but because women lack interest in traditionally male jobs such as retail sales,[145] bakery operations,[146] and manufacturing.[147] In the leading “lack of interest” case, EEOC v. Sears, Roebuck & Co.,[148] the court accepted the employer’s assertions “that the gendered characteristics Sears ascribed to the commission sales position were an inherent, necessary part of the job.”[149] Sears’s male-biased criteria reflected the characteristics of the existing male-dominated workforce; the absence of women from the workforce meant that they would be disfavored for not matching the prevailing archetype of a commissioned salesman.

Even partial exclusion facilitates stereotyping by allowing exceptions to be written off as special cases, as in the Supreme Court decision upholding a state’s ban on women practicing law.[150] Furthermore, due to stereotype threat,[151] it is hard for a woman in a male-dominated workplace to excel as an “exception,” further decreasing the likelihood of dispelling stereotypes in such a workplace.

*** Top of Page 31 ***

ii. The Prevalence of Hostile Work Environments in Traditionally Male Jobs

The fewer women in a workplace, the more likely it is that the work environment will be hostile to women. The problem is greatest for women entering male-dominated occupations, as Schultz has noted.[152] The hostility often “is neither driven by the desire for sexual relations nor even sexual in content”; rather, it is “designed to maintain work—particularly the more highly rewarded lines of work—as bastions of masculine competence and authority.”[153] Harassment thus can stem from men’s desire to preserve male-dominated fields by both driving out women who enter (with hostility, sabotage, etc.) and marking the work as male territory (with sexual images and indicia of male-dominated culture).

Michelle Adams has elaborated on the general principle of social psychology that dominant-group members “often allocate resources to those within their group and discriminate against outgroup members because to do so contributes to their own individual self-conception.”[154] Social psychology theories such as “social identity theory” emphasize the tendency to favor one’s own group, which helps dominant groups perpetuate their power.[155] Other theories focus on darker motivations, such as “realistic-group-conflict theory” (i.e., groups have incompatible goals and compete for scarce resources)[156] and “social dominance theory” (i.e., groups aim to protect their privilege, power, and prestige).[157] Despite their differences in focus, each of these theories helps explain the prevalence of hostility to women in male-dominated workplaces.

High female attrition is the predictable result of hostility to women in predominantly male workplaces.[158] As Schultz has described the phenomenon, “harassment is . . . driving the small number of women in nontraditional jobs away.”[159] Like other discrimination, harassment is a “powerful disincentive for women to move into and remain in nontraditional occupations.”[160] Thus, even if a gender disparity did not result from discrimination, it can lead to discriminatory outcomes.

A strategy of gender-based hostility aimed at excluding women is more feasible and more likely to succeed where women are already underrepresented, sometimes even despite the presence of women in positions of authority. Presumably, women with authority would be less likely to

*** Top of Page 32 ***

tolerate gender discrimination and harassment. However, in male-dominated workplaces, the few women who make it to the top may spend little effort improving conditions for future women.[161]

iii. Subjective and Informal Decisionmaking That Favors Men

An additional problem in male-dominated workplaces is what might be called “soft” discrimination—social networks favoring men and other dominant groups in a workplace. Specifically, selection criteria that tend to favor men, as well as word-of-mouth hiring and other informal decisionmaking processes, reinforce gender disparities even absent any intent to discriminate.[162] The classic example is the interviewer who begins job interviews with questions aimed at determining whether the interviewer and the applicant know people in common. Interviewers ask such questions, in the words of one interviewer at a major hiring conference, “to see if [the applicants are] the same species” as the incumbents.[163]

*** Top of Page 33 ***

Male decisionmakers may tend to select job applicants and employees to mentor based on preexisting contacts and various commonalities with those employees that correlate with gender. This has been the experience of women in the legal profession, for example, where “[w]omen are often not mentored, and are frequently not included in important informal and social contacts. This is particularly true in firms with few women.”[164] Similarly, white dominance is reinforced, Daria Roithmayr has noted, by selection criteria that correlate with race, such as standardized testing, cultural experiences, and informal contacts. Decisionmakers perceive these measures to have nondiscriminatory utility as reasonably good selection criteria—at least good enough to outweigh the “switching costs” of changing the process.[165]

