Content uploaded by Michael Selmi
Author content
All content in this area was uploaded by Michael Selmi on May 28, 2016
Content may be subject to copyright.
THE GEORGE WASHINGTON UNIVERSITY LAW SCHOOL
PUBLIC LAW AND LEGAL THEORY WORKING PAPER NO. 188
Was the Disparate Impact
Theory a Mistake?
Michael Selmi
This paper can be downloaded free of charge from the
Social Science Research Network at:
http://ssrn.com/abstract=881606
701
WAS THE DISPARATE IMPACT THEORY A MISTAKE?
Michael Selmi*
The disparate impact theory long has been viewed as one of the most
important and controversial developments in antidiscrimination law. In this
Article, Professor Selmi assesses the theory’s legacy and challenges much of the
conventional wisdom. Professor Selmi initially charts the development of the
theory, including a close look at Griggs v. Duke Power Co. and Washington
v. Davis, to demonstrate that the theory arose to deal with specific instances of
past discrimination rather than as a broad theory of equality. In the next
section, Professor Selmi reviews the success of the theory in the courts through
an empirical analysis and concludes that it has had a strikingly limited impact
outside of the context of written employment tests and is, in fact, an extremely
difficult theory on which to succeed. In the final section, Professor Selmi contends
that whatever gains the disparate impact theory has produced likely could have
been obtained through other means, particularly in large urban cities, and that
the theory may have had the unintended effect of limiting our conception of
intentional discrimination. Disparate impact theory always has been seen as
beginning where intentional discrimination ends, and by pushing an expansive
theory of impact, we were left with a truncated theory of intentional
discrimination that continues to turn on animus and motive. Rather than a new
legal theory of discrimination, Professor Selmi concludes, a greater societal
commitment to remedying inequities was needed, as the ultimate mistake behind
the disparate impact theory was the belief that legal theory could do the work that
politics could not.
INTRODUCTION .............................................................................................................702
I. THE ROAD TO GRIGGS AND BEYOND .....................................................................708
A. The Seniority Cases and the Origins of the Disparate Impact Theory..........708
B. Testing Cases and the EEOC.........................................................................714
C. The Griggs Decision .......................................................................................717
* Professor of Law, George Washington University Law School. This Article was
completed while I was a Visiting Professor at Harvard Law School where I received excellent
research assistance from Courtney Chai and Michael Stein. At George Washington University, I
also benefited by terrific research assistance from Kim Sikora and Terra Nevitt. Many friends and
colleagues have offered generous comments on prior drafts, including Dean Fred Lawrence, Paul
Butler, Charlie Craver, Wendy Parker, Michael Van Alstine, Marion Crain, Pauline Kim,
Michael Stein, and Charlie Sullivan, and I am particularly indebted to Christine Jolls for
discussions and comments. Earlier versions of this Article were presented at the Annual Law and
Society Conference, and at faculty workshops at Boston University and George Washington
University Law School where I also received valuable feedback.
702 53 UCLA LAW REVIEW 701 (2006)
D. Washington v. Davis and the Turn to Intent ..................................................725
II. ASSESSING THE THEORY IN THE COURTS...............................................................734
A. The Scope of the Study..................................................................................734
B. The Success of the Disparate Impact Theory ................................................738
1. Empirical Assessment .............................................................................738
2. Subjective Employment Practices and Age Cases..................................744
3. Pregnant Women and Bearded Men ......................................................749
III. ASSESSING THE EFFECT OF THE DISPARATE IMPACT THEORY ................................753
A. The Gains of the Disparate Impact Theory ...................................................755
1. Testing Cases as Intentional Discrimination Claims .............................757
2. Politics and the Disparate Impact Theory..............................................763
B. The Mistake in the Disparate Impact Theory................................................767
1. The Disparate Impact Theory Is Easier to Prove....................................768
2. The Law Should Move Away From a Focus on Blame...........................773
3. The Limitations of Intent.......................................................................776
CONCLUSION.................................................................................................................782
INTRODUCTION
Within antidiscrimination law, no theory has attracted more attention
or controversy than the disparate impact theory, which allows proof of
discrimination without the need to prove an intent to discriminate. The
general outlines of the controversy are well known. In the 1971 landmark
decision of Griggs v. Duke Power Co.,1 the U.S. Supreme Court unanimously
approved of the theory in the context of statutory employment
discrimination claims.2 Five years later in the equally momentous Washington
v. Davis3 the Court refused to extend the theory to constitutional claims,
holding instead that intentional discrimination is required to establish a
violation of the Equal Protection Clause.4 Both of these cases involved
written employment examinations, but advocates have sought to extend
the theory to virtually every civil rights context under the perception that
the disparate impact theory would reach discrimination that was otherwise
out of reach for claims of intentional discrimination.5 Just last term, the
1. 401 U.S. 424 (1971).
2. Id. at 436 (interpreting Title VII to include disparate impact cause of action).
3. 426 U.S. 229 (1976).
4. Id. at 239.
5. See, e.g., Alexander v. Sandoval, 532 U.S. 275 (2001) (holding that Title VI requires
proof of intentional discrimination); Gen. Bldg. Contractors Ass’n v. Pennsylvania, 458 U.S. 375
(1982) (holding that § 1981 prohibiting discrimination in contracting only applies to claims of
intentional discrimination); City of Mobile v. Bolden, 446 U.S. 55 (1980) (holding that a voting
Was the Disparate Impact Theory a Mistake? 703
Supreme Court applied the cause of action to age discrimination
claims, settling a longstanding dispute in the lower courts.6
Together Griggs and Washington v. Davis are widely seen as two of the
most influential civil rights decisions ever issued. The Griggs decision has
been universally hailed as the most important development in employment
discrimination law.7 Even critics concede its influence and have long sug-
gested that the theory encourages employers to rely on quotas as a means of
avoiding disparate impact lawsuits.8 And when adverse Supreme Court
decisions threatened to eviscerate the Griggs decision, Congress responded
by passing the Civil Rights Act of 1991.9
The reaction to Washington v. Davis has been equally spirited, though
in many respects the polar opposite of that for Griggs. Liberal academics
have denounced the decision as unjustifiably limiting the scope of the Equal
Protection Clause, and Charles Lawrence’s renowned article that helped
rights claim brought under the Fifteenth Amendment required proof of intentional discrimination).
Before the Supreme Court issued its decision in Sandoval, there was some room for plaintiffs to pursue
disparate impact claims under Title VI, a statute that requires nondiscrimination by federal
contractors, pursuant to the regulations that were issued to implement the statute. See Guardians
Ass’n v. Civil Serv. Comm’n, 463 U.S. 582 (1983).
