GOLDSCHEID 10/28/2011 10:31 AM [33] JULIE GOLDSCHEID* Disparate Impact’s Impact: The Gender Violence Lens I. Introduction ............................................................................. 34 A. Disparate Impact: A Theory of Limits or Untapped Potential? .......................................................................... 36 B. Context: Domestic Violence ............................................ 38 II. Disparate Impact: Testing Theory and Doctrine ..................... 42 A. Theory .............................................................................. 42 B. Doctrine ............................................................................ 45 1. Statutory Requirements .............................................. 45 2. “Single Decision” Cases ............................................. 48 C. Establishing Impact .......................................................... 59 D. Impact of Impact Analysis ............................................... 65 Conclusion .......................................................................................... 66 Although disparate impact theory often is touted as an important remedy for workplace inequality, in practice, it is less frequently used. Nevertheless, the theory can be a meaningful remedy for gender violence survivors who are subjected to adverse employment actions or termination for reasons that may not appear facially gendered. An employee may argue that the action had an impermissible gender-based disparate impact due to the disproportionate number of survivors who are women. Consequently, disparate impact would seem a natural remedy. This Article reviews * Professor, CUNY School of Law. This work has been funded by the City University of New York School of Law. It was supported (in part) by a grant from the City University of New York PSC-CUNY Research Award Program. Thanks to Beryl Blaustone for comments on an earlier draft and to Deborah Rubin, Nate Treadwell, and Mona Moayad for their able research assistance. GOLDSCHEID 10/28/2011 10:31 AM 34 OREGON LAW REVIEW [Vol. 90, 33 infrequently explored issues that could be interpreted to limit the theory’s utility. It concludes that those limitations should not bar the theory’s use. Instead, disparate impact should offer an alternative remedy for the hidden role domestic and sexual violence plays in perpetuating women’s economic inequality. I INTRODUCTION lthough disparate impact theory has been touted as one of the most important and controversial developments in employment discrimination law,1 in practice, it is less frequently used.2 On its face, this may seem surprising, given the challenges of proving discriminatory intent as part of a disparate treatment claim and recent scholarship demonstrating the ways that unconscious bias, rather than intentional discrimination, accounts for inequalities at work.3 Although some recent commentators have criticized the doctrine,4 others have called for its revival.5 This Article joins the dialog about the practical and potential utility of disparate impact theory by arguing that it should be applied to cases evaluating the theory’s applicability to a particular set of cases: those involving terminations of domestic violence survivors. The Article situates the issue in the context of two debates: one about the applicability of disparate impact theory and the other about the economic stability of domestic violence survivors. In cases in which a survivor is terminated or otherwise subjected to an adverse job 1 See, e.g., Michael Selmi, Was the Disparate Impact Theory a Mistake?, 53 UCLA L. REV. 701, 702 (2006). 2 See, e.g., Elaine W. Shoben, Disparate Impact Theory in Employment Discrimination: What’s Griggs Still Good For? What Not?, 42 BRANDEIS L.J. 597 (2004) (arguing that disparate impact litigation is a largely untapped resource that currently is not making a major impact on employment discrimination law). 3 Recent articles discussing the role of unconscious bias include, e.g., Tristin K. Green, Discrimination in Workplace Dynamics: Toward a Structural Account of Disparate Treatment Theory, 38 HARV. C.R.-C.L. L. REV. 91, 95–99 (2003); Anthony G. Greenwald & Linda Hamilton Krieger, Implicit Bias: Scientific Foundations, 94 CALIF. L. REV. 945 (2006); Melissa Hart, Subjective Decisionmaking and Unconscious Discrimination, 56 ALA. L. REV. 741, 745–50 (2005); Linda Hamilton Krieger & Susan T. Fiske, Behavioral Realism in Employment Discrimination Law: Implicit Bias and Disparate Treatment, 94 CALIF. L. REV. 997 (2006). 4 See infra notes 14–17 and accompanying text. 5 See infra notes 18–22 and accompanying text. A GOLDSCHEID 10/28/2011 10:31 AM 2011] Disparate Impact’s Impact 35 action because she6 has been the victim of abuse, she may argue that the action had an impermissible gender-based disparate impact due to the disproportionate number of survivors who are women.7 This Article evaluates the viability of that claim. The Article reviews and dismisses doctrinal arguments that could be raised to bar survivors’ claims. Instead, allowing such claims would advance the doctrine’s purpose of eradicating employment practices with discriminatory consequences.8 Elsewhere I have argued that adverse actions taken against domestic or sexual violence survivors should be analyzed as sex discrimination under disparate treatment theory.9 Accordingly, courts would recognize an unexplained adverse action against a domestic or sexual violence survivor as raising an inference of discrimination that the employer then could rebut through proof that the employer engaged in an interactive process with the employee but could not resolve the employment-related concern without taking the adverse action.10 That approach would encourage human resources’ “best practices” and safety planning at work.11 The approach would require decisions to be based on work-related determinations rather than on inaccurate stereotypes about abuse survivors.12 This analysis of disparate impact complements that proposal by evaluating an alternative legal theory. I argue that the theory should be available to domestic violence survivors, just as others subjected to adverse employment actions may bring both disparate treatment and disparate impact claims, and that doctrinal objections flatly barring such claims are inconsistent with the doctrine’s history and purpose. The Article first reviews both the theoretical underpinnings and the current doctrinal requirements of the statutory disparate impact claim. 6 Studies consistently show that the vast majority of victims of domestic and sexual violence are women. See, e.g., CALLIE MARIE RENNISON, U.S. DEP’T OF JUSTICE, BUREAU OF JUSTICE STATISTICS, INTIMATE PARTNER VIOLENCE, 1993–2001 (2003), http://bjs.ojp.usdoj.gov/content/pub/pdf/ipv01.pdf (concluding that eighty-five percent of all victimizations by intimate partners in 2001 were against women); U.S. DEP’T OF JUSTICE, BUREAU OF JUSTICE STATISTICS, CRIMINAL VICTIMIZATION IN THE UNITED STATES, STATISTICAL TABLES INDEX, tbl.2 (reporting that ninety-two percent of all sexual assaults in 2005 were committed against women). 7 See, e.g., RENNISON, supra note 6; U.S. DEP’T OF JUSTICE, supra note 6. 8 Griggs v. Duke Power Co., 401 U.S. 424, 432 (1971). 9 See Julie Goldscheid, Gender Violence and Work: Reckoning with the Boundaries of Sex Discrimination Law, 18 COLUM. J. GENDER & L. 61, 73–78 (2008). 10 Id. at 111. 11 Id. at 102–03. 12 Id. at 107–14. GOLDSCHEID 10/28/2011 10:31 AM 36 OREGON LAW REVIEW [Vol. 90, 33 The Article then focuses on two issues that might pose challenges to domestic violence survivors’ bringing claims. The first is the question whether a single employment decision should be considered a “particular employment practice” subject to disparate impact review. The second is the question of how to define the appropriate comparator group in cases such as these in which the criterion for the decision is closely linked to membership in a protected class. The Article concludes that disparate impact claims brought by domestic violence survivors subjected to adverse employment actions because of their experience with abuse fall squarely within the category of cases contemplated by Congress and the Court in recognizing disparate impact claims. A. Disparate Impact: A Theory of Limits or Untapped Potential? Generally speaking, disparate impact theory allows recovery when an employment policy or practice that is neutral on its face has a disparate impact on a protected group and the policy or practice is not justified by business necessity.13 Disparate impact theory has most frequently been used in the context of written employment tests or explicit job requirements; the theory has been less successful when invoked outside those contexts.14 Some theorists argue that the theory has had a limited effect and that it is not a promising vehicle for future change. For example, Michael Selmi argues that the theory ultimately was a mistake and that a broader definition of intentional discrimination (under disparate treatment) would more effectively eradicate employment discrimination.15 Other theorists similarly critique disparate impact as a tool for redressing inequality in today’s workplaces.16 Some scholars have identified practical limits, such as 13 42 U.S.C. § 2000e-2(k)(1) (2006). 14 See, e.g., Selmi, supra note 1, at 705, 738–67 (noting that disparate impact theory has produced limited results outside of testing context); Shoben, supra note 2, at 598 (asserting that disparate impact theory has largely untapped potential); Charles A. Sullivan, Disparate Impact: Looking Past the Desert Palace Mirage, 47 WM. & MARY L. REV. 911, 954 (2005) (noting “little development” in disparate impact law since the 1991 Civil Rights Act). 15 Selmi, supra note 1, at 706–07. 16 See, e.g., Green, supra note 3 at 136–44 (urging a focus on structural workplace dynamics); Linda Hamilton Krieger, The Content of Our Categories: A Cognitive Bias Approach to Discrimination and Equal Employment Opportunity, 47 STAN. L. REV. 1161, 1231–37 (1995) (arguing that implicit bias should be recognized as a form of intentional discrimination). GOLDSCHEID 10/28/2011 10:31 AM 2011] Disparate Impact’s Impact 37 the lack of jury trial, limited remedies, and statutory exemptions, as reasons disparate impact theory is not more widely used.17 Yet others argue for the theory’s expanded application. For example, Charles Sullivan urges a “return to, and revival of, the disparate impact theory.”18 He reviews and counters theoretical as well as doctrinal objections while arguing that disparate impact holds a “brighter promise” for expansion than does disparate treatment doctrine, particularly given constrained interpretations of disparate treatment requirements and our enhanced understandings of unconscious discrimination.19 Michelle Travis argues that disparate impact theory holds great potential for addressing gender-based workplace inequalities that stem from organizational norms that enshrine longstanding inequalities such as those involving caregiving.20 Joan Williams and Nancy Segal similarly highlight disparate impact’s utility in challenging workplace practices and norms that penalize those with caregiving responsibilities.21 Elaine Shoben argues that disparate impact is unnecessarily underutilized and calls for “creative thinking” about additional applications of the theory.22 This Article picks up on that call and offers one example of an application for which the doctrine could afford needed relief. Domestic violence survivors may be terminated or subjected to 17 Sullivan, supra note 14, at 968; see also Samuel R. Bagenstos, The Structural Turn and the Limits of Antidiscrimination Law, 94 CALIF. L. REV. 1, 21–24 (2006) (detailing limited applications of disparate impact law); Shoben, supra note 2, at 598–99 (citing limited damages as well as difficulties of proof, employer savvy, and failure of the plaintiffs’ bar to appreciate the theory’s potential as reasons for the theory’s underutilization). 18 Sullivan, supra note 14, at 984. 