In the context of gender, the problem is not so much tangible selection criteria, such as standardized testing and educational attainment, but the subjectivity and informality of many employment decisions. Choices among candidates ultimately entail a great deal of discretion—especially to the extent that there is a surplus of similarly qualified candidates, whether because of a depressed local economy or above-market efficiency wages attracting a deep applicant pool.[166] Once employees are on the job, a substantial amount of training and mentoring is informal. Schultz has noted that this is true of “virtually all training” in traditionally male blue-collar jobs,[167] and it is just as true in many white-collar fields as well.[168]

*** Top of Page 34 ***

Not only does this informality facilitate actual discrimination, it also favors workers who happen to share natural ties and affinities with existing decisionmakers. Social contacts are invaluable in many fields as a way “to form relationships with powerful incumbents”;[169] in fact, part of the benefit of an education at an elite institution is developing crucial professional contacts.[170] Shared “cultural knowledge” is often important for workers hoping to bond with, and be viewed as successful by, those incumbents.[171]

These cultural advantages are usually not considered discrimination within the limited definitions of current antidiscrimination laws. An employer has not violated Title VII, for example, because an employee gained informal knowledge and contacts by belonging to the same country club as the employer’s white male officers, or because the white male employee’s upbringing works to his advantage in interviews with (or workplace schmoozing of) white male decisionmakers. However, the effect is much the same as overt discrimination: the in-group worker gains very real advantages for reasons entirely divorced from the worker’s merit as an employee. In short, subjective and informal decisionmaking leaves in place the “old boys’ network” of male decisionmakers selecting and grooming male employees. This is an illustration of the prediction of social identity theory:[172] whether consciously discriminating or not, empowered groups will maintain their power by favoring their own members.

The problem is a circular one because the relevant contacts and cultural knowledge come in large part from the workplace, but the lack of those contacts and knowledge is a barrier to workplace entry and success. If the entire workforce, including workplace decisionmakers, were evenly

*** Top of Page 35 ***

divided between men and women, men would possess no advantage in contacts or culture; even if contacts and culture split somewhat along gendered lines, new female workers would have just as many ties to female decisionmakers as new male workers have with male decisionmakers.[173] Because fewer decisionmakers are female, however,[174] new male workers have contacts and cultural advantages that would not exist absent current gender disparities.

IV. A Dynamic Model of Preferences for Diversity: Initial Integration Fails To Eliminate Segregation

This Part sets out a model of how women’s preference for diversity can lead to increased occupational segregation, even as more women enter the labor force. Though nonquantitative, this model is deeply rooted in economics; its premise is rational, strategic decisionmaking, and it presents a dynamic analysis that considers how one change causes subsequent adjustments, and so on.[175] In labor economics, dynamic analyses can yield surprising yet accurate predictions because they consider how changed circumstances affect workers’ expectations and thus affect whether or not they enter a field or particular workplace.[176]

The model in this Part assumes that the labor market entirely excludes women at first. Then, only certain fields, and certain specific firms,

*** Top of Page 36 ***

start including women. The employers that begin including women will find themselves with more female job applicants, beginning a dynamic process yielding a counterintuitive result: even as more women enter the workforce, making the overall workforce more integrated, occupational segregation increases because the firms initially lagging in integration never catch up, even if they do not discriminate.

A. Initial Conditions: Complete Segregation Followed by Uneven Initial Integration

In this model, at first no women participated in any fields traditionally considered “men’s work.” Then integration began, but not at the same pace in all fields or firms. Initial integration focused only on certain industries or certain job types for several reasons. First, some firms or industries may be more welcoming of women because “[i]n a market of many sellers the intensity of the prejudice . . . will vary.”[177] Second, some jobs may conflict less with long-standing stereotypes of women’s proper role (e.g., the wide range of caregiving occupations that are more than ninety-five percent female, in contrast to the physical labor occupations that are more than ninety-five percent male).[178] Third, some firms may be more sensitive to, or successful at, recruiting and including women because of, for example, overtly or subtly “gendered” job postings.[179] Fourth, some firms may be earlier targets of lawsuits or public scrutiny of their exclusionary practices. For example, many law firms faced sex discrimination suits in the 1980s,[180] but many major financial institutions did not face similar suits until the late 1990s and early 2000s.[181] Fifth, some male-dominated fields, more dependent on a steady influx of men in their early twenties than others, experienced war-induced labor shortages of

*** Top of Page 37 ***

men, leading them to bring women into the fold as a matter of economic necessity.[182] Random chance probably also explains some uneven integration; even if all firms face the same stereotypes and incentives, one firm or industry may still find itself with more female workers.[183]

B. Widening of Initial Gender Gaps over Time

Once there is an initially uneven pace of integration, the gender gap may widen due to various forces of labor supply and demand.[184] Specifi-cally, the level of gender diversity affects both female workers’ incentives to enter a workplace and employers’ incentives to hire women.