6. Smith v. City of Jackson, 125 S. Ct. 1536 (2005). The battle over applying the
disparate impact theory to age discrimination cases is discussed in Part II.B.2, infra.
7. See, e.g., HUGH DAVIS GRAHAM, THE CIVIL RIGHTS ERA: ORIGINS AND DEVELOPMENT
OF NATIONAL POLICY 1960–1972, at 383–86 (1990) (stating that Griggs “burst like a bombshell”
and discussing its importance); 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, 433
(2005) (labeling Griggs “[a]side from Brown . . . the single most influential civil rights case during the
past forty years”); Alfred W. Blumrosen, The Legacy of Griggs: Social Progress and Subjective
Judgments, 63 CHI.-KENT L. REV. 1, 1–2 (1987) (“Few decisions in our time—perhaps only Brown
v. Board of Educ.—have had such momentous social consequences [as Griggs].”).
8. For example, Richard Epstein, a fierce critic of the theory, has called Griggs “the first
and single most important Supreme Court decision under Title VII . . . .” RICHARD A. EPSTEIN,
FORBIDDEN GROUND S: THE CASE AGAINST EMPLOYMENT DISCRIMINATION LAWS 183 (1992).
Professor Epstein discusses how the force of the disparate impact theory encourages employers to adopt
“implicit quotas.” Id. at 234–36; see also John Hasnas, Equal Opportunity, Affirmative Action, and
the Anti-discrimination Principle: The Philosophical Basis for the Legal Prohibition of Discrimination, 71 FORDH AM
L. REV. 423, 480 (2002) (claiming that disparate impact theory creates a “strong incentive to give
preferential treatment to minorities”). Although the link to quotas has been a particular concern
for conservative critics, the link has long been discussed in the context of the disparate impact
theory. See, e.g., Hugh Steven Wilson, A Second Look at Griggs v. Duke Power Company:
Ruminations on Job Testing, Discrimination, and the Role of the Federal Courts, 58 VA. L. REV. 844,
873 (1972) (noting that “employers may use privately imposed quotas” to avoid disparate impact
liability). For an incisive rebuttal to the quota argument, see Ian Ayres & Peter Siegelman, The
Q-Word as Red Herring: Why Disparate Impact Liability Does Not Induce Hiring Quotas, 74 TEX. L.
REV. 1487 (1996).
9. For a discussion of the importance of the disparate impact provisions to the Civil
Rights Act of 1991, see Neal Devins, Reagan Redux: Civil Rights Under Bush, 68 NOTRE DAME L.
REV. 955, 984–99 (1993).
704 53 UCLA LAW REVIEW 701 (2006)
spawn critical race theory was principally a critique of the Court’s adoption
of an intent standard.10 In many quarters, constitutional law is still taught
as if a different, more meaningful concept of equality would have emerged if
only the Supreme Court had reached a different conclusion in Davis.
Professor Reva Siegel is representative of this position when she writes that
had the disparate impact theory been available, “equal protection litigation
might [have] move[d] the nation closer to disestablishing historic patterns
of race and gender stratification than current constitutional doctrines now do.”11
One of the more interesting aspects of the disparate impact theory is that
its mythology has arisen without any serious exploration of its reality. For
example, in the last several years, scholars have offered numerous proposals
to extend the disparate impact theory to cure all manner of social ills;
extending the disparate impact doctrine has long been one of the primary
obsessions of liberal academics and advocates alike.12 Three prominent
10. See Charles R. Lawrence, III, The Id, the Ego, and Equal Protection: Reckoning With
Unconscious Racism, 39 STAN. L. REV. 317, 323 (1987) (“[R]equiring proof of intentional motivation
as a prerequisite to constitutional recognition that a decision is race-dependent ignores much of what
we understand about how the human mind works.”). Several other critical scholars have staked
their claims with critiques aimed at the intent requirement. See, e.g., Barbara J. Flagg, “Was Blind,
But Now I See”: White Race Consciousness and the Requirement of Discriminatory Intent, 91 MICH. L.
REV. 953, 968 (1993) (“[T]he Davis rule reflects a distinctively white way of thinking about
race.”); Alan David Freeman, Legitimizing Racial Discrimination Through Antidiscrimination Law: A
Critical Review of Supreme Court Doctrine, 62 MINN. L. REV. 1049, 1052 (1978) (critiquing focus
on intent as part of “perpetrator” perspective). Paul Brest, on the other hand, helped solidify his
scholarly reputation with a modest defense of the intent principle. See Paul Brest, Foreword: In
Defense of the Antidiscrimination Principle, 90 HARV. L. REV. 1, 19–22 (1976) (suggesting that
racially disproportionate impact should not constitute unlawful discrimination).
11. Reva Siegel, Why Equal Protection No Longer Protects: The Evolving Forms of Status-
Enforcing State Action, 49 STAN. L. REV. 1111, 1145 (1997).
12. Just in the last few years, the following articles have appeared: Henry L. Chambers, Jr.,
Colorblindness, Race Neutrality, and Voting Rights, 51 EMORY L.J. 1397 (2002) (urging application
of disparate impact theory to literacy tests); Carl H. Coleman, The “Disparate Impact” Argument
Reconsidered: Making Room for Justice in the Assisted Suicide Debate, 30 J.L. MED. & ETHICS 17 (2002)
(assisted suicide); Mary Crossley, Reasonable Accommodation as Part and Parcel of the
Antidiscrimination Project, 35 RUTGERS L.J. 861 (2004) (discussing disparate impact theory as applied
to a disabilities statute); Lara M. Gardner, A Step Toward True Equality in the Workplace: Requiring
Employer Accommodation for Breastfeeding Women, 17 WIS. WOMEN’S L.J. 259 (2002) (discussing
the applicability of the disparate impact model to prohibit discrimination against breastfeeding);
Robert A. Kearney, The Disparate Impact Hostile Environment Claim: Sexual Harassment Scholarship
at a Crossroads, 20 HOFSTRA LAB. & EMP. L.J. 185 (2003) (advocating creation of a disparate
impact hostile environment claim); R.A. Lenhardt, Understanding the Mark: Race, Stigma, and
Equality in Context, 79 N.Y.U. L. REV. 803, 895–98 (2004) (arguing for use of disparate impact
analysis to ferret out racial stigma); Charles A. Sullivan, The World Turned Upside Down?:
Disparate Impact Claims by White Males, 98 NW. U. L. REV. 1505 (2004) (exploring application of the
theory to white male plaintiffs); Michelle A. Travis, Equality in the Virtual Workplace, 24 BERKELEY J.