19 Id. at 968–1000. 20 Michelle A. Travis, Recapturing the Transformative Potential of Employment Discrimination Law, 62 WASH. & LEE L. REV. 3, 37 (2005). 21 Joan C. Williams & Nancy Segal, Beyond the Maternal Wall: Relief for Family Caregivers Who Are Discriminated Against on the Job, 26 HARV. WOMEN’S L.J. 77, 108– 10 (2003); see also, e.g., Joan C. Williams & Stephanie Bornstein, The Evolution of “FReD”: Family Responsibilities Discrimination and Developments in the Law of Stereotyping and Implicit Bias, 59 HASTINGS L.J. 1311, 1344–45 (2008) (arguing for application of disparate impact theory to family responsibilities discrimination); accord Reva B. Siegel, Note, Employment Equality Under the Pregnancy Discrimination Act of 1978, 94 YALE L.J. 929, 956 (1985) (arguing that disparate impact theory should apply to pregnancy-related discrimination under the Pregnancy Discrimination Act). But see, e.g., Joanna L. Grossman, Pregnancy, Work, and the Promise of Equal Citizenship, 98 GEO. L.J. 567, 615–19 (2010) (arguing that despite its theoretical applicability, disparate impact has proven to be of limited utility to pregnant workers). 22 Shoben, supra note 2, at 607. GOLDSCHEID 10/28/2011 10:31 AM 38 OREGON LAW REVIEW [Vol. 90, 33 adverse employment consequences as a result of the abuse.23 Because domestic violence survivors overwhelmingly are women, those adverse employment actions will have a disproportionate impact based on sex. Disparate impact would seem a natural basis for redress. Critics no doubt are correct that the reach of the theory will be limited, that law cannot do the work of social movements in shifting social norms,24 and that our lack of consensus regarding employers’ responsibility for remedying subtle discrimination limits prospects for both legal and social change.25 That said, disparate impact remains a viable theory, devised to provide redress in cases such as these, when “neutral” practices perpetuate prohibited discrimination. Although the number of cases to which this argument may apply likely will be small, the analysis offers an example of how disparate impact could be a more effective tool in remedying subtle discrimination at work. B. Context: Domestic Violence The experience of domestic or sexual violence survivors illustrates the potential, and some of the challenges, of disparate impact theory. Domestic or sexual violence survivors often are subjected to adverse employment actions as a result of their experiences with the abuse. For example, a domestic violence survivor may be stalked at work by her partner, and her employer may terminate her, even in the absence of evidence that her partner actually posed a safety threat to the workplace.26 Survivors may reflexively be terminated, perhaps due to unconscious stereotypes that someone coping with abuse would not be able to simultaneously maintain her job or inaccurate assumptions about her needs as a result of the abuse.27 In other cases, survivors may be disciplined or terminated because the abuser interferes with their ability to perform their job functions on time.28 In those cases, survivors may bring claims of disparate treatment or hostile environment harassment,29 claims of wrongful discharge,30 or claims 23 See infra Part I.B. 24 See Selmi, supra note 1, at 780–81. 25 Bagenstos, supra note 17, at 34–40 (calling for enhanced normative principles). 26 See, e.g., Goldscheid, supra note 9, at 73–78. 27 Id. at 87–88, 95–101. 28 Id. at 73–77. 29 Id. at 85–88, 91–95. 30 See, e.g., Apessos v. Mem’l Press Grp., 15 Mass. L. Rptr. 322 (Mass. Super. Ct. 2002); Danny v. Laidlaw Transit Servs., Inc., No. 78421-3 (W.D. Wash. Oct. 3, 2008). GOLDSCHEID 10/28/2011 10:31 AM 2011] Disparate Impact’s Impact 39 under state or local laws in those jurisdictions that explicitly prohibit adverse actions taken on the basis of an employee’s experience as a survivor of abuse.31 Those remedies may or may not afford adequate relief.32 Alternatively, a survivor who is terminated because she has been subjected to abuse may argue that the termination reflects a facially gender-neutral policy that has a disparate impact based on sex.33 For example, a survivor may be terminated and told that her termination was due to circumstances involving her abuse. Whether or not a survivor discloses that she is coping with abuse, her employer may be aware of the problem and may terminate her as a result. The employer may cite performance or attendance issues or may base the decision on fears that her abuser will pose safety threats to the workplace. In those cases in which an employer disciplines or terminates an employee because of her experience with abuse, the But see, e.g., Imes v. City of Asheville, 594 S.E.2d 397 (N.C. Ct. App. 2004), aff’d, 606 S.E.2d 117 (N.C. 2004) (rejecting the claim); Green v. Bryant, 887 F. Supp. 798 (E.D. Pa. 1995) (same). See generally Margaret C. Hobday, Protecting Economic Stability: The Washington Supreme Court Breathes New Life in the Public-Policy Exception to At-Will Employment for Domestic Violence Victims, 17 WM. & MARY J. WOMEN & L. 87 (2010); Sandra S. Park, Working Towards Freedom from Abuse: Recognizing a “Public Policy” Exception to Employment-At-Will for Domestic Violence Victims, 59 N.Y.U. ANN. SURV. AM. L. 121 (2003). 31 See, e.g., LEGAL MOMENTUM, WOMEN’S LEGAL DEF. & EDUC. FUND, STATE LAW GUIDE: EMPLOYMENT RIGHTS FOR VICTIMS OF DOMESTIC OR SEXUAL VIOLENCE (Sept. 2010), http://www.legalmomentum.org/assets/pdfs/employment-rights.pdf; see also, e.g., Deborah A. Widiss, Domestic Violence and the Workplace: The Explosion of State Legislation and the Need for a Comprehensive Strategy, 35 FLA. ST. U. L. REV. 669 (2008). 32 See, e.g., Goldscheid, supra note 9, at 95–104. 33 Similar arguments have met with mixed results in cases brought by survivors and their families who allege that law enforcement’s failure to respond appropriately to domestic violence victims’ calls for help led them to suffer harm. For example, a number of courts have rejected claims. See, e.g., Ricketts v. City of Columbia, 36 F.3d 775, 781– 82 (8th Cir. 1994); Eagleston v. Guido, 41 F.3d 865, 878 (2d Cir. 1994); McKee v. City of Rockwall, 877 F.2d 409, 416 (5th Cir. 1989). Others have upheld claims under equal protection and substantive due process theories. See, e.g., Navarro v. Block, 72 F.3d 712, 715–17 (9th Cir. 1995) (upholding an equal protection claim that domestic violence victims were treated less seriously than similar nondomestic violence victims under rational basis review); Thurman v. City of Torrington, 595 F. Supp. 1521, 1529 (D. Conn. 1984) (rejecting a motion to dismiss an equal protection claim challenging police policy of differential treatment of domestic violence victims); see also, e.g., Okin v. Village of Cornwall-on-Hudson Police Dep’t, 577 F.3d 415 (2d Cir. 2009) (upholding a survivor’s claim that police response violated substantive due process based on state-created danger theory). But see, e.g., Castle Rock v. Gonzales, 545 U.S. 748 (2005) (rejecting a procedural due process claim by a survivor when law enforcement failed to take action to enforce a protective order). GOLDSCHEID 10/28/2011 10:31 AM 40 OREGON LAW REVIEW [Vol. 90, 33 decision may be challenged as a facially neutral practice that has a disparate impact based on sex.34 There is an intuitive appeal to this argument, given domestic and sexual violence’s disparate impact on women.35 Certainly, if an employer had an explicit policy approving adverse employment actions taken against domestic violence survivors, such a policy would easily be seen as having a gendered impact on the workplace, given the disproportionate percentage of women who are victims. Indeed, some commentators have concluded that the theory should apply easily to these cases.36 But employers are not likely to expressly adopt such policies. Consequently, the issue becomes whether an individual employment decision should be insulated from disparate impact review because, due to the nature of the decision, it is not likely to be either formalized or publicly disclosed. If disparate impact is a tool for rooting out discriminatory workplace decisions that either reflect pretext or perpetuate historic inequalities based on membership in protected categories including sex, disparate impact’s application would seem natural.37 34 The claim might be fashioned in a variety of ways. It could be framed as a disparate treatment claim. See Goldscheid, supra note 9. Alternatively, a claim could be framed as a challenge to the underlying personnel policy, e.g., the policy regarding absenteeism or performance. But that approach would not accurately capture the claim. Unlike other cases, for example, the “caregiving” challenges in which employees challenge an employer’s leave policies as having a disparate impact based on sex or pregnancy, a survivor would not be challenging a stated policy. See infra notes 147–150 for discussion of those cases. Instead, she would be challenging an employment decision for its impact on domestic violence survivors. 35 See RENNISON, supra note 6. 36 Maria Amelia Calaf, Comment, Breaking the Cycle: Title VII, Domestic Violence, and Workplace Discrimination, 21 LAW & INEQ. 167, 186–91 (2003); John E. Matejkovic, Which Suit Would You Like? The Employer’s Dilemma in Dealing with Domestic Violence, 33 CAP. U. L. REV. 309, 336 (2004); Nicole Buonocore Porter, Victimizing the Abused?: Is Termination the Solution When Domestic Violence Comes to Work?, 12 MICH. J. GENDER & L. 275, 294–95 (2006); Nina W. Tarr, Employment and Economic Security for Victims of Domestic Abuse, 16 S. CAL. REV. L. & SOC. JUST. 371, 393–94 (2007); Wendy R. Weiser & Deborah A. Widiss, Employment Protection for Domestic Violence Victims, 38 CLEARINGHOUSE REV. J. POV. L. & POL’Y 3, 6–7 (2004). 37 Indeed, the connection between domestic violence and sex discrimination is so strong that international treaties explicitly equate domestic violence with sex discrimination. See, e.g., Declaration on the Elimination of Violence Against Women, G.A. Res. 48/104, U.N. Doc. A/RES/48/104 (Dec. 20, 1993) (stating that violence against women is an “obstacle to the achievement of equality” and that it “constitutes a violation of the rights and fundamental freedoms of women,” and recognizing that “violence against women is a manifestation of historically unequal power relations between men and women, which have led to domination over and discrimination against women by men”). The persistence of sex-based discriminatory attitudes towards survivors could be extended to the GOLDSCHEID 10/28/2011 10:31 AM 2011] Disparate Impact’s Impact 41 Of course, the fact of a termination or other adverse action taken against an employee would not by itself trigger liability. As with other cases in which a plaintiff establishes a prima facie showing of disparate impact, the employer would be able to defend against liability by establishing that the action was job related for the position in question and consistent with business necessity.38 Consequently, in cases in which performance or attendance problems persisted despite the employer’s good faith efforts to help the employee safely negotiate ongoing employment despite the abuse, or when safety risks to the workplace overall could not be reduced without altering the survivor’s job status, the adverse action might be justified and the disparate impact claim would fail. Analogous arguments have proven successful on behalf of domestic violence victims who lose housing based on policies disadvantaging domestic violence victims.39 However, employment differs from housing in that employers are not likely to maintain explicit policies disadvantaging domestic violence survivors whereas landlords have maintained such explicit policies. In addition, at least two doctrinal limitations that have evolved in the workplace context may render disparate impact arguments difficult to sustain. First, some courts have narrowly construed the statutory requirement that a plaintiff prove that a “particular employment practice”40 caused the employment context, in which termination of domestic violence survivors because of their experience as survivors could similarly be treated as sex discrimination. 