1. Labor Supply: Women’s Incentives To Enter a Workplace Increase as the Representation of Women Increases

a. The Preference for Diversity: Variation in Intensity of Preference

Women prefer, all else being equal, to work for employers with more female workers,[185] at least when workplaces range from male-dominated to roughly balanced.[186] The model becomes a bit more complicated, but

*** Top of Page 38 ***

more realistic, if it allows for varied intensity of preferences, a widely recognized element of labor supply models. For example, some workers value high wages above all else, whereas others place more value on reasonable hours.

Some women will have an unusually strong preference for a diverse workplace—e.g., those who have already experienced discrimination, or those who are most risk averse (such as women who are their families’ primary wage earners). On the other end of the spectrum, some women will have little or no preference for diversity—e.g., those who have never experienced discrimination, those who are least risk-averse, or those who, because they have no personal or professional ties to the community, least fear having to switch jobs.

b. Widening of Small Initial Gender Gap Because of Sorting by Intensity of Preference for Diversity

At the point of initial diversification, when firms first start hiring women, female workers will tend to sort themselves by the strength of their preference for diversity. Women with little or no preference for diversity may choose jobs without regard for diversity, while those with stronger preferences for diversity will focus on the firms with somewhat higher female representation.

At first, the gender gap is modest. Only those women with the strongest preference for diversity—“Type I women”—will make job decisions based on a very small initial gender gap, such as a 2% gap, where one field has 10% women and another has 12%. Whereas the Type I women will disproportionately choose the 12% female workplace, most other women will not notice or care about such a small difference and will make job choices based on other criteria.

Because the Type I women disproportionately enter the slightly more diverse workplace, the gender gap widens a bit—perhaps from a 2% to a 5% gap between workplaces. Because this gender gap is wider, it becomes noticeable and a subject of concern to somewhat more women—specifically, women who are quite concerned about the diversity of their workplace, though not so much as the Type I women. Call these the “Type II women.” Most women still may not care about this relatively small gender gap; now, however, the gap is big enough to influence not only the Type I women but also the Type II women. Accordingly, the Type I and Type II women in the job market will disproportionately choose the more diverse workplace.

The process continues: with the Type I and Type II women choosing the field with 5% more women, the gap grows—perhaps to 10%, perhaps larger. This larger gap will start to influence additional women who pre-

*** Top of Page 39 ***

fer more diverse workplaces, though not as strongly as the Type I and Type II women. This next group—“Type III women”—will start disproportionately choosing the more diverse workplace, making it still more diverse.

Thus, an initially modest gender gap widens due to positive feedback resulting from a sorting process. Those with the strongest preference for diversity flock to slightly more diverse firms. Because this widens the gender gap, more women are drawn to the more diverse workplaces, making them increasingly diverse in comparison to other firms. Formally, there is positive feedback because an increase in X (the gender gap) leads to an increase in Y (female interest in the more diverse workplace), and the reverse is true as well: an increase in Y leads to an increase in X. Thus, an initial increase in X (which was initially modest) leads to snowballing increases in both X and Y—rapidly increasing diversity in the initially slightly more diverse workplaces and growing gender gaps.

c. Widening Gender Gaps as a Form of “Female Flight”: The Reverse of “White Flight” Residential Segregation

This model of widening gender gaps—an initial gap widening due to sorting by intensity of preference for diversity—is a logical reversal of Professor Thomas Schelling’s “tipping” model of the “white flight” that yields residential racial segregation. As Adams has summarized:

Schelling’s tipping model presumes that white residents have varying levels of racial tolerance. Thus, when the first black person enters a neighborhood, only the most intolerant whites depart. Then, those vacancies are filled by more blacks, causing those whites that are only marginally tolerant to leave. This phenomenon repeats creating a tipping effect that drives more and more whites out . . . .[187]