EMP. & LAB. L. 283 (2003) (advocating use of disparate impact model for gender issues in the
Was the Disparate Impact Theory a Mistake? 705
employment discrimination scholars have recently called for a revival of the
theory,13 and several articles in leading law reviews have explored its
ramifications.14 Yet none of these articles comes to grips with a central
facet of the theory: Outside of the original context in which the theory
arose, namely written employment tests, the disparate impact theory has
produced no substantial social change and there is no reason to think that
extending the theory to other contexts would have produced meaningful
reform. In other words, had Washington v. Davis been decided differently,
the end results would have been pretty much the same. Even with written
tests the theory did not achieve the expected reform, as the vast majority of
tests continue to have significant adverse impact.15
As discussed below, the disparate impact theory arose initially to deal
with specific practices, seniority systems and written tests, that were
perpetuating past intentional discrimination.16 Although courts have never
restricted the theory to those particular contexts, the reality has been that
the theory has proved an ill fit for any challenge other than to written
examinations, the only category of claim for which legal standards have
evolved to evaluate the permissibility of employment practices.17 This latter
point is important and too easily glossed over by both advocates and critics.
While it is true that the disparate impact theory allows proof of
discrimination without the need to prove intent, employers are allowed to justify
their practices under a business necessity test. Because that test allows for
virtual workplace); Martha Chamallas, The Market Excuse, 68 U. CHI. L. REV. 579 (2001) (book
review) (arguing that disparate impact is the most appropriate theory for pay equity claims).
13. See Belton, supra note 7, at 469–72 (speculating on the possible future of the theory); Elaine
W. Shoben, Disparate Impact Theory in Employment Discrimination: What’s Griggs Still Good For? What
Not?, 42 BRANDEIS L.J. 597 (2004) (emphasizing the potential of disparate impact theory); Charles
Sullivan, Re-Reviving Disparate Impact 59 (unpublished manuscript, on file with author) (“The overall
theme of this Article is that future development of the antidiscrimination project should focus far more
on the disparate impact as a theory of liability than on disparate treatment.”).
14. See, e.g., Christine Jolls, Antidiscrimination and Accommodation, 115 HARV. L. REV. 642
(2001) (comparing the accommodation requirement under a disabilities statute to disparate impact
theory); Richard A. Primus, Equal Protection and Disparate Impact: Round Three, 117 HARV. L.
REV. 494 (2003) (exploring the purpose and constitutionality of disparate impact theory).
15. See Paul R. Sackett et al., High Stakes Testing in Employment, Credentialing, and Higher
Education: Prospects in a Post-Affirmative-Action World, 56 AM. PSYCHOLOGIST 302, 302 (2001)
(“In education, employment, and credentialing contexts, test score distributions consistently
reveal significant mean differences by race.”).
16. The two most important formative cases both involved seniority issues. See Local 189,
United Papermakers v. United States, 416 F.2d 980 (5th Cir. 1969); Quarles v. Philip Morris,
Inc., 279 F. Supp. 505 (E.D. Va. 1968). These cases are discussed in Part I.A, infra.
17. Written tests can be validated pursuant to professionally established guidelines. See
Albemarle Paper Co. v. Moody, 422 U.S. 405 (1975) (discussing guidelines). As discussed below,
the guidelines are ill-suited for anything other than written examinations.
706 53 UCLA LAW REVIEW 701 (2006)
normative judgments regarding what practices are properly defined as
discriminatory, courts readily accept most proffered justifications.18
A central reason for courts’ general acceptance of justifications, I will
suggest, is that courts never fully accepted the disparate impact theory as a
legitimate definition of discrimination, or as a legitimate means of proving
discrimination, and it was a mistake to think that they would. The dispa-
rate impact theory has often been justified based on the difficulty of proving
intentional discrimination, particularly in cases where evidence of overt
bias or animus is lacking.19 Yet, there was no reason to believe that courts
would be more willing to see discrimination through the lens of disparate
impact theory when they were unable to do so even through the far more
common mix of circumstantial evidence of intentional discrimination.
Suggesting otherwise was like offering a pair of glasses to cure blindness, and
the reality has been that disparate impact claims are more difficult—not
easier—to prove than claims of intentional discrimination.
But to suggest that the disparate impact theory has produced less change
than typically assumed is a far cry from demonstrating that the theory was a
mistake. On this score, I will argue that the theory had the rather perverse
effect of limiting our conception of intentional discrimination, which, in the
end, may have hindered our efforts to eradicate discrimination more than it
has plausibly helped. As a concept, the disparate impact theory begins where
intentional discrimination ends, and seeking an expansive role for the
disparate impact theory ultimately has left us with a truncated definition of
intentional discrimination. The disparate impact theory has always been
contrasted with racial animus and motive, and despite the familiar refrains
regarding how discrimination has become more subtle over time, we continue
to define intentional discrimination in the context of animus and consciously
impermissible motives.20 The disparate impact theory also has proven a poor
vehicle for uncovering subtle discrimination while the intentional discrimina-
tion framework has remained seriously undeveloped, even though it likely
18. This issue is discussed in more detail in Part II.B, infra.
19. See Mark S. Brodin, Costs, Profits, and Equal Employment Opportunity, 62 NOTRE DAME
L. REV. 318, 358 (1987) (suggesting that the Supreme Court established disparate impact theory
in part based on the difficulty of proving intent); George Rutherglen, Disparate Impact Under Title VII:
An Objective Theory of Discrimination, 73 VA. L. REV. 1297 (1987) (arguing that one justification
for disparate impact theory is the difficulty of proving intent under disparate treatment models).
20. See, e.g., Linda Hamilton Krieger, The Content of Our Categories: A Cognitive Bias
Approach to Discrimination and Equal Employment Opportunity, 47 STAN. L. REV. 1161, 1172
(1995) (noting that in pretext cases involving intentional discrimination, “liability is premised on
the presence of conscious discriminatory animus”). I return to this theme in Part III.B.3, infra.
Was the Disparate Impact Theory a Mistake? 707
could have been expanded to include much of what the disparate impact
theory ultimately captured, while also including more subtle discrimination.