38 42 U.S.C. § 2000e-2(k)(1)(A)(i) (2006). See infra Part II.B. 39 See, e.g., Bouley v. Young-Sabourin, 394 F. Supp. 2d 675 (D. Vt. 2005) (upholding sex discrimination claim under Fair Housing Act when landlord sought to evict domestic violence victim less than seventy-two hours after her husband assaulted her); Alvera v. Creekside Vill. Apts., HUD ALJ No. 10-99-0538-8 (U.S. Dep’t of Hous. & Urban Dev., Portland, Or., Apr. 16, 2001) (policy discriminating against domestic violence victims violate prohibitions on sex discrimination); see also, e.g., 42 U.S.C. § 1437f(c)(9) (2006) (providing, inter alia, that applicants for Section 8 housing vouchers cannot be denied housing because of their experiences with domestic violence and that an abuser’s actual or threatened violence cannot be the basis for adverse decisions regarding the victim’s housing); Emily J. Martin, Fair Housing for Battered Women: Preventing Homelessness Through Civil Rights Laws, 27 CORNERSTONE 6 (2005), available at http://www.aclu.org /pdfs/fairhousingforbatteredwomen072806.pdf (discussing theories and citing cases); PUB. HOUS. MGMT. & OCCUPANCY DIV., U.S. DEP’T OF HOUS. & URBAN DEV., PUBLIC HOUSING OCCUPANCY GUIDEBOOK 216–19 (2003) (recommending preferences for domestic violence victims and discussing best practices for maintaining victims’ housing status). See generally Lenora M. Lapidus, Doubly Victimized: Housing Discrimination Against Victims of Domestic Violence, 11 AM. U. J. GENDER SOC. POL’Y & L. 377 (2002). 40 42 U.S.C. §2000e-2(k)(1)(A)(i). GOLDSCHEID 10/28/2011 10:31 AM 42 OREGON LAW REVIEW [Vol. 90, 33 disparate impact.41 Second, the survivor would have to establish the requisite statistical impact, even though she may have been the only employee who she was aware was similarly treated.42 This Article will evaluate each of those issues in turn. II DISPARATE IMPACT: TESTING THEORY AND DOCTRINE A. Theory The history of disparate impact doctrine’s development as a tool to eliminate employment discrimination has been well documented and won’t be repeated here.43 What is less settled are its purposes, rationales, and goals. Drawing from the oft-stated original articulation in Griggs v. Duke Power Co., disparate impact theory has been seen as prohibiting practices that are “fair in form, but discriminatory in operation.”44 As the Supreme Court stated: The objective of Congress in the enactment of Title VII is plain from the language of the statute. It was to achieve equality of employment opportunities and remove barriers that have operated in the past to favor an identifiable group of white employees over other employees. Under the Act, practices, procedures, or tests neutral on their face, and even neutral in terms of intent, cannot be maintained if they operate to “freeze” the status quo of prior discriminatory employment practices.45 The Court saw Title VII’s antidiscrimination mandate as demanding the removal of all “artificial, arbitrary, and unnecessary barriers to employment,” when those barriers have the effect of 41 See, e.g., Sullivan, supra note 14, at 976–79 (discussing limitations); Travis, supra note 20, at 36–40. 42 Cf., e.g., Barbara J. Flagg, Fashioning a Title VII Remedy for Transparently White Subjective Decisionmaking, 104 YALE L.J. 2009, 2027 (1995) (discussing difficulty in identifying policies and their impact in a hypothetical claim by an African American woman who was denied promotion). 43 See, e.g., Robert Belton, Title VII at Forty: A Brief Look at the Birth, Death, and Resurrection of the Disparate Impact Theory of Discrimination, 22 HOFSTRA LAB. & EMP. L.J. 431 (2005); Alfred W. Blumrosen, Strangers in Paradise: Griggs v. Duke Power Co. and the Concept of Employment Discrimination, 71 MICH. L. REV. 59 (1972); Kingsley R. Browne, The Civil Rights Act of 1991: A “Quota Bill,” a Codification of Griggs, a Partial Return to Wards Cove, or All of the Above?, 43 CASE W. RES. L. REV. 287 (1993); Martha Chamallas, Evolving Conceptions of Equality Under Title VII: Disparate Impact Theory and the Demise of the Bottom Line Principle, 31 UCLA L. REV. 305 (1983); Selmi, supra note 1, at 708–33. 44 401 U.S. 424, 431 (1971). 45 Id. at 429–30. GOLDSCHEID 10/28/2011 10:31 AM 2011] Disparate Impact’s Impact 43 discriminating against a historically disadvantaged group.46 It saw Congress’s goal of equal opportunity as encompassing the removal of all “‘built-in headwinds’” to equality.47 In other words, if the employer’s conduct operates to exclude or disadvantage protected groups, the conduct would be permitted only if it is related to job performance.48 Yet this language doesn’t fully capture the doctrine’s underlying theory or its goals. Richard Primus has posited two primary potential rationales.49 In his view, disparate impact can be seen either as an “evidentiary dragnet designed to discover hidden instances of intentional discrimination” or as a “more aggressive attempt to dismantle racial hierarchies regardless of whether anything like intentional discrimination is present.”50 He concludes that an account that focuses on breaking down historical hierarchy and persistent segregation is preferable to approaches that focus on the employer’s state of mind.51 Sullivan identifies two potential rationales that describe the same, or very similar concerns: reaching intentional discrimination in cases that lack proof of disparate treatment and removing unnecessary barriers based on the history of discrimination and subordination.52 Under both views, disparate impact draws on antidiscrimination law’s conceptual roots in tort law by recognizing that discrimination consists of conduct rather than a state of mind and by defining “intent” to include the likelihood that a given result will flow from a given action.53 The two rationales generally track two 46 Id. at 431. 47 Id. at 432. 48 Id. at 431. 49 Richard A. Primus, Equal Protection and Disparate Impact: Round Three, 117 HARV. L. REV. 494, 518–36 (2003). Others have adopted his framework in their analyses of disparate impact discrimination. See, e.g., Jennifer L. Peresie, Toward a Coherent Test for Disparate Impact Discrimination, 84 IND. L.J. 773, 779 (2009). 50 Primus, supra note 49, at 518. Others have also seen the focus of disparate impact as focusing on the risk that the challenged practice incorporates or reflects pretext. See, e.g., George Rutherglen, Disparate Impact Under Title VII: An Objective Theory of Discrimination, 73 VA. L. REV. 1297, 1339–43 (1987). 51 Primus, supra note 49, at 536. Others agree that framing the theory in terms of effects rather than intent holds greater potential to advance equality. See, e.g., Pamela L. Perry, Two Faces of Disparate Impact Discrimination, 59 FORDHAM L. REV. 523, 526 (1991). 52 Sullivan, supra note 14, at 964–66. 53 Blumrosen, supra note 43, at 71–72. GOLDSCHEID 10/28/2011 10:31 AM 44 OREGON LAW REVIEW [Vol. 90, 33 competing visions of workplace equality: equal treatment and equal achievement.54 The latter rationale in particular, in which disparate impact is viewed as a tool to address historic barriers, seems consistent with the structural approach advocated by those who advocate new approaches to workplace inequality.55 In that sense, it may be thought to be an effective tool for addressing subconscious discrimination.56 Notably, Linda Krieger, in her landmark work identifying the role of “unconscious” discrimination in employment inequality, argues that disparate impact is an ill-suited tool for addressing unconscious discrimination.57 She criticizes disparate impact as a vehicle for addressing the subjective practices that often are not recognized under disparate treatment’s current requirement of discriminatory intent.58 She views the original purpose of disparate impact law as addressing particular problems that had arisen as Title VII evolved.59 She argues that disparate impact was not designed to go beyond those contexts to address systems of subjective practices.60 Instead, in her view, disparate treatment should be reconceptualized to reflect the role of cognition, rather than intent, in the realistic operation of discriminatory decisions.61 This critique stands in some tension with accounts of those involved in the Griggs litigation and related policy development, which define disparate impact discrimination broadly in terms of consequence.62 54 Robert Belton, The Dismantling of the Griggs Disparate Impact Theory and the Future of Title VII: The Need for a Third Reconstruction, 8 YALE L. & POL’Y REV. 223, 224–25 (1990); see also, e.g., Julia Lamber, Discretionary Decisionmaking: The Application of Title VII’s Disparate Impact Theory, 1985 U. ILL. L. REV. 869, 903 (1985) (arguing that disparate impact reflects an equal opportunity concept of equality with goals of eliminating historical, social, and structural barriers that impede achievement for minority group members and recognizing the present effects of past discrimination). 55 See sources cited supra note 3. 56 See, e.g., Primus, supra note 49, at 532–35. 57 Krieger, supra note 3, at 1231. 58 Id. at 1230. She rejects disparate impact as a “fix” for challenging those practices, citing practical concerns with validation studies, political problems with the cost of the validation studies that would be required, and theoretical problems with the fit between the original intent of disparate impact theory and the subjective practices with which she is concerned. Id. at 1231–37. 59 Those problems are employers’ substitution of facially neutral criteria for previously used race-based classifications and similar reliance on eligibility criteria associated with successful performance in stereotypically male jobs. Id. at 1237. 60 Id. 61 Id. at 1241–43. 62 Blumrosen, supra note 43, at 62, 73. GOLDSCHEID 10/28/2011 10:31 AM 2011] Disparate Impact’s Impact 45 From a policy perspective, both approaches require courts to balance the impact of neutral policies that have a discriminatory effect against an employer’s legitimate right to promulgate policies regardless of racial or gendered effects if they truly are directed to central business (or governmental) concerns.63 Cases involving domestic violence survivors may fall under both categories, in that they may reflect masked discriminatory intent (pretext) or decisions that perpetuate historic gender-based discriminatory practices. Accordingly, disparate impact would be applicable under either theoretical approach. B. Doctrine 1. Statutory Requirements Congress’s 1991 codification of disparate impact as a theory of relief for employment discrimination provides that a plaintiff can establish disparate impact discrimination by showing that a “particular employment practice” causes a disparate impact against an employee in a group protected by the statute.64 The employer then has the burden of establishing that the practice is job related for the position and consistent with business necessity.65 Even if the employer makes that showing, the plaintiff can prevail if she establishes that an “alternative employment practice” exists and that the employer refuses to adopt it.66 Consequently, proof of both a “particular employment practice” and an impermissible disparate impact are key to sustaining a claim. For disparate impact theory to apply to a termination or other adverse action taken against a domestic violence survivor, a court would first have to recognize the adverse action taken against an individual as a “particular employment practice” that falls within the statute. As one court noted, there is “precious little” case law on the meaning of 63 Lamber, supra note 54, at 903. 64 42 U.S.C. § 2000e-2(k)(1)(A)(i) (2006) (prohibiting discrimination based on race, color, religion, sex, or national origin). Even if the plaintiff cannot demonstrate that each particular challenged employment practice causes a disparate impact, the court may treat the entire decision-making process as one employment practice if the plaintiff demonstrates that the elements of an employer’s decision-making process are not capable of separation for analysis. Id. § 2000e-2(k)(1)(B)(i). 