The key point of the Schelling’s “tipping” model, as Christina Ho has explained, is that

start[ing] with the proposition that a population . . . will exhibit a wide range of preferences about racial makeup[,] . . . Schelling offers an account of how any binary population can manage repeatedly to generate the same highly segregated residential patterns . . . . [E]ven a small perturbation in composition, such

*** Top of Page 40 ***

as one caused by a desegregation order, can operate in Schelling’s model to produce “tipping” . . . .[188]

Schelling’s analysis is generalizable to gender segregation in the workplace.[189] Both in Schelling’s “white flight” model and in this “female flight” model, the theory can be generalized to four elements. Where a market features (1) an initially modest demographic “gap” (whatever the cause) and (2) varied preferences for diversity, then (3) a dynamic positive feedback process of sorting by intensity of preference will occur, resulting in (4) a widening demographic gap.

2. Labor Demand: Employers’ Incentives To Hire Women Increase as the Representation of Women Increases

As discussed above, a gender gap can widen in a dynamic process of positive feedback on the supply side of the labor market. On the demand side, a gender gap can increase for different reasons, but with the same result: an initially small gender gap enlarges due to market pressures that yield positive feedback.

a. Decreasing Sexism as a Workforce Includes More Women

An underrepresentation of women facilitates discrimination, hostile work environments, and a general lack of equal opportunity by making inaccurate stereotypes of (and bigotry against) women more likely and hostile environments easier to effectuate, while allowing men to enjoy network benefits based on the dominance of male decisionmakers.[190] Whether arising from inaccurate negative stereotypes, pure bigotry, or a desire to accommodate the biases of customers or employees, an aversion to hiring women is, in labor-economic terms, a decreased employer demand for female labor.[191] As a workplace becomes more diverse, these aversions decrease, which means employers’ demand for female labor increases.

*** Top of Page 41 ***

b. Search Costs: Advantages in Outreach for Workplaces with More Women

Labor demand is partly based on recruiting expense. An employer’s labor costs are lower, and labor demand therefore higher, when recruiting is easier, cheaper, and more successful. Workplaces with more women may find it easier to engage in the sort of efforts that firms often undertake to recruit female candidates, such as having female employees make campus visits, conduct interviews, and perform other recruiting work.

Men can recruit women, of course, but having female employees participating in recruiting may make it more successful. Typical word-of-mouth recruiting benefits those groups already represented in the workplace, usually men, and signals a lack of openness to employees in underrepresented groups, such as women and minorities.[192] Yet the reverse also is possible: recruiting by women can signal to applicants that the firm values female employees, especially when female employees participate in recruiting events expressly designed for women or minorities. Female employees may be more likely to know, through personal connections, other female applicants or places to find female applicants. Just as word-of-mouth recruiting perpetuates the status quo where the decisionmakers are white men, women’s involvement in recruiting can change that status quo. Moreover, female employees may have an advantage in communicating with female applicants. They can speak with more credibility about how women are treated at the firm, and they might be better able to phrase advertisements, postings, and recruiting speeches in ways that are more appealing to women, which can have a substantial impact on whether women apply for jobs.[193]

A firm need not have a very high percentage of women to reap these benefits, but a firm with a nontrivial representation of women will find it easier to recruit women successfully than a firm with fewer women. Consequently, firms with more women will have a higher demand for female labor because they will find it easier to find and procure female workers.

c. Startup Costs of Integration

There are certain costs of integration, including: (1) inducing or forcing incumbent male employees to allow women equal opportunity, (2) adapting firm policies to address issues that have not arisen before (e.g., the medical leave policy’s coverage of pregnancy/maternity leave), and (3) creating new policies necessitated by workforce diversity (e.g., harassment reporting and antidiscrimination training). Many of these

*** Top of Page 42 ***

costs are fixed costs,[194] varying little, if at all, with the number of employees or size of the employer. Once a firm begins to diversify, it is likely to adopt an antiharassment policy and set up antidiscrimination training. After it does, that cost does not increase as it further diversifies.[195] Once a firm incurs these costs, however, it can employ the newly included workers more cheaply than its competitors who have not yet borne the same costs. A firm that has already hired many women and developed antiharassment, antidiscrimination, and pregnancy leave policies can hire more women without again bearing the expense of transforming its workplace.