Developing a more expansive concept of intent would have required a
far stronger push to redefine intentional discrimination to encompass acts
that were not tied to old-style discrimination, and it also would have required a
greater societal commitment to remedying racial, gender and other disparities
linked to what is often defined as societal discrimination. Ironically, the
move to the disparate impact theory may have alleviated some of that
perceived need as it sent a signal that intentional discrimination was largely a
thing of the past. At the same time, it moved away from notions of fault or
blame that were necessary to trigger greater social responsibility. With this
framework in mind, it becomes easier to see how the theory was a mistake
and, equally important, why it was premised on a deeply mistaken foundation.
This Article proceeds in three parts. Part I explores the neglected his-
tory of the disparate impact theory, its context, and its evolution in the
Supreme Court culminating in Washington v. Davis, which I suggest was a
poorly developed case for extending the theory because the plaintiffs never
articulated a reason why the city of Washington, D.C. should have been
held responsible for the disparate effects of its test under any theory of dis-
crimination. Part II embarks on a different course, as it presents an empiri-
cal analysis of how the disparate impact theory has fared in both the
appellate and district courts, and I conclude that the theory has had a
strikingly limited impact. One of the more important findings of this case
survey is that a substantial number of successful disparate impact cases also
succeed under a theory of intentional discrimination, suggesting that the
impact theory is often superfluous. I also argue that the theory has rarely
been successfully stretched beyond the testing context. Part III lays out my
argument for why the theory was ultimately a mistake by first demonstrating
that even the benefits produced in the testing cases likely could have been
procured either through a broader definition of intentional discrimination,
one that was available at the time the disparate impact theory arose, or by
ordinary politics. Following that discussion is an analysis of how the disparate
impact theory significantly contributed to our limited understanding of
intentional discrimination.
In addition to revealing the limits and restrictions of the disparate
impact theory, another purpose of this Article is to help revive a more criti-
cal analysis within legal scholarship. The faith so many scholars and
advocates have imbued in the disparate impact theory largely ignores much
of what we have learned about the way in which the law works to preserve
708 53 UCLA LAW REVIEW 701 (2006)
social norms rather than to upend them. Taking seriously the disparate
impact theory would have posed a substantial challenge to existing prac-
tices, which is precisely why the theory never has been taken particularly
seriously by courts. But more than that, one lesson we ought to relearn is
that we cannot expect the law to do the work that politics cannot.
I. THE ROAD TO GRIGGS AND BEYOND
The disparate impact theory is generally associated with the Supreme
Court’s decision in Griggs v. Duke Power Co.21 Somewhat remarkably, Griggs
was only the Supreme Court’s second interpretation of the employment
provisions of the Civil Rights Act of 1964 (Title VII), and that interpretation,
like its first, appeared to greatly expand the scope of the statute.22 At the same
time, the theory did not arise spontaneously, and one of the least chronicled
aspects of an otherwise heavily analyzed case is the origins of the disparate
impact cause of action.23 As discussed in detail below, two important cases, two
influential law review articles, and a strategic decision by the Equal
Employment Opportunity Commission (EEOC) all contributed to the
creation of the theory, which the Court ultimately sanctioned in Griggs.
A. The Seniority Cases and the Origins of the Disparate Impact Theory
Much of the concern that gave rise to the disparate impact theory cen-
tered on the perpetuation of past lawful discrimination through what
appeared to be neutral practices, initially seniority systems and later written
tests that were imposed after Title VII became applicable. Prior to the pas-
sage of the Civil Rights Act of 1964, African Americans were widely segre-
gated into undesirable positions that were located within different job
progression lines from the white workers.24 At the time, many seniority
systems afforded seniority only within particular jobs. As a result, regardless
21. 401 U.S. 424 (1971).
22. Like Griggs, the Supreme Court’s first Title VII decision offered a potentially broad
interpretation of the statute by creating a “sex-plus” claim in which it was possible for plaintiffs to
allege that a defendant discriminated against a subclass of women, in this case women with pre–
school age children. See Phillips v. Martin Marietta Corp., 400 U.S. 542 (1971). Not
coincidentally, similar to Griggs, the sex-plus theory has been significantly limited and is now a
marginal part of discrimination doctrine.
23. Professor Robert Belton, one of the lead attorneys in Griggs, recently offered his own
retrospective, which accords with some of my own treatment though we part company over the
theory’s ultimate influence. See Belton, supra note 7, at 435–54.
24. See William B. Gould, Employment Security, Seniority and Race: The Role of Title VII of
the Civil Rights Act of 1964, 13 HOWARD L.J. 1 (1967) (describing seniority cases and systems).
Was the Disparate Impact Theory a Mistake? 709
of their seniority within a particular company, African Americans had to
start at the bottom of the white job ladder in order to move into the more
desirable positions.25 Complicating matters further, many of the jobs at the
bottom of the white progression paid less than the top black positions, and
many black workers were reluctant to take a pay cut as a way of moving into
the better jobs.26
In one of the most important early cases challenging the discriminatory
effects of seniority systems, Local 189, United Papermakers v. United States,27
Judge Wisdom framed the matter as
how to reconcile equal employment opportunity today with seniority
expectations based on yesterday’s built-in racial discrimination. May
an employer continue to award formerly “white jobs” on the basis of
seniority attained in other formerly white jobs, or must the employer
consider the employee’s experience in formerly “Negro jobs” as an
equivalent measure of seniority?28
As Judge Wisdom noted, unless service experience in the “Negro jobs” was
applied to the now opened white jobs, senior African American employees
would find themselves competing with whites for entry-level jobs, and it
would be many years before they would be able to move up the ranks to bet-
ter positions. If this were to occur, it would contravene the specific purpose
of the Act.
There was, however, an important twist to the Papermakers case that
reveals an underlying motive for the company’s seniority system that could
not be properly ascribed to unintentional discrimination, and also illustrates
why many of the early cases were treated as forms of intentional dis-
crimination. In the Papermakers case, the EEOC initially approved of the
seniority system within job titles for future jobs, but the Department of
Labor’s Office of Federal Contract Compliance, reviewing companies that
had federal contracts, objected to this arrangement as impermissibly per-
petuating the effects of past discrimination.29 As a condition of maintaining
federal contracts, the Department of Labor required the company to com-
bine an employee’s time in job with his entire time at the company, and the
company agreed to this arrangement.30 The union that represented the
25. See Alfred W. Blumrosen, Seniority and Equal Employment Opportunity: A Glimmer of
Hope, 23 RUTGERS L. REV. 268, 276–77 (1969).
26. Id.
27. 416 F.2d 980 (5th Cir. 1969).