65 Id. § 2000e-2(k)(1)(A)(i). 66 Id. § 2000e-2(k)(1)(A)(ii). GOLDSCHEID 10/28/2011 10:31 AM 46 OREGON LAW REVIEW [Vol. 90, 33 “employment practice” as the term is used in disparate impact cases.67 Courts consistently apply the doctrine to “practices” such as discrete testing procedures or job requirements68 or express personnel policies.69 However, some courts conclude that the theory would not extend, for example, to employment decisions that affect a single or only a few employees, or to employment actions that could be characterized as “single decisions” or “isolated incidents.”70 This could be problematic for a domestic violence survivor. In addition to the likely absence of an explicit policy authorizing adverse actions against domestic violence survivors, a survivor may have difficulty identifying others who also were subject to an adverse action because of their status as survivors. Yet the statutory language, legislative history, and the principles underlying the theory do not justify precluding a disparate impact claim on this ground alone. Although Title VII generally defines “unlawful employment practices,”71 it does not define or elaborate how that definition is meant to differ from the “particular employment practice[s]” required by the separate provision spelling out proof requirements for disparate impact claims.72 As Michelle Travis argues, the term “particular” was used to clarify the issue addressed in Wards Cove73 concerning whether an employee could state a prima facie disparate impact case solely by identifying a “bottom-line” statistical disparity between the 67 Council 31, Am. Fed’n of State, Cnty. & Mun. Emps. v. Ward, 978 F.2d 373, 377 (7th Cir. 1992). 68 See Griggs v. Duke Power Co., 401 U.S. 424 (1971) (applying disparate impact analysis to high school diploma requirement); see also, e.g., Dothard v. Rawlinson, 433 U.S. 321 (1977) (applying disparate impact to challenge height and weight requirements). 69 For example, policies requiring commencement of leave upon pregnancy would be subject to disparate impact review. See Harriss v. Pan Am. World Airways, Inc., 649 F.2d 670 (9th Cir. 1980). 70 See, e.g., infra Part II.B.2. 71 The statute declares: It shall be an unlawful employment practice for an employer—(1) to fail or refuse to hire or to discharge any individual, or otherwise to discriminate against any individual with respect to his compensation, terms, conditions, or privileges of employment, because of such individual’s race, color, religion, sex, or national origin; or (2) to limit, segregate, or classify his employees or applicants for employment in any way which would deprive or tend to deprive any individual of employment opportunities or otherwise adversely affect his status as an employee, because of such individual’s race, color, religion, sex, or national origin. 42 U.S.C. § 2000e-2(a). 72 Id. § 2000e-2(k)(1)(A)(i). 73 Wards Cove Packing Co. v. Atonio, 490 U.S. 642 (1989). GOLDSCHEID 10/28/2011 10:31 AM 2011] Disparate Impact’s Impact 47 percentage of the employer’s workforce that was made up of members of a protected category and the percentage of that category in the relevant labor pool.74 That issue addresses a concern with causation, not a concern about the type of neutral decision that might create a disparate impact. In other words, by using the term “particular,” Congress sought to ensure that an employer-authorized practice, rather than some other unrelated factor, caused the challenged disparity.75 Accordingly, the statute requires a plaintiff to demonstrate that each practice causes the disparity, unless she can establish that the elements of the employer’s decision-making process cannot be separated for analysis, in which case the combined processes would be treated as a “particular employment practice.”76 The use of the term “particular” therefore clarifies how to handle “bottom-line disparities” and does nothing to disturb the Griggs decision’s recognition that a broad range of employment decisions would be subject to disparate impact challenge.77 The 1991 Civil Rights Act’s legislative history confirms that the use of the word “particular” reflects Congress’s concern with causation in cases involving “bottom-line disparities” rather than an effort to limit the types of practices subject to disparate impact challenges.78 As L. Camille Hébert chronicles, the legislative history reflects Congress’s recognition that disparate impact has been, and may be, applied to a “wide variety of practices” including tests, job requirements, leave or other personnel policies, or “other subjective or objective” evaluation procedures or practices.79 Indeed, the Supreme Court has recognized that disparate impact may be used to challenge “subjective” employment decisions, such as the method for awarding or denying promotions.80 That disparate impact challenges have not been widely successful outside the testing context does not mean that they could or should not be successful in appropriate cases. 74 Travis, supra note 20, at 80; see also Muñoz v. Orr, 200 F.3d 291, 304 (5th Cir. 2000) (recognizing that disparate impact cannot be used to challenge “cumulative effects of an employer’s practices”). 75 Travis, supra note 20, at 80–82. 76 42 U.S.C. § 2000e-2(k)(1)(B). 77 Griggs v. Duke Power Co., 401 U.S. 424, 432 (1971). 78 See, e.g., Browne, supra note 43, at 328–39 (reviewing legislative history of “particular employment practice” requirement). 79 L. Camille Hébert, The Disparate Impact of Sexual Harassment: Does Motive Matter?, 53 U. KAN. L. REV. 341, 380 & n.151 (citing H.R. REP. NO. 102-40(I), at 34 (1991), reprinted in 1991 U.S.C.C.A.N. 549, 572–76; H.R. REP. NO. 102-40(II), at 16 (1991), reprinted in 1991 U.S.C.C.A.N. 694, 709). 80 Watson v. Fort Worth Bank & Trust, 487 U.S. 977, 991 (1988). GOLDSCHEID 10/28/2011 10:31 AM 48 OREGON LAW REVIEW [Vol. 90, 33 2. “Single Decision” Cases This Part reviews the limited universe of reported cases in which defendants have challenged the applicability of disparate impact theory by arguing that the employment decision at issue was a “single decision” and not a “particular employment practice” that would be subject to disparate impact review. If a domestic violence survivor brought a disparate impact challenge to her termination or other adverse action, the claim might be characterized as raising this “single decision” issue.81 The question whether disparate impact theory applies when subjective decisions may be characterized as “single employment decisions” has been considered, but not rigorously analyzed, in a relatively small number of decisions. Some courts have recognized that a decision can have a discriminatory disparate impact even if it was a “single decision” that did not reflect an official or repeatedly applied policy. Watson v. Fort Worth Bank & Trust itself was a case involving a small employer, and the Court there recognized that single decisions could perpetuate discrimination.82 If the purposes of disparate impact theory are to eliminate the “headwinds” that interfere with equality and to facilitate challenges to decisions that have the effect of adversely impacting a protected group and lack business justification, one would expect all manner of neutral decisions that have a negative impact on a protected group to be subject to review, regardless of the number of employees affected. Accordingly, a domestic violence survivor should not have to establish that her employer adopted a formal policy sanctioning adverse actions against survivors of domestic violence in order to bring a disparate impact challenge. In many cases, courts treat disparate impact in a manner that is consistent with the theory’s stated purpose of eliminating hidden discrimination by upholding claims that suggest pretext, even absent traditional proof of disparate treatment. For example, in Council 31, 81 A preliminary question may be whether the decision is the type that is properly subject to disparate impact review. Although disparate impact is most often thought of in the context of objective criteria such as tests, it is an important tool for challenging subjective practices such as terminations or other adverse actions. See, e.g., id. (confirming applicability of disparate impact analysis to subjective employment decisions such as hiring, compensation, and promotions); Wards Cove Packing Co. v. Atonio, 490 U.S. 642, 657 (1989) (listing “subjective decision making” as among the practices that could be challenged under disparate impact theory). 82 487 U.S. at 990–91. GOLDSCHEID 10/28/2011 10:31 AM 2011] Disparate Impact’s Impact 49 American Federation of State, County & Municipal Employees v. Ward, the Seventh Circuit Court of Appeals upheld a disparate impact challenge to a layoff decision that allegedly had a racially discriminatory impact.83 The appellate court reversed the lower court’s dismissal and rejected the conclusion that a disparate impact challenge must identify a “repeated, customary method of operation,” and that it would not apply to a single layoff decision.84 The appellate court reasoned that a decision barring from disparate impact challenges employment practices that may be characterized as a “single decision” would be “analytically unmanageable” because “almost any repeated course of conduct can be traced back to a single decision.”85 As another court stated, the criteria should be whether a practice is sufficiently specific and focused for the court to be able to address “whether it is a pretext for discrimination in light of the employer’s explanation for the practice.”86 Indeed, a number of claims involving single layoff decisions have been recognized as cognizable under disparate impact theory, even if the plaintiffs ultimately did not prevail on the merits.87 Similarly, other challenges to termination decisions or to aspects of hiring practices have been upheld even though the challenged practice might have been characterized as a “single decision” and did not involve formal policies or procedures. For example, in Chaney v. Southern Railway Co., the court allowed the plaintiff’s disparate impact claim to proceed based on the former employee’s argument that he had been terminated as a result of an employer-administered drug test that had a disparate impact based on race.88 The court did not require proof that the test had been used in other instances or that its use was part of the employer’s official policy or procedures. Similarly, in Conroy v. Johanns, the court allowed a claim to proceed based on the argument that the employer’s decision to advertise a 83 978 F.2d 373, 377 (7th Cir. 1992). 84 Id. at 376. 85 Id. at 377. 86 Am. Fed’n of State, Cnty. & Mun. Emps. v. Washington, 770 F.2d 1401, 1405 (9th Cir. 1985). 87 See, e.g., Frank v. Xerox Corp., 347 F.3d 130 (5th Cir. 2003) (alleging so-called “balanced workforce initiative” generated reports that set specific racial goals for jobs at particular grade levels led to subsequent reduction in percentage of black employees); cf. Sengupta v. Morrison-Knudsen Co., 804 F.2d 1072 (9th Cir. 1986) (allowing disparate impact challenge to layoff decision though finding no statistical impact); Nolting v. Yellow Freight Sys., Inc., 799 F.2d 1192 (8th Cir. 1986) (same). 