In contrast, a firm that has not yet diversified still faces these initial fixed costs of integration. “Many companies are unwilling to incur the short-term costs of developing effective human resource practices and addressing cultural conflict . . . . Moreover, smaller companies often lack the resources to invest adequately in improving their human resource practices.”[196] These are fixed costs that, before a firm pays them, are a disincentive to hiring and creating a work environment palatable to the workers in question. Thus, the cost of hiring women is lower, and labor demand for women is higher, for firms that have already adapted their workplaces as necessary.

V. The Quantitative Models: When Women Prefer Diversity and Diversification Is Uneven, Gender Gaps Remain Despite Overall Diversification

This Part sets forth more detailed quantitative models of how women’s rational preferences for diverse workplaces can lead to durable gender segregation.

A. The Three Quantitative Models

This subpart sets out three quantitative models, each simulating the labor market under different assumptions. Each model is based on a labor market characterized by initial exclusion of women, slightly uneven initial integration, and a female preference for gender-diverse workplaces. The first model assumes that all women prefer diversity. The second model assumes that only some fraction of women have that preference. The third model expands on the second model by incorporating the ef-

*** Top of Page 43 ***

fects of discrimination, especially in the more male-dominated workplaces.

In short, the models in this Part predict that labor market segregation will arise, and persist, as a result of three simple conditions: (1) an initial state of near-complete exclusion of women from most occupations (whether due to social pressures, government regulation, or both), (2) early gender integration that is uneven among different occupations and different firms, some becoming slightly diverse before others (whether due to random fluctuation or meaningful differences among occupations, firms, or regions), and (3) forces of labor supply and demand making diversification cheaper, easier, and more rewarding in the initially more diverse workplaces, as well as relatively costlier, harder, and less rewarding in the initially less diverse workplaces. Given these conditions, a small gender gap in the early days of integration will widen over time—even as more women enter the labor market, and even without actual discrimination, because women will disproportionately choose more gender-diverse workplaces. Ultimately, the labor market will display a striking contrast of progress and stagnation—increasing female labor force participation and increasing occupational segregation. Discrimination, of course, will widen the gender gap further.

Also strikingly, although the gender gap widens because of women’s preferences for more gender-balanced workplaces, it results in a degree of segregation that is suboptimal from women’s own perspective. Even as women come to be almost half of the overall labor force, many workplaces will not be at all gender-balanced: some workplaces will become majority-female, while others will remain male-dominated. Ironically, women’s rational preference for diversity yields less gender-balanced workplaces than we might expect in a labor force that is fairly gender-balanced in the aggregate. This lack of gender balance among workplaces, moreover, leaves the overall labor force slightly less than evenly gender-balanced. Even if equal numbers of men and women want to work, the persistence of segregation discourages some women from the labor force.

1. The First Model: All Women Prefer Gender-Diverse Workplaces

This basic model examines a labor market in which all women have the same preference for a diverse workplace, unlike the qualitative model in the previous Part (in which women had a spectrum of preferences) and unlike the models discussed later in this Part (which create more realism by relaxing the assumption of identical preferences). I make eight assumptions, the first two of which set out the basic theory.

Assumption #1: Women prefer diversity. All else being equal, women prefer workplaces with more women, at least to the extent that workforces

*** Top of Page 44 ***

well below 50% women are less appealing than workforces that are closer to 50%.[197]

Assumption #2: Initially slight and uneven integration. One firm or sector hires a few women first—here, 10% for mathematical ease.[198]

The third, fourth, and fifth assumptions rule out other causes of gender disparities, such as discrimination and women being discouraged from the labor force. These other causes surely exist in the real world, but this model examines how much of the gender gap can be explained solely by the theory at hand—that an initially small gender gap and women’s preference for diversity will widen the gender gap even absent actual discrimination.

Assumption #3: Men have no diversity preference. Men have no preference as to the presence of women in the workplace—i.e., male applicants do not seek workplaces that are predominately male or ones that are gender-balanced.