28. Id. at 982–83.
29. Id. at 984–85.
30. Id. at 985.
710 53 UCLA LAW REVIEW 701 (2006)
white employees, however, refused to accept the proposal, and voted to
strike once the company indicated that it planned to impose the new sys-
tem unilaterally.31 The Justice Department then filed suit to enjoin the
strike, and perhaps for good measure staked out a third governmental posi-
tion by seeking to enjoin the use of seniority altogether.32
The company, a codefendant in the Justice Department lawsuit, subse-
quently defended its seniority system by focusing narrowly on the absence
of a present intent to discriminate. The company argued that it had ceased
to discriminate once Title VII became effective and “[t]he fact that the
system continues to prefer whites over previously hired Negroes in filling
certain vacancies does not in itself show racial discrimination. That effect,
the defendants argue[d, was] merely an ineradicable consequence of extinct
racial discrimination.”33 As with the practices subsequently challenged in
Griggs, no one asserted that the seniority system had been preserved with
a specific, or primary, intent to discriminate against African American work-
ers, and for this reason there did not appear to be intentional discrimination
as traditionally defined, although later in the opinion the court suggested
that such intent was present.34 The nature of that intent, however, was tied
to the central purpose of Title VII rather than to the motive of the actor.
The purpose of the Act, the court noted, was to provide employment
opportunities previously denied routinely and systematically to African
Americans; allowing job seniority to determine employment opportunities
would continue to “freeze” out the intended beneficiaries of the legislation,
a fact the Supreme Court had previously recognized and sought to remedy
in the voting rights context.35 As Judge Wisdom explained, “It is not
31. See id. at 984–85.
32. Id. at 985. Toward the end of its opinion, the Fifth Circuit noted, “We cannot help
sharing Crown Zellerbach’s bewilderment at the twists and turns indulged in by government
agencies in this case.” Id. at 997.
33. Id. at 986.
34. Id. at 997 (“The requisite intent may be inferred from the fact that the defendants
persisted in the conduct after its racial implications had become known to them. Section 707(a)
demands no more.”).
35. Id. at 987–88, 990–91. With respect to voting rights, the notion of a “freeze out” had
two meanings. On the one hand, entrenched past practices could freeze out African Americans,
even if not implemented with a specific intent to do so. See Louisiana v. United States, 380 U.S. 145
(1965). Conversely, the Fifth Circuit developed what was known as a “freeze theory” as a remedy
for voting rights violations. Under that theory, no new practices could be implemented because of
their discriminatory effects, and courts froze the standards at a particular time as a way of avoiding
new hurdles. See United States v. Duke, 332 F.2d 759 (5th Cir. 1964). In Griggs, the Supreme
Court mentioned the concept, which was also discussed more extensively by the dissenting judge
in the prior appellate decision. See Griggs v. Duke Power Co., 401 U.S. 424, 430 (1971) (“Under
the Act, practices, procedures, or tests neutral on their face, and even neutral in terms of intent,
Was the Disparate Impact Theory a Mistake? 711
decisive therefore that a seniority system may appear to be neutral on its
face if the inevitable effect of tying the system to the past is to cut into the
employees present right not to be discriminated against on the ground of
race.”36 To the court, this was more than remedying the present effects of
past discrimination; it was present discrimination that could be tolerated
only through a legitimate justification. This is where the business necessity
language entered the analysis: “When an employer or union has dis-
criminated in the past and when its present policies renew or exaggerate
discriminatory effects, those policies must yield, unless there is an overrid-
ing legitimate, non-racial business purpose.”37
The Fifth Circuit in Papermakers specifically referred to this business
purpose as a “business necessity,” a concept it borrowed from a pre–Title
VII case involving dual seniority systems where the court had struggled with
a similar issue. In Whitfield v. United Steelworkers of America, Local 2708,38
the employer had maintained two racially separate job classifications but
was required to merge them in light of Supreme Court precedent interpret-
ing the National Labor Relations Act.39 The two job lines, however, “were
not so functionally related that experience at the top of the formerly black
line could provide adequate training for the bottom jobs in the white
line.”40 To comply with the Supreme Court mandate, and ensure that the
employees would be qualified for the jobs, the company required black
employees to take a qualifications test to move into the positions that for-
merly had been held exclusively by white employees, even though the
incumbent white employees were not required to take the test. The Fifth
Circuit upheld the requirement noting that “[s]uch a system was conceived out of
business necessity, not out of racial discrimination. An employee without
proper training and with no proof of potential ability to rise higher, cannot
expect to start in the middle of the ladder, regardless of plant seniority.”41
In contrast to the Whitfield case, the Fifth Circuit saw no business jus-
tification in Papermakers for allocating positions based on job rather than
cannot be maintained if they operate to “freeze” the status quo of prior discriminatory
employment practices.”); Griggs v. Duke Power Co., 420 F.2d 1225, 1247 (4th Cir. 1970)
(Sobeloff, C.J., concurring in part and dissenting in part) (discussing the freeze theory in voting rights).
36. Papermakers, 416 F.2d at 988.
37. Id. at 989.
38. 263 F.2d 546 (5th Cir. 1959), cert. denied, 360 U.S. 902 (1959).
39. See Steele v. Louisville & Nashville R.R. Co., 323 U.S. 192 (1944) (requiring certified
unions to represent members of the bargaining unit on a nondiscriminatory basis).
40. Papermakers, 416 F.2d at 993.
41. Id. (quoting Whitfield, 263 F.2d at 550). Following Papermakers, the Fifth Circuit
repudiated Whitfield as no longer defensible after Title VII. See Taylor v. Armco Steel Corp., 429 F.2d
498 (5th Cir. 1970).