88 847 F.2d 718, 725 (11th Cir. 1988). GOLDSCHEID 10/28/2011 10:31 AM 50 OREGON LAW REVIEW [Vol. 90, 33 position in a job category that required a college degree had a gendered disparate impact.89 The court rejected the argument that the decision did not constitute an “employment practice” because the plaintiff had established inconsistencies in how jobs were posted that could constitute a “standard practice” and that might indicate a pretext for discrimination.90 The court did not require a formal showing of a stated policy; instead it allowed the plaintiff the opportunity to establish facts that would show a practice that had a gendered impact. In other cases, courts have recognized that claims involving an adverse action taken against an individual could be subject to a disparate impact challenge, even though the claims in those cases ultimately failed for lack of statistical evidence. A concern about whether the decisions reflected a pretext for discrimination seemed to drive the court. For example, in Coe v. Yellow Freight System, Inc., the court recognized that an employer’s refusal to admit the plaintiff to its management training program could be subject to a disparate impact challenge.91 The court reasoned that the employer’s policy with respect to program admission was sufficiently narrow to suggest that it might be a “ploy” to “practice discrimination.”92 Rather than taking a formalistic approach and requiring a stated policy or evidence of repeated conduct, the court recognized that a lack of apparent business justification for the employment decision might reflect the type of discriminatory decision disparate impact theory was intended to address.93 This flexible approach hews more closely to the broad-sweeping intent reflected in both Griggs and the 1991 Civil Rights Act (CRA). To the extent courts are charged with balancing a concern with perpetuating the effects of discrimination with employers’ interests in maintaining valid business practices, the “business necessity” aspect of the equation,94 rather than the “employment practice” requirement, 89 No. 2:06-cv-867, 2007 WL 1830725 (D. Utah June 22, 2007). 90 Id. at *7. 91 646 F.2d 444, 452 (10th Cir. 1981). 92 Id. 93 Id. 94 See 42 U.S.C. § 2000e-2(k)(1)(A) (2006) (setting forth proof requirements under which a claim will fail if an employer establishes that the challenged practice is job related for the position in question and consistent with business necessity). For discussion of the business necessity defense, see, e.g., Susan S. Grover, The Business Necessity Defense in Disparate Impact Discrimination Cases, 30 GA. L. REV. 387 (1996); Andrew C. Spiropoulos, Defining the Business Necessity Defense to the Disparate Impact Cause of Action: Finding the Golden Mean, 74 N.C. L. REV. 1479 (1996). GOLDSCHEID 10/28/2011 10:31 AM 2011] Disparate Impact’s Impact 51 should do the work of protecting the employer’s interest. The employer should be required to justify its decision if its practice raises the specter of enshrining or reproducing historic discrimination. Nevertheless, a number of courts have rejected claims on the ground that the plaintiff did not identify a “particular employment practice” within the meaning of the statute. As Sullivan argues, the objections are not categorically unfounded, but they are “overdrawn.”95 When closely examined, virtually all of these decisions reflect what seem to be concerns with causation or concerns that disparate impact theory should not be applied in a way that would endorse quotas or other public policies or that would interfere with employers’ legitimate, or at least traditional, business practices (though that analysis would more properly fall within the “business necessity” defense).96 Nevertheless, some courts go further in rejecting claims even though the challenged practices could reasonably be seen as falling within the theory’s reach. One group of cases purports to reject claims because the plaintiff failed to identify a particular employment practice, but a close reading of the decisions reveals an underlying concern with a failure to establish causation between the employment decision and the prohibited disparate impact.97 For example, in Pouncy v. Prudential Insurance Co. of America, plaintiffs challenged three aspects of the employer’s practices they claimed had a disparate impact based on race.98 Although the court stated that those practices were different from selection procedures such as educational requirements and aptitude tests, it really was concerned about whether the challenged practices caused the alleged racial disparities.99 The court recognized that the plaintiffs presented statistics showing that blacks were overrepresented in the lower levels of the employer’s workforce but 95 Sullivan, supra note 14, at 976. 96 See, e.g., Browne, supra note 43, at 327–29 (explaining how a “particularity” requirement addressed concerns that disparate impact challenges to general decision- making processes would lead to routine bottom-line challenges, impair an employer’s ability to select the most qualified candidate, and pressure employers to engage in quota hiring); accord Selmi, supra note 1, at 753 (noting courts’ willingness to approve “common business practices” despite their disparate impact). 97 See, e.g., Pouncy v. Prudential Ins. Co. of Am., 668 F.2d 795, 800 (5th Cir. 1982); accord Anderson v. Douglas & Lomason Co., 26 F.3d 1277 (5th Cir. 1994). 98 Pouncy, 668 F.2d at 801 (dismissing the three “practices” of failing to post job openings, the use of a “level” system, and the use of subjective evaluation criteria, as unlike the selection procedures to which the disparate impact model traditionally has been applied). 99 Id. GOLDSCHEID 10/28/2011 10:31 AM 52 OREGON LAW REVIEW [Vol. 90, 33 found that the statistics did not establish that the three selected “practices” caused those disparities.100 Thus, although the court couched its decision in terms of the plaintiff’s failure to identify specific practices, its real concern seemed to reflect the issue addressed in the 1991 CRA regarding whether or when “bottom line” disparities could be subject to disparate impact review.101 Similarly, in Anderson v. Douglas & Lomason Co., the court found that the plaintiffs’ challenge to a “practice” of requiring applicants to fill out job applications at the plant failed because plaintiffs had not identified a “specific aspect” of the employers’ practice.102 But, as in Pouncy, the court’s real concern was causation: the court explained that the plaintiffs had failed to establish a “causal connection” between the alleged policy and the racial imbalance with which the plaintiffs were concerned.103 Another group of decisions involved “comparable worth” cases in which courts rejected disparate impact challenges to compensation systems that resulted in gender-based disparities.104 These decisions state that they rejected claims because the challenges did not involve sufficiently specific “employment practices.”105 However, the courts’ reasoning reflected policy concerns about comparable worth and a reluctance to invoke disparate impact law to impose liability when an employer applied longstanding market-based compensation systems.106 For example, the court in American Federation of State, County and Municipal Employees v. Washington, found that the challenge did not involve the type of “specific, clearly delineated employment practice applied at a single point in the job selection 100 Id. at 801–02. 101 See sources cited supra notes 78–80. 102 Anderson, 26 F.3d at 1284. 103 Id. 104 See, e.g., Am. Fed’n of State, Cnty. & Muni. Emps. v. Washington, 770 F.2d 1401 (9th Cir. 1985); Spaulding v. Univ. of Wash., 740 F.2d 686 (9th Cir. 1984). 105 See, e.g., Am. Fed’n, 770 F.2d at 1405–06; Spaulding, 740 F.2d at 707. 106 In another decision that also can be seen as declining to use disparate impact analysis to challenge an employer’s settled policies, the court in Kelber v. Forest Electric Corp., 799 F. Supp. 326 (S.D.N.Y. 1992), rejected a disparate impact claim by an electrician who was fired due to excessive absences because, in part, her claims that the defendant’s “job assignment and termination policies have a disparate impact on pregnant women” was not sufficiently specific. Id. at 333. The court concluded that the plaintiff had failed to identify any “specific” employment practice that affected pregnant women. The court also found that the plaintiff had failed to establish the requisite impact. Id. But it is difficult to imagine how the plaintiff, an electrician in a field and job no doubt dominated by men, could establish that any policy had a disparate impact on women, never mind pregnant women. GOLDSCHEID 10/28/2011 10:31 AM 2011] Disparate Impact’s Impact 53 process,” which, in its view, properly would be the subject of disparate impact challenges.107 It objected to the use of disparate impact theory to “interfere” with complex compensation systems grounded in “surveys, agency hearings, administrative recommendations, budget proposals, executive actions and legislative enactments,” and which were “responsive to supply and demand and other market forces.”108 The court in Spaulding v. University of Washington identified related concerns that allowing disparate impact challenges in comparable worth claims would stretch the theory beyond where it was intended to go.109 That court saw the disparate impact model as a means to challenge particular employment practices that were not obviously job related and was concerned that the theory could be used instead as a means to interfere with the market.110 Regardless of whether the policy judgment that accepted market-based practices fall outside the scope of disparate impact review (presumably because they would not impermissibly enshrine prohibited discrimination) is correct, these cases should be seen as hinging on particular policy concerns rather than on an analysis of the “particular employment action” requirement. Decisions in related contexts reflect a similar judicial reluctance to interfere with established market-based practices. In Finnegan v. Trans World Airlines, Inc., the court rejected an age discrimination claim, concluding that the employer’s financially driven wage and fringe-benefits cuts were not the type of employment practices disparate impact was designed to address.111 In the court’s opinion, Judge Posner described the policies subject to disparate impact challenge as policies that “gratuitously––needlessly—although not necessarily deliberately, excluded black or female workers from equal employment opportunities.”112 Judge Posner recognized that the theory readily would apply to policies that originally had been adopted for discriminatory purposes, but that had not been changed when the employer eliminated deliberate discrimination.113 He distinguished those cases from the type of across-the-board compensation reductions brought on by financial adversity that were 107 Am. Fed’n, 770 F.2d at 1405. 108 Id. at 1406. 109 Spaulding, 740 F.2d at 707. 110 Id. 111 967 F.2d 1161 (7th Cir. 1992). 112 Id. at 1164. 113 Id. GOLDSCHEID 10/28/2011 10:31 AM 54 OREGON LAW REVIEW [Vol. 90, 33 not the legacy of deliberate age discrimination.114 In this sense, the court explicitly drew on one of the identified rationales of disparate impact theory—eliminating the effects of past discrimination—to reject a claim it found outside the doctrine’s historic scope. That reasoning would not preclude claims based on a single decision that perpetuated historically discriminatory biases. Other pre-1991 CRA cases have rejected challenges based on determinations that the policy could not be attributed to actions that were not employer initiated. These decisions can be seen as related to those discussed above concerning causation.115 They draw on the concept, embedded in the 1991 CRA, that the “policy” must be driven by the employer, rather than other factors such as historical practice, or the market, to be subject to disparate impact review. But courts apply this concept in a highly formalistic way. For example, in Equal Employment Opportunity Commission (EEOC) v. Chicago Miniature Lamp Works, the court found that the employer’s passive reliance on employees’ word of mouth for recruitment was not a “particular employment practice” within the statutory meaning because it relied on the employees’ rather than the employer’s actions.116 Similarly, in Beard v. Whitley County REMC, the court rejected a challenge to differential wages and benefits earned by two categories of employees because the compensation packages were found to be the result of separate negotiation processes with different unions, not any affirmative steps taken or practices undertaken by the employer.117 Although the reasoning identifies a credible concern that challenged decisions should stem from employer-driven acts, these decisions draw an artificial line between employer and employee- or union- driven conduct, when the policies might instead be traced to the employer’s actions or acquiescence.118 Notably, other courts have taken different approaches to similar claims. For example, in contrast to EEOC v. Chicago Miniature Lamp Works, the Fourth Circuit Court of Appeals in Thomas v. Washington County School Board, recognized that policies and practices of nepotism and word-of-mouth 114 Id. 115 See, e.g., sources cited supra notes 98–103. 