Assumption #4: Nondiscriminatory hiring and attrition. Each firm/sector hires in proportion to the population and suffers attrition in proportion to its workforce—i.e., no discrimination causes any part of the gender gap, and less diverse firms do not find it harder to retain female employees.

Assumption #5: 50-50 split of job-seekers. Women and men apply for jobs in equal numbers—i.e., at the point when they apply for jobs, women are not discouraged from entering the labor force as a whole.

Importantly, we would see more of a gender gap, not less, to the extent that the third, fourth, and fifth assumptions are incorrect. Thus, making these assumptions puts the theory of women’s diversity preferences to the

*** Top of Page 45 ***

test: can it explain gender gaps if the other likely causes of such gaps are excluded?

The sixth, seventh, and eighth assumptions are basically model specifications that should not meaningfully affect the ultimate conclusions of the analysis.

Assumption #6: Two relevant firms/sectors, denoted “Alpha” and “Beta,” each with (for ease of math) 100 workers. Critically, the analysis is identical regardless of its scale. Alpha and Beta may be two different sectors of the economy, such as heavy industry and retail sales; two occupations, such as teaching and law; two individual companies, such as two particular law firms; or two departments in the same employer, such as a firm’s sales division and its human resources division.

Assumption #7: Workers apply for only one job. Applicants choose which of the two employers—Alpha or Beta—to which to apply. This is not as unrealistic as it first sounds. In a broader model, with many employers, this assumption would be merely that applicants cannot apply to all employers/sectors simultaneously but instead must focus their efforts on one sector, one job type, or a limited number of employers.

Assumption #8: 100 employees, 10% attrition each year. At both Alpha and Beta, hiring recurs each year to replace the ten departing employees.

a. The Initial State of the Labor Market and Breakdown of Job Applicants

With few women in the labor market, Alpha has a workforce of 10% women (assumption #2), while Beta has no women. This is the relevant initial state.

Table 1.1: Initial State of Labor Market (First Model)

Workplace or sector

% women
in workforce

% men
in workforce

Alpha

10% (10 of 100)

90% (90 of 100)

Beta

0% (0 of 100)

100% (100 of 100)

Each year, X men and X women apply for jobs in the labor market in which the only workplaces are Alpha and Beta, since under assumption #5 men and women enter and participate in the labor force in equal numbers. Because the men are indifferent to the level of gender diversity in their workforce (assumption #3), they apply in equal numbers to Alpha and Beta, each of which receives X/2 male applicants. Women, however, prefer the more gender-diverse employer (assumption #1)—here, Alpha.

*** Top of Page 46 ***

Thus, women apply only to Alpha, which receives X female applicants; Beta, because it is the less diverse employer, receives no female applicants. The result of these conditions is that while Beta hires from a pool of only male applicants, Alpha hires from a mixed, but disproportionately female, pool of applicants.

Table 1.2: Annual Job Applications to Alpha and Beta
(First Model)

Job applications each year

# applying to Alpha

# applying to Beta

Of the X men

X/2

X/2

Of the X women

X

0

Total (men & women)

X women, X/2 men

0 women, X/2 men

Per 10 slots

6.7 women,
3.3 men

0 women,
10 men

Because both Alpha and Beta engage in nondiscriminatory hiring (assumption #4), each hires in proportion to its applicant pool. The last row, above, illustrates this: Alpha’s applicant pool is two-thirds women, so for every 10 slots, it can be expected to hire 6.7 women and 3.3 men.[199] Beta’s applicant pool is all male, so it hires 0 women and 10 men.

b. The Evolution of Diversity at Alpha and Beta

Each year, 10 employees leave Alpha and 10 leave Beta (assumption #8). Because the model assumes nondiscriminatory attrition (assumption #4), employees depart in proportion to their population in each workforce. So in the first year, when Alpha is 10% female, 10% of the departing employees are women—i.e., of the 10 departing, 1 is female and 9 are male. At Beta, where all employees are male, all 10 departures are men.

The new employees change the female population at Alpha. Of Alpha’s 10 new employees, 6.7 are women, so in the first year, Alpha loses 1 woman (from attrition) but gains 6.7 (from hiring). Beta loses no women and gains no women. At the end of the first year, then, Alpha has a net gain of 5.7 women while Beta has no net gain.

*** Top of Page 47 ***

Table 1.3: Female Population Resulting from Attrition and Replacement—