712 53 UCLA LAW REVIEW 701 (2006)
plant seniority. In making this determination, the court focused exclusively
on the qualifications necessary to perform the job while ignoring whatever
morale issues might otherwise justify the policy. The court also specifically
noted that only qualified individuals would be eligible for the positions
regardless of seniority.42 Equally important, without an adequate business
justification, the court saw the perpetuation of past discrimination as a form
of intentional ongoing discrimination: “When an employer adopts a system
that necessarily carries forward the incidents of discrimination into the pre-
sent, his practice constitutes on-going discrimination, unless the incidents
are limited to those that safety and efficiency require.”43
Although the Papermakers case offers critical insight into the devel-
opment of the disparate impact theory, it was not the sole source of the
theory. As the case makes clear, the Fifth Circuit relied heavily on two
articles published in the Harvard Law Review,44 and a Virginia federal dis-
trict court case that proved especially influential. In Quarles v. Philip
Morris, Inc.,45 the plaintiffs challenged a number of the defendant’s employ-
ment practices at its Richmond, Virginia cigarette factory as intentionally
discriminatory, relying primarily on statistical disparities in the hiring and
promotion process to establish discriminatory intent. The plaintiffs lost
those claims with little analysis by the district court and prevailed only on
the wage claims of two African American employees.46 In addition to chal-
lenging those employment practices, the plaintiffs also challenged the
transfer and seniority provisions of the collective bargaining agreement as
“intentional, unlawful employment practices.”47 The substance of this claim
was effectively identical to that raised in the Papermakers case, but the
analysis was subtly different, including a more detailed focus on how the
maintenance of the seniority system was a form of intentional
42. As a defense to the policy, the employer offered an expert witness to testify about the
effect the government’s policy would have on the plant but, according to the court, his testi-
mony centered on allowing unqualified individuals to bid on jobs or jump over jobs based on seniority.
Papermakers, 416 F.2d at 989–90. It is worth noting that the company’s defense was certainly
weakened in that it was willing to go along with the policy but for the white union’s opposition,
and it appears that the union was seeking to protect its members. The company also seemed to go
to substantial lengths to protect the interests of that union.
43. Id. at 994.
44. See George Cooper & Richard B. Sobol, Seniority and Testing Under Fair Employment
Laws: A General Approach to Objective Criteria of Hiring and Promotion, 82 HARV. L. REV. 1598
(1969) and Note, Title VII, Seniority Discrimination, and the Incumbent Negro, 80 HARV. L. REV.
1260 (1967). Cooper and Sobol were attorneys involved in the Papermakers case, as well as in
Washington v. Davis, 426 U.S. 229 (1976), discussed infra Part I.D.
45. 279 F. Supp. 505 (E.D. Va. 1968).
46. Id. at 509–10.
47. Id. at 510.
Was the Disparate Impact Theory a Mistake? 713
discrimination. At one point, the court even equated the two: “The court
finds that the defendants have intentionally engaged in unlawful
employment practices by discriminating on the ground of race against Quarles,
and other Negroes similarly situated. This discrimination, embedded in sen-
iority and transfer provisions of collective bargaining agreements, adversely
affects the conditions of employment and opportunities for advancement of
the class.”48
By the time of Papermakers, many scholars already were highlighting
the importance of altering seniority systems to ensure that the purposes of
Title VII were fulfilled. From Quarles and Papermakers, it seems clear that
what has come to be defined as the disparate impact theory arose initially, and
primarily, in the context of the discriminatory effects of seniority systems.49
In its enforcement efforts, the United States also targeted unions’ desire to
preserve discriminatory seniority systems.50 These seniority cases were
distinctive in a number of important ways. For example, none of the courts
specifically discussed a disparate impact theory, although they did discuss
the discriminatory effects of what appeared to be a neutral seniority
system—neutral in that it was not implemented with the explicit, or estab-
lished, motive of excluding African Americans from desirable jobs. But as
the discussions in both Papermakers and Quarles suggest, there was a significant
question as to why job seniority was being preserved in these plants on the
basis of what were now unlawful job distinctions. Moreover, the Quarles
case specifically defined the continuance of the system as a form of
intentional discrimination. The vast majority of seniority cases were also
distinctive in that they involved employers that had previously discriminated
48. Id. at 519.
49. See, e.g., Blumrosen, supra note 25, at 294 (“To require that Negro employees remain
subordinate to white employees based on historic discrimination would constitute a deprivation of
employment opportunity and an adverse effect on employment status because of race in violation
of the statute.”); Gould, supra note 24 (describing the way seniority systems preserved past
intentional discrimination); William B. Gould, Seniority and the Black Worker: Reflections on
Quarles and Its Implications, 47 TEX. L. REV. 1039 (1969). Gould observes:
If Congress intended to bring into being an integrated work force, however, and not
merely to create a paper plan meaningless to Negro workers, the only acceptable
legislative intent on past discrimination is one that requires unions and employers to root
out the past discrimination embodied in presently nondiscriminatory seniority
arrangements so that black and white workers have equal job advancement rights.
Id. at 1042.
50. See, e.g., Int’l Bhd. of Teamsters v. United States, 431 U.S. 324 (1977); United States
v. Sheet Metal Workers Int’l Ass’n, Local Union No. 36, 416 F.2d 123 (8th Cir. 1969); United
States v. H.K. Porter Co., 296 F. Supp. 40 (N.D. Ala. 1968). The Papermakers case was also
brought by the United States.
714 53 UCLA LAW REVIEW 701 (2006)
explicitly against their African American employees, so that the
challenged systems perpetuated past intentional discrimination.51
B. Testing Cases and the EEOC
The disparate impact theory clearly took a different turn in the testing
context, but it was a turn anticipated in the seniority cases and one based
on similar concerns regarding the perpetuation of past lawful discrimina-
tion.52 Given the vast inequities in school education systems among white
and black schoolchildren, imposing written tests as a condition of employ-
ment predictably would have the effect of perpetuating segregated job clas-
sifications. Like the seniority cases, the testing cases arose principally
among employers that had engaged in prior intentional discrimination.
Although it may have been difficult to prove that an employer instituted a test
with the specific motive to continue to segregate African Americans, it
took no great leap of faith to understand that the testing requirement did
just that. Indeed, Professor Alfred Blumrosen, an important scholar and
partial architect of the disparate impact theory, argued that instituting tests
or maintaining seniority systems that had the probable effect of excluding
African Americans should be defined as intentional discrimination under
basic tort law.53 Ultimately, this was a road not taken, but it is important to
emphasize that these cases could have been, and in some instances were,
defined as involving intentional discrimination.54
51. See, e.g., Robinson v. Lorillard Corp., 444 F.2d 791 (4th Cir. 1971) (invalidating a
departmental seniority system based on prior discrimination); Jones v. Lee Way Motor Freight, Inc.,
431 F.2d 245 (10th Cir. 1970) (treating a no-transfer policy as akin to seniority as a means to perpetuate
discrimination); United States v. Dillon Supply Co., 429 F.2d 800 (4th Cir. 1970) (finding that
seniority in segregated lines prevented transfers).
52. As was true with the seniority issue, one of the more influential articles was a student
note. See Note, Legal Implications of the Use of Standardized Ability Tests in Employment and
Education, 68 COLUM. L. REV. 691 (1968). Others likewise focused on the way tests had been
introduced into the employment process. See Cooper & Sobol, supra note 44.