116 947 F.2d 292 (7th Cir. 1991). 117 840 F.2d 405 (7th Cir. 1988). 118 See Travis, supra note 20, at 42–43 (criticizing the reasoning of EEOC v. Chicago Miniature Lamp Works). GOLDSCHEID 10/28/2011 10:31 AM 2011] Disparate Impact’s Impact 55 hiring served to “freeze the effects of past discrimination” and had a disparate impact based on race.119 A few other decisions may prove problematic for a domestic violence survivor because they explicitly found alleged “practices” that affected only a single or a few individuals inappropriate for disparate impact review. Some decisions reflexively state that disparate impact claims must be directed towards “systemic” results of employment practices.120 At least one court indicated that a “discretionary” decision to deny a new mother leave would not be “appropriate” for disparate impact review.121 These cases seem to reflect the view that disparate impact review is only suitable for explicit, formal policies—a view that was rejected in Watson v. Fort Worth Bank and Trust.122 On close review, however, the facts of those cases seem to reflect the courts’ judgment that the claims lacked merit rather than on a categorical judgment whether a “single decision” might ever be subject to a successful disparate impact challenge. For example, in Harper v. Godfrey Co., which involved circumstances surrounding and following a strike by unionized workers, the court rejected a challenge alleging that the employer’s creation of a seniority list had a racially disparate impact.123 The court first reasoned that the composition of a seniority list cannot be called a “policy or practice” for disparate impact purposes because it “was a single decision.”124 The court insisted that to be subject to disparate impact review, the plaintiffs had to establish that the “manner” in which the list is prepared constitutes a policy that has a disparate impact.125 The court dismissed the policies identified by the plaintiffs but found that even if the court recognized the alleged policies for purposes of the challenge, their “limited magnitude” would nevertheless doom the claims.126 Similarly, in Wright v. National Archives and Records Service, the court rejected a claim by an African American civil service employee 119 915 F.2d 922, 925–26 (4th Cir. 1990). 120 See Harper v. Godfrey Co., 839 F. Supp. 583, 604 (E.D. Wis. 1993), aff’d in part, rev’d in part on other grounds, 45 F.3d 143 (7th Cir. 1995); Wright v. Nat’l Archives & Records Svc., 609 F.2d 702, 712 (4th Cir. 1979). 121 Barrash v. Bowen, 846 F.2d 927, 931 (4th Cir. 1988). 122 See Watson v. Fort Worth Bank & Trust, 487 U.S. 977, 991 (1988). 123 Harper, 839 F. Supp. at 604. 124 Id. 125 Id. 126 Id. at 605. GOLDSCHEID 10/28/2011 10:31 AM 56 OREGON LAW REVIEW [Vol. 90, 33 that a series of incidents relating to the agency’s training programs discriminated on the basis of race.127 The court found that the various actions that allegedly led to the plaintiff’s denial of promotion were not the type of facially neutral “policy or practice” subject to disparate impact challenge.128 The court was concerned that of the four employees allegedly affected by the “practices,” only one was alleged to have suffered any harm that was allegedly discriminatory.129 The decision went on to reject the plaintiff’s disparate treatment claims notwithstanding a dissenting opinion that would have upheld plaintiff’s claims on that theory, and concluded by characterizing the grievance as growing “out of a positively motivated intention to better the employment conditions of minority employees.”130 Thus, in addition to the court’s concern that no adverse impact was established, the court also appeared unwilling to hold an employer liable for what the court deemed to be good faith efforts to diversify the workplace.131 Likewise, in Bramble v. American Postal Workers Union, a case sometimes referred to for the proposition that disparate impact does not apply to decisions that affect a single employee,132 the court rejected an age discrimination claim by a former union president who alleged that a newly adopted salary structure that had eliminated his salary had an age-based disparate impact.133 The court was troubled both that there was no indication that the plaintiff’s age was a factor in the decision, whether stated or unstated, and that the plaintiff could not establish that others would be similarly affected.134 The court seemed to be concerned that the employer’s decision was not sufficiently linked—either through veiled intent or impact—to the protected category of age. At least two other decisions state that they reject sex-based disparate impact claims because the challenged employment decisions were “isolated incidents.” In Ilhardt v. Sara Lee Corp., the court 127 609 F.2d 702, 706 (4th Cir. 1979). 128 Id. at 711–13. 129 Id. at 712. 130 Id. at 718. 131 This concern seems misplaced because an employer’s good faith efforts typically are not considered as part of a disparate impact claim, which addresses discriminatory impact rather than intent. 132 See MICHAEL J. ZIMMER ET AL., CASES AND MATERIALS ON EMPLOYMENT DISCRIMINATION 346 (6th ed. 2003). 133 135 F.3d 21, 23 (1st Cir. 1998). 134 Id. at 26. GOLDSCHEID 10/28/2011 10:31 AM 2011] Disparate Impact’s Impact 57 rejected a challenge by a pregnant female attorney who was also the only part-time employee, that the elimination of her position had a sex-based disparate impact.135 The court first rejected her disparate treatment claim because her position was terminated before her employer learned that she was pregnant.136 The court then rejected her disparate impact claim based both on the conclusion that the decision to eliminate her position was only an isolated, one-time decision and not an employment practice, and on the plaintiff’s failure to prove that an alleged policy of terminating part-time employees would have a disparate impact on women.137 According to the court, the plaintiff’s cited studies, showing that the majority of part-time workers were women with child-care responsibilities, were not sufficiently “indisputable” for the court to take judicial notice of them.138 The decision has been criticized as improperly relieving the employer of any obligation to demonstrate that its criteria were consistent with business needs.139 Its limited reasoning—both with respect to the reason why the decision doesn’t qualify as a sufficiently specific practice and why the proffered studies were rejected without analysis—offers little guidance for future courts. Similarly, in Wynn v. Columbus Municipal Separate School District, the court rejected a female athletic coach’s challenge to her school board’s decision to combine the position of head football coach and athletic director because the decision effectively precluded women from the position.140 The court rejected the plaintiff’s proffered statistical proof based on the conclusion that the job description would have a disparate impact on both men and women because both men and women would not be qualified for the position as it had been structured.141 In its focus on the statistical deficiencies, the court stated that “discriminatory impact cannot be established by one isolated decision such as the one involved here.”142 Because she was the only female coach who had ever applied, she could not establish that the requirement would have the actual effect of 135 118 F.3d 1151, 1152 (7th Cir. 1997). 136 Id. at 1156. 137 Id. at 1156–57. 138 Id. 139 Travis, supra note 20, at 43; see also, e.g., Grossman, supra note 21, at 617 (critiquing the decision). 140 692 F. Supp. 672, 682–83 (N.D. Miss. 1988). 141 Id. 142 Id. at 684. GOLDSCHEID 10/28/2011 10:31 AM 58 OREGON LAW REVIEW [Vol. 90, 33 excluding a disproportionate number of women.143 The court required a comparison of the female coaches who had applied to and been rejected for the position of athletic director because of their lack of qualifications for the job of head football coach.144 Thus, even though the court stated that it rested its decision, at least in part, on the plaintiff’s failure to identify a “particular employment practice,” the court’s reasoning focused more on the perceived lack of disparate impact.145 One could imagine the “practice” being framed in terms of the changed job description combining the football coach and athletic director positions. So framed, a court might have reached a different result. Notably, the court upheld the plaintiff’s disparate treatment claim.146 Thus, despite a number of decisions involving “single employment decisions” that seemingly reject cases based on the failure to satisfy the “particular employment practice” requirement, in fact, only a few cases actually stand for the premise that the “practice” must affect more than a single employee, and even those may be explained by other concerns. Nevertheless, courts have interpreted the “particular employment practice” requirement narrowly in related contexts even when not addressing whether the challenged practice constituted a “single employment decision.” Grossman and Travis detail courts’ narrow interpretations of the “particular employment practice” requirement in a number of cases with claims involving “caregiving”- related employment decisions.147 Some of these decisions insulate traditional practices, particularly practices involving leaves, from disparate impact review.148 But other courts have recognized similar leave policies as “particular employment practices” that had a sex- 143 Id. 144 Id. at 683. 145 See discussion of proper comparison groups infra Part I.C. 146 Wynn, 692 F. Supp. at 686. 147 See Grossman, supra note 21, at 617–18; Travis, supra note 20, at 39–46. 148 Travis, supra note 20, at 39–46. For example, in Stout v. Baxter Healthcare Corp., the court characterized the employer’s policy offering employees with more than three absences during a ninety-day probationary period (which disproportionately affected pregnant women) as a challenge to the absence of sufficient leave. 282 F.3d 856, 861 (5th Cir. 2002). In the court’s view, the leave policy would produce an effect that was contrary to the Pregnancy Discrimination Act’s explicit refusal to guarantee leave to pregnant employees, and the court refused to recognize a policy that it would subject to disparate impact review. Id. at 861–62; see also Grossman, supra note 21, at 617; Travis, supra note 20, at 43. GOLDSCHEID 10/28/2011 10:31 AM 2011] Disparate Impact’s Impact 59 based disparate impact.149 Rather than reflecting a reasoned analysis of the “single employment decision” requirement, some of these decisions seem to turn on courts’ views that upholding the challenge would be tantamount to requiring leave, which, in some courts’ views, is contrary to the Pregnancy Discrimination Act (PDA), and which would be a “warrant for favoritism.”150 Whether or not that judgment is correct, either as a normative matter or as a matter of interpretation of the PDA’s reach, the decisions illustrate how disparate impact decisions may instead reflect substantive policy judgments when the claims invoke controversial national policy issues such as comparable worth151 or mandatory parenting leave. Cases involving adverse actions taken against domestic violence victims should not invoke those same concerns.152 C. Establishing Impact The next question is how a court would evaluate whether a claim establishes the requisite statistical impact. Disparate impact theory arose, and is most often used, in the context of exclusionary hiring practices.153 Accordingly, in analyzing impact, courts generally compare the percentage of employees in the protected class in the workplace with the percentage of qualified individuals in the relevant population pool.154 Indeed, employers are required to maintain records of the impact of their “tests and other selection procedures” on employment opportunities for those in protected classes.155 The so-called “four-fifths rule” provides that a selection rate for any 149 See, e.g., Williams & Bornstein, supra note 21, at 1344 n.228 (citing favorable results in cases challenging an employer lifting requirement of 150 pounds, a police department policy allowing light duty only for on-the-job injuries, and a policy that employees could not use sick days to care for sick children). 