53. ALFRED W. BLUMROSEN, BLACK EMPLOYMENT AND THE LAW 176 (1971) (“I
conclude that the intent requirement of Title VII is the intent requirement of a civil action in
tort—that the defendant be aware of the consequences of his action which are reasonably certain
to flow from his behavior.”).
54. This issue is discussed in more detail in Part IV.A., infra. As one example, a Mississippi
district court invalidated the use of Graduate Record Examinations (GRE) scores and a master’s
degree requirement for school teachers because of the expected effects of the policies. See
Armstead v. Starkville Mun. Separate Sch. Dist., 325 F. Supp. 560 (N.D. Miss. 1971). The court
noted, “Racial differentials like those above mentioned were expected by defendants when they
adopted [the p]olicy,” id. at 568, and went on, “[T]he court concludes that the School
Board, in adopting the policy, knew or should have known that its implementation would bar
more black than white teachers from reemployment and hiring by the district.” Id. On appeal,
Was the Disparate Impact Theory a Mistake? 715
Despite the acknowledgements that both tests and the seniority
systems perpetuated past intentional discrimination, courts were largely
receptive to the employers’ claims that they needed to ensure that the
African American employees were qualified for the newly opened
positions.55 This contention was not easily dismissed given that African
Americans had been denied the opportunity for training and advancement
in the past. Not surprisingly, past intentional discrimination rendered
many African Americans unqualified, or less qualified, for the jobs that
were now open to them. Again, this is where the business necessity test
came in—to ensure that employees were truly qualified for the positions
they sought.56 But this also clearly could be a pretext for frustrating the
purposes of the statute: If employers’ past discrimination could serve as a
basis for denying new opportunities to African Americans, it would have
taken many years for them to realize any gains in the workplace—just as was
true with the seniority systems.
Although seniority systems and written tests were surely different,
their similarities are more important to understanding the origins of the
disparate impact theory and its ultimate limits. In both the seniority and
the testing cases, the issue was the perpetuation of past intentional but
lawful discrimination that would contravene the purposes of the legislation.
As a result, the disparate impact theory was not seen initially as a broad
alternative concept of discrimination, but rather, the cause of action origi-
nated to deal with specific issues involving past intentional discrimination.57
It was the EEOC, not the Courts, that conceived of the theory as a
potential alternative approach to discrimination, and it did so in part for
strategic considerations. According to Professor Blumrosen, who was a
high-level EEOC official at the time, it quickly became clear that negotiations
with employers would be smoother if they could move away from a focus on
intentional discrimination, which carried with it an implicit label of blame
the decision was upheld with respect to the GRE scores but not the master’s degree
requirement. See Armstead v. Starkville Mun. Separate Sch. Dist., 461 F.2d 276 (5th Cir. 1972).
55. Professor William Gould was critical of this aspect of the cases, and singled out the
Quarles decision, otherwise commonly viewed as beneficial to plaintiffs, as more “harmful than
helpful.” Gould, supra note 49, at 1074.
56. See Local 189, United Papermakers v. United States, 416 F.2d 980, 990 (5th Cir. 1969)
(emphasizing that “no employee would have a right to a job that he could not perform properly”);
Quarles v. Philip Morris, Inc., 279 F. Supp. 505, 518 (E.D. Va. 1968) (noting that “[m]any
Negroes, regardless of seniority, are not qualified for supervisory positions”).
57. Ironically, the Supreme Court ultimately rejected the argument that seniority systems
that perpetuated pre-Act discrimination violated Title VII. See Int’l Bhd. of Teamsters v. United
States, 431 U.S. 324 (1977) .
716 53 UCLA LAW REVIEW 701 (2006)
that employers were expected to resist.58 To the EEOC, and to plaintiffs
more generally, it mattered little how a particular act was defined so long as
the power to remedy the effects was available. It was also conceivable that
judges would be more receptive to a theory based on unintentional
discrimination, whereas they might have been reluctant to identify
employers as intentional discriminators with all that such a label was meant
to convey. As discussed below, this strategic decision may have had an
understandable appeal at the time, but it was based on a mistaken founda-
tion because it required courts and employers to accept the disparate impact
theory as embodying an identifiable form of discrimination. Yet, the farther
removed one became from remedying the present effects of past intentional
discrimination, the more difficult such acceptance became.
There is another aspect of the specific contexts in which the disparate
impact theory arose that is worth highlighting. Seniority systems and
employment tests were specific practices that were easy to identify and for
which there was no difficult causal question; the adverse impact of these
practices was clear and all that was at issue were the employers’ attempts to
justify the relevance, or legitimacy, of their practices. The employers’
rationales were likewise relatively easy to define, and they were objective
rather than subjective in nature. Employers should have been able to
explain the importance of a seniority system based on job titles or the need
for a particular employment test, and to the extent that any were unable to
do so in the face of clear and substantial disparate impact, one might question
the employer’s motive in establishing or maintaining the practice.59 As
soon as one moved away from these contexts, however, it became far more
difficult to establish causation or a clear business justification conducive to
objective proof. The theory began to weaken and, ultimately, to dissolve.60
58. See ALFRED W. BLUMROSEN, MODERN LAW: THE LAW TRANSMISSION SYSTEM AND
EQUAL EMPLOYMENT OPPORTUNITY 73 (1993) (discussing the Equal Employment Opportunity
Commission (EEOC) position seeking to avoid moral blame). Professor Robert Belton has
highlighted the role that the NAACP Legal Defense and Education Fund played in the early years
of the theory. See Robert Belton, A Comparative Review of Public and Private Enforcement of Title
VII of the Civil Rights Act of 1964, 31 VAND. L. REV. 905, 936 (1978).
59. See, e.g., Developments in the Law—Employment Discrimination and Title VII of the Civil
Rights Act of 1964, 84 HARV. L. REV. 1109, 1142 (1971) (“If [an employer] is fixing an irrelevant
qualification which has a significant differential impact on black employment, he is discriminating
on the basis of race.”). This is similar to the point made many years ago by Professor George
Rutherglen, and one with which I largely agree, although I will later suggest that the disparate impact
theory was unnecessary to achieve this goal. See Rutherglen, supra note 19, at 1311 (“The theory
of disparate impact only addresses the difficulty of proving pretextual discrimination and of using
objective evidence more clearly and systematically.”); discussion infra Part III.B.
60. There are other practices that courts typically treat as presumptively valid and
therefore needing no justification, as discussed in more detail in the next section.