150 Grossman, supra note 21, at 616 (noting cases that state the Pregnancy Discrimination Act does not provide preferential treatment); see, e.g., Stout, 282 F.3d at 861–62 (rejecting a challenge to an absenteeism policy reasoning that to uphold challenge would require special leave for pregnant employees). 151 See cases cited supra notes 104–10. 152 Indeed, at least some courts have recognized that an employer’s adverse action against a domestic violence survivor would violate public policy. See sources cited supra note 30. 153 Calaf, supra note 36, at 187; see also, e.g., Selmi, supra note 1, at 733–53 (arguing that disparate impact theory has not been successful outside of cases challenging employment tests). 154 See, e.g., LEX K. LARSON, 2 EMPLOYMENT DISCRIMINATION § 22.02 (Lexis 2d ed. 2011). 155 29 C.F.R. § 1607.4 (2010). GOLDSCHEID 10/28/2011 10:31 AM 60 OREGON LAW REVIEW [Vol. 90, 33 protected group that is less than four-fifths of the rate for the group with the highest selection rate will generally be regarded as evidence of adverse impact.156 Federal regulations directing employers how to comply with antidiscrimination law’s disparate impact prohibition recognize that smaller differences in selection rates may constitute an adverse impact when the difference is significant statistically or practically, or when the employer’s actions have disproportionately discouraged applicants.157 Similarly, greater differences in selection rates may not constitute adverse impacts where the differences are based on small numbers or where other factors, such as special recruiting, impact the numbers.158 This statistical calculation applies most readily to hiring tests and explicit job requirements in which the percentage of employees who identify as members of a particular protected class can be compared against the relevant labor pool. With an employment practice such as a decision to terminate, transfer, or demote an individual worker, the comparison works somewhat differently. Courts that have addressed these cases have taken different approaches, but their results vary and often employ limited analyses. The Supreme Court has made clear that no rigid mathematical formula is required in order to establish a disparate impact.159 Instead, courts are directed to judge the significance of the disparity flexibly on a case-by-case basis.160 Some courts have permitted proof without relying on statistics.161 Others explicitly recognize that cases in which there is a “paucity” of statistics may nevertheless reflect impermissible disparate impact discrimination.162 Contrary to the Supreme Court’s authority, however, some decisions state that the plaintiff must present a statistical showing of impact,163 156 Id. § 1607.4(D). 157 Id. 158 Id. 159 Watson v. Fort Worth Bank & Trust, 487 U.S. 977, 994–95 (1988). 160 Id. at 995. 161 Shoben, supra note 2, at 606 (citing Garcia v. Woman’s Hosp. of Texas, 97 F.3d 810, 814 (5th Cir. 1996)); accord Angela Onwuachi-Willig, Another Hair Piece: Exploring New Strands of Analysis Under Title VII, 98 GEO. L.J. 1079, 1123 & n.224 (2010). 162 Thomas v. Wash. Cnty. Sch. Bd., 915 F.2d 922, 926 (4th Cir. 1990) (upholding a disparate impact challenge to nepotism and word-of-mouth job posting practices). 163 Grossman, supra note 21, at 618 (citing cases). The cases raise the question whether courts insist on statistical evidence because it is doctrinally required or because it affords a justification for rejecting a claim about which the court is unsympathetic. For example, in Lang v. Star Herald, the court rejected the plaintiff’s disparate impact claim because the plaintiff provided no statistical support for her challenge to the newspaper’s policy of GOLDSCHEID 10/28/2011 10:31 AM 2011] Disparate Impact’s Impact 61 and some seem to require a workplace large enough to establish a statistically significant claim.164 In the domestic violence context, one question will be whether a plaintiff must establish proof of a disparate impact on her workplace or whether proof of impact on the general population will suffice.165 Some courts evaluate the impact of the decision on the protected class generally.166 Courts seem to be most willing to use general population statistics in cases in which the decision is based on some characteristic that is closely related to membership in the protected class, what might be called a “class-linked” characteristic. For example, courts in some of the cases involving “no-beard” policies have recognized that those policies would have a disparate impact on African American males, given their vulnerability to pseudofolliculitis barbae (PFB).167 A number of courts have upheld disparate impact challenges to “no-beard” policies based on evidence of how that policy would affect the class (African American males) generally.168 However, at least one court required proof that the “no- allowing indefinite unpaid leaves of absence but with no guarantee that the employer would hold open the employee’s position. 107 F.3d 1308, 1314 (8th Cir. 1997). Like cases described in notes 148 and 150, supra, the court may have been unwilling to issue a decision it viewed as requiring an employer to offer a particular type of leave. Similarly, in Maganuco v. Leyden Community High School District 212, the court ostensibly required statistics but did so in the course of rejecting a claim that, in the court’s view, would require the employer to extend the maternity leave it already offered. 939 F.2d 440, 443– 44 (7th Cir. 1991). 164 See, e.g., Lang, 107 F.3d at 1314 (noting employee’s concession that employer was too small for statistical analysis); Council 31, Am. Fed’n of State, Cnty. & Mun. Emps. v. Ward, 978 F.2d 373, 378 (7th Cir. 1992) (noting the potential difficulty in establishing the requisite statistical impact in cases involving “single decisions”). 165 See Flagg, supra note 42, at 2026–27 (recognizing the issue); Grossman, supra note 21, at 618 (citing cases). 166 See, e.g., Bradley v. Pizzaco of Neb., Inc., 939 F.2d 610, 612–13 (8th Cir. 1991) (upholding a disparate impact claim involving a “no-beard” policy based on studies and expert testimony that policy discriminates against black males who disproportionately suffer from skin disorder brought on by shaving); accord Abraham v. Graphic Arts Int’l Union, 660 F.2d 811, 819 (D.C. Cir. 1981) (upholding a disparate impact claim concerning a pregnancy-related leave without analysis of statistical impact in plaintiff’s workplace). 167 PFB is a skin disorder resulting from ingrown hairs when people with certain kind of hair are clean shaven. If affects virtually only African American males. See, e.g., Equal Emp’t Opportunity Comm’n v. Trailways, Inc., 530 F. Supp. 54, 56 (D. Colo. 1981) (discussing medical evidence). 168 Equal Emp’t Opportunity Comm’n v. Trailways, Inc., 530 F. Supp. at 56–59 (rejecting the argument that statistics regarding percentage of African American men in defendant’s workforce would preclude disparate impact claim based on “no-beard” policy); accord Bradley v. Pizzaco of Neb., Inc., 7 F.3d 795, 796 (8th Cir. 1993) (citing the district court’s conclusion that a “no-beard” policy has a disparate impact on African American males); Johnson v. Memphis Police Dep’t, 713 F. Supp. 244, 247 (W.D. Tenn. GOLDSCHEID 10/28/2011 10:31 AM 62 OREGON LAW REVIEW [Vol. 90, 33 beard” policy had a disparate impact in the defendant’s workplace itself. In that case, it may have mattered that the protected group was well represented in the workplace notwithstanding the policy.169 In the case of a class-linked policy such as a “no-beard” policy, requiring proof that the policy has had a disparate impact on the protected group’s representation at a particular workplace would seem to hide the fact that the policy would have a disproportionate negative effect on African Americans overall. Employment decisions based on an employee’s experience with domestic violence would be similarly class linked, given its predominant impact on women. Disparate impact claims brought on the basis of abuse similarly should be analyzed in terms of their impact in the general population rather than in the plaintiff’s workplace itself. Courts have used general population statistics in cases challenging pregnancy or caregiving-related policies when the criteria at issue are arguably class linked. For example, in Roberts v. United States Postmaster General, the court found that a policy prohibiting an employee from using her accumulated sick leave to attend to family members’ medical needs may have a sex-based disparate impact.170 That policy could be seen as class linked, given women’s disproportionate assumption of caregiving responsibilities. In a similar ruling, the court in Chambers v. Omaha Girls Club, Inc. found 1989) (recognizing that a “no-beard” policy would have racial disparate impact unless the employer allowed a medical exception to accommodate PFB); Richardson v. Quik Trip Corp., 591 F. Supp. 1151, 1155 (S.D. Iowa 1984) (same). But see, e.g., Equal Emp’t Opportunity Comm’n v. Greyhound Lines, Inc., 635 F.2d 188, 190–93 (3d Cir. 1980) (rejecting a claim based on the plaintiff’s failure to prove that a “no-beard” policy had a racially discriminatory impact in defendant’s workforce). Even cases rejecting challenges to “no-beard” policies did so based on a finding crediting the defendant’s satisfaction of the business necessity defense rather than the absence of a disparate impact. See, e.g., Fitzpatrick v. City of Atlanta, 2 F.3d 1112, 1118 (11th Cir. 1993) (rejecting a claim based on business necessity grounds, while “assuming” that firefighters adequately pled prima facie case that a “no-beard” policy had racial disparate impact); Stewart v. City of Houston, No. H-07-4021, 2009 WL 2849728 (S.D. Tex. Sept. 3, 2009) (upholding a “no- beard” policy for uniformed officers based on evidence that facial protective equipment could not safely be worn with a beard); Woods v. Safeway Stores, Inc., 420 F. Supp. 35, 42 (E.D. Va. 1976) (recognizing that a “no-beard” policy had “some discriminatory impact” but finding the policy justified by a legitimate business purpose). 169 See, e.g., Equal Emp’t Opportunity Comm’n v. Greyhound Lines, Inc., 635 F.2d at 190–93 (rejecting a claim based on failure to prove that a “no-beard” policy had a racially discriminatory impact in the defendant’s workforce). 170 947 F. Supp. 282, 289 (E.D. Texas 1996). Although the court relied on the general assertion that the policy would have a disparate impact on women due to their “more frequent role as child-rearers,” the court did not elaborate on the type of proof the plaintiff would be required to provide to ultimately prevail. GOLDSCHEID 10/28/2011 10:31 AM 2011] Disparate Impact’s Impact 63 that a rule barring pregnancies in single employees would have a gender-based disparate impact based on higher fertility rates among black women in the population.171 The explicit distinction based on the status of being unmarried and pregnant could be seen as a class- linked criterion that should be evaluated in terms of its impact on the general population, not on a particular workforce. Similarly, other courts have used general population statistics in recognizing that pregnancy and caregiving-related policies might have a disparate impact.172 At the same time, other courts have used the plaintiff’s workplace as a comparator and nevertheless upheld claims.173 On the other hand, some courts, particularly in cases challenging policies that could be seen as less closely class linked, have rejected claims based on the absence of proof that the policy disproportionately affected members of a protected class in the plaintiff’s workplace.174 As with cases analyzing whether the decision satisfied the “single employment decision” requirement, a close look reveals that some of these decisions seem driven by substantive policy-related concerns rather than by the plaintiff’s failure to establish impact per se.175 For example, some decisions 171 834 F.2d 697, 701 (8th Cir. 1987). 172 For example, one court recognized that a policy requiring nurses to lift over 150 pounds might have a disparate impact on pregnant woman. See Garcia v. Woman’s Hosp. of Tex., 97 F.3d 810, 812 (5th Cir. 1996), aff’d, 143 F.3d 227, 231 (5th Cir. 1998) (rejecting the claim based on a lack of evidence that no pregnant woman could lift 150 pounds). In another case, the court recognized that a policy limiting temporary employees’ leave to ten days affected women more severely than men. Abraham v. Graphic Arts Int’l Union, 660 F.2d 811, 819 (D.C. Cir. 1981). 173 See, e.g., Equal Emp’t Opportunity Comm’n v. Warshawsky & Co., 768 F. Supp. 647, 651–55 (N.D. Ill. 1991) (holding that a policy requiring employees to work for one year before taking paid sick leave had a disparate impact on pregnant women); Germain v. County of Suffolk, No. 07-CV-2523, 2009 U.S. Dist. LEXIS 45434 (E.D.N.Y. May 29, 2009) (holding that a policy limiting light-duty assignments only to officers who suffer occupational injuries had a disproportionate impact on pregnant women). 174 See, e.g., Kelber v. Forest Elec. Corp., 799 F. Supp. 326, 333 (S.D.N.Y. 1992) (rejecting a challenge to absenteeism and job assignment policies by an electrician who was fired for excessive pregnancy-related absences both because she failed to identify a specific policy and because she failed to establish impact); Davidson v. Franciscan Health Sys., 82 F. Supp. 2d 768, 774–75 (S.D. Ohio 2000) (rejecting the challenge to a twenty- six-week leave policy by an employee who already had taken a leave absent evidence that any other employee who had been terminated for exceeding the policy was pregnant); see also, e.g., Ruth Colker, Anti-Subordination Above All: Sex, Race, and Equal Protection, 61 N.Y.U. L. REV. 1003, 1034 (1986) (criticizing disparate impact theory for requiring proof of impact on a particular employer’s workplace over a limited time period); Porter, supra note 36, at 294–95 (recognizing cases requiring workplace statistics); Tarr, supra note 36, at 393–94 (same). 175 See supra notes 106–14 and accompanying text. GOLDSCHEID 10/28/2011 10:31 AM 64 OREGON LAW REVIEW [Vol. 90, 33 suggest an unwillingness to use disparate impact doctrine to disturb settled workplace policies such as light-duty assignments.176 Others reflect a view that granting the plaintiff’s claim would effectively authorize “preferential treatment” for pregnant employees, a policy determination Congress explicitly rejected in the PDA.177 Other courts have recognized, or have assumed for purposes of analysis, that the plaintiff may have established a prima facie case, but found that the employer nevertheless established a business necessity.178 Overall, this body of decisions offers little analysis of the proper statistical comparison courts should use when cases invoke statistics. One of the most extensive discussions can be found in Equal Employment Opportunity Commission v. Warshawsky & Co., in which the court considered a challenge that the employer’s policy requiring all employees to work at least one year before they were eligible for sick leave had a disparate impact based on pregnancy and sex.179 The court rejected the statistical approaches urged by both parties and concluded that the applicable comparison was the percentage of females terminated because of the policy during their first year with the percentage of males similarly terminated during that same period.180 That approach seems reasonably suited to the 176 See, e.g., Spivey v. Beverly Enters., Inc., 196 F.3d 1309, 1314 (11th Cir. 1999) (rejecting a nurse’s assistant’s challenge to a policy limiting modified duty to employees who suffered from work-related injuries absent evidence that the policy had a disproportionate impact on pregnant employees); Dimino v. N.Y.C. Transit Auth., 64 F. Supp. 2d 136, 157–58 (E.D.N.Y. 1999) (rejecting a challenge to a policy prohibiting light duty for medically limited employees absent evidence that women transit workers lose more work time than men due to policy); Woodard v. Rest Haven Christian Servs., No. 07- C-0665, 2009 U.S. Dist. LEXIS 21086 (N.D. Ill. Mar. 16, 2009) (rejecting a nurse’s challenge to a policy limiting light-duty work to those who are disabled on the job because the plaintiff did not present evidence of impact at her place of employment). 177 See, e.g., Armstrong v. Flowers Hosp., Inc., 33 F.3d 1308, 1314–16 (11th Cir. 1994); Maganuco v. Leyden Cmty. High Sch. Dist., 939 F.2d 440, 443–44 (7th Cir. 1991) (rejecting the challenge to a policy requiring teachers to use either sick leave or maternity leave based on the plaintiff’s failure to establish that women who have been disabled due to pregnancy accumulate sick days at a greater rate annually than male coworkers or women who have not experienced pregnancy-related disability; concluding that the effect instead was to prevent women who choose to remain at home after the end of pregnancy- related disability from using sick days to cover periods of disability); see also supra note 163 and accompanying text. 178 See, e.g., Ahmad v. Loyal Am. Life Ins. Co., 767 F. Supp. 1114, 1115 (S.D. Ala. 1991); Porter v. Kansas, 757 F. Supp. 1224, 1230 (D. Kan. 1991). 179 768 F. Supp. 647 (N.D. Ill. 1991). 180 Id. at 654. The Equal Employment Opportunity Commission argued that the court should compare the percentage of pregnant first-year employees who were discharged because of the policy with the percentage of all nonpregnant first-year employees who were discharged because of the policy. Id. at 651. By contrast, Warshawsky argued that GOLDSCHEID 10/28/2011 10:31 AM 2011] Disparate Impact’s Impact 65 particular policy involved, which was explicit and which would operate affirmatively to terminate existing employees. But in cases of single decisions that are either explicitly based on or have an impact based on a class-linked characteristic, such as a “no-beard” policy,181 a policy banning braided, locked, or twisted hairstyles,182 or a decision to terminate an employee because she has survived domestic violence, an analysis of the impact of a decision on the plaintiff’s workplace itself would not shed light on its overall effect. In those cases, an evaluation of the impact of the decision on the population at large seems the better approach. D. Impact of Impact Analysis By analyzing how disparate impact doctrine would apply to a particular set of cases, this Article does not urge an expansion of the doctrine; it simply evaluates infrequently explored aspects of the doctrine and argues that some courts have interpreted statutory requirements in an unduly restricted way. The impact of the analysis likely will be limited, given the relatively specific nature of the claim and its context. A disparate impact challenge brought by a domestic violence survivor is different from many disparate impact cases in that the decision does not involve an explicit policy. In that sense, it differs, for example, from those cases involving pregnant women and bearded men,183 dress codes,184 or cases involving “English-only” rules,185 because those challenges involve facially neutral but expressly stated policies, whereas a domestic violence survivor may be told that she was terminated because of the abuse, but the the court should compare the number of pregnant first-year employees who required sick leave with the nonpregnant first-year employees who required sick leave. Id. at 652. This case illustrates the difficulty of identifying the appropriate comparison groups. 181 See supra notes 167–69 and accompanying text. 182 See Onwuachi-Willig, supra note 161, at 37–39. 183 See, e.g., Selmi, supra note 1, at 749–53 (discussing cases); see also, e.g., Travis, supra note 20, at 39–46 (discussing disparate impact challenges to caregiving-related inequalities). 184 See, e.g., Jespersen v. Harrah’s Operating Co., 444 F.3d 1104 (9th Cir. 2006). 185 See, e.g., Garcia v. Spun Steak Co., 998 F.2d 1480 (9th Cir. 1993); see also, e.g., Christopher David Ruiz Cameron, How the García Cousins Lost Their Accents: Understanding the Language of Title VII Decisions Approving English-Only Rules as the Product of Racial Dualism, Latino Invisibility, and Legal Indeterminacy, 85 CALIF. L. REV. 1347 (1997) (discussing cases). GOLDSCHEID 10/28/2011 10:31 AM 66 OREGON LAW REVIEW [Vol. 90, 33 employment decision is unlikely to involve an officially stated policy.186 Of course, the decision to terminate a domestic violence survivor may be seen as similar to other challenges that generally have not succeeded. For example, a claim by a survivor who was terminated due to excessive absenteeism (related to the abuse) may be seen as a challenge to the employer’s (gender-neutral) absenteeism policy. That challenge could be seen as similar to challenges to absenteeism policies brought by employees seeking pregnancy or childcare leave. However, cases involving domestic violence survivors generally should not invoke the same policy-based concerns to which courts have alluded in those cases.187 Alternately, claims by domestic violence survivors may be seen as similar to the often unsuccessful challenges brought to dress codes or “English-only” policies because those cases may be seen as involving issues of the employee’s choices.188 Regardless whether that distinction is correct, it should not apply to cases involving adverse actions against domestic violence survivors, because in those cases the employee’s circumstances are caused by the abuser’s, not the employee’s, conduct.189 CONCLUSION This Article has evaluated the question of whether adverse employment actions taken against survivors of domestic violence should be subject to disparate impact challenges. The Article reviews objections that may be raised and concludes that those arguments should not bar claims outright. The issue offers one example of ways in which cramped doctrinal interpretations contribute to a perception that disparate impact theory is unavailable, particularly outside the testing context. This Article challenges that perception. The Article shows how adverse employment actions taken against domestic 186 Of course, a domestic violence survivor who is terminated for the ostensible reason that she violated a standing workplace rule, for example, that she had excessive absences, could bring a disparate impact challenge to the impact of that policy on domestic violence survivors. Framed that way, she likely would have difficulty establishing that the overall impact of that policy had a discriminatory effect, both because of the relatively small percentage of survivors likely to be affected and because the employer likely would successfully defend the business necessity of a generic absentee policy. 187 See supra notes 147–50 and accompanying text. 188 See Flagg, supra note 42, at 2028. 189 Indeed, any decision based on the employee’s role in the abuse, absent additional facts, would perpetuate a stereotype of its own. GOLDSCHEID 10/28/2011 10:31 AM 2011] Disparate Impact’s Impact 67 violence survivors fall within the scope of neutral employment decisions the Court and Congress have recognized may perpetuate impermissible discrimination. GOLDSCHEID 10/28/2011 10:31 AM 68 OREGON LAW REVIEW [Vol. 90, 33