Was the Disparate Impact Theory a Mistake? 717
C. The Griggs Decision
At the time it arose, the Griggs case fit easily within the developing
case law. However, it also offered a potentially more expansive theory of
liability than the seniority cases because it was generally easier to confine
the seniority cases to their particular context, while the testing cases could
have had a broader application. Prior to the passage of Title VII, the Duke
Power Company (Duke Power) had intentionally segregated its workforce,
restricting its African American employees to generally undesirable jobs
within the labor department, including primarily outdoor maintenance and
janitorial work.61 The highest pay within the labor department was lower
than the starting pay in the other departments.62 In 1955, the company
imposed a high school degree requirement for initial assignment to any
department other than the labor department, and after Title VII became
applicable, the employer required those seeking employment or transfers
also to pass two written examinations.63 At the request of employees within
the coal handling division, the employer subsequently allowed existing
employees without high school degrees to transfer from the coal or labor
departments by passing the two examinations.64 For both the high school
degree and testing requirements, the company exempted incumbent
employees, all of whom were white.65 Indeed, several of the white employees
and five of the foremen did not have high school degrees, but they were all
allowed to stay in their positions without taking the examinations.66 This
fact proved crucial to the appellate court, which found the company liable
for intentional discrimination on this basis and required that the black
employees hired before the high school degree requirement was instituted
be provided with the same exemption as the white employees.67 By the
61. See Griggs v. Duke Power Co., 420 F.2d 1225, 1227–29 (4th Cir. 1970) (“Until 1966, no
Negro had ever held a position at [the plant] in any department other than the Labor Department.”).
62. Id. at 1228. The labor department had a maximum wage of $1.565 per hour while the
minimum wage in other departments in the plant was $1.705 with a maximum of $3.18 to $3.65
per hour. Id.
63. Griggs v. Duke Power Co., 401 U.S. 424, 427–28 (1971). Although the tests likely
would have had their greatest impact at the hiring stage, the class in Griggs consisted of existing
employees, and the question with respect to applicants was never addressed. It has been noted
widely that the tests were instituted the day Title VII became effective, creating a presumption
that they were instituted intentionally to disadvantage African Americans. The high school
degree requirement, however, was instituted nearly ten years earlier and actually had a greater
effect on the black employees, only one of whom had a high school degree.
64. Id. at 428.
65. Id. at 427–28.
66. Griggs v. Duke Power Co., 292 F. Supp. 243, 247 (M.D.N.C. 1968).
67. Griggs, 420 F.2d at 1231.
718 53 UCLA LAW REVIEW 701 (2006)
time the case reached the Supreme Court, only four of the original thirteen
employees were seeking relief, and the defendants did not seek review from
the appellate court’s finding of intentional discrimination.
As was common at the time, Duke Power took no steps to determine
whether the tests or the education requirement would actually ensure
qualified employees but instead sought to justify the practices by asserting
that the requirements would provide more educated—and by implication,
better—employees.68 Some of the briefs filed in the Supreme Court
included sample questions from the tests, and there was no obvious connec-
tion between the questions and the jobs at issue.69 The district court
specifically noted that the test, widely used by employers at the time, was
not job related in the sense that it would provide valuable information
regarding the ability to perform job duties. That court nevertheless held
that the employer’s desire to upgrade the quality of its workforce was a valid
business justification.70 From the cases and the briefs, it did not appear that
Duke Power ever administered the test, but the EEOC had documented
that whites typically fared three times better than African Americans on
one of the tests the company sought to implement.71
68. Griggs, 401 U.S. at 431 (noting that both the tests and the high school degree
requirement “were adopted . . . without meaningful study of their relationship to job-performance
ability”). The court of appeals explained the company’s justification as follows: “Duke claims that
the policy was instituted because its business was becoming more complex, it had employees who
were unable to grasp situations, to read, to reason, and who did not have an intelligence level high
enough to enable them to progress upward through the company’s line of advancement.” Griggs,
420 F.2d at 1231.
69. The company used two commonly available tests, the Wonderlic Personnel Test and
the Bennett Mechanical AA Test, and set the passing scores at the median for high school
graduates. The questions included in the briefs included comparisons of proverbs, sentence
comprehension, and definitions: “Does B.C. mean ‘before Christ?’” These questions prompted the
Steelworkers, which filed a brief in support of the petitioners, to conclude, “These questions
perhaps might have utility on a law school aptitude exam. As a measure of ability to fill jobs in an
industrial plant they are ludicrous.” Brief for United Steelworkers of America, AFL-CIO as
Amicus Curiae at 4, Griggs, 401 U.S. 424 (No. 70-124). In contrast, the employer highlighted
some of what it deemed the more pertinent questions. In a footnote it asked:
Does it take “formal schooling” or “cultural background” to know that November is the
eleventh month of the year (Question No. 1 [Wonderlic]), or that chew is related to
teeth as smell is to nose (Question No. 7) or that if 3 lemons sell at 15 cents, one and
one-half dozen would cost 90 cents (Question No. 12)?
Brief for Respondent at 21 n.6, Griggs, 401 U.S. 424 (1971) (No. 70-124).
70. Griggs, 292 F. Supp. at 250 (noting that “[a] test which measures the level of general
intelligence, but is unrelated to the job to be performed is just as reasonably a prerequisite to
hiring or promotion as is a high school diploma”).
71. In a footnote, the Court observed that in one case, 58 percent of whites passed the tests
in question, while only 6 percent of blacks passed. Griggs, 401 U.S. at 430 n.6. The Court also
noted that based on the 1960 census, within North Carolina, 34 percent of white males had
completed high school compared to 12 percent of blacks. Id.
Was the Disparate Impact Theory a Mistake? 719
The company’s explanation for the test, and its willingness to pay
some of the education costs for those who sought to finish high school,72
transformed the case, at least in the Supreme Court’s eyes, from one of
intentional discrimination to something different. There was no indication
that the company adopted the requirements with the express purpose of
confining African Americans to the labor department. At the same time, there
was little question of what the effect of the requirements would be: to
exclude the vast majority of African Americans and thereby to preserve the
segregated job lines within the company. Importantly, all of the courts to
analyze the issue accepted the company’s stated explanation of a desire to
upgrade the quality of its workforce at face value. This was true even
though the test had not been shown to provide reliable information regard-
ing the necessary skills for the positions, and even though the controversy
over standardized tests was not a new one. The United States government
had suspended the administration of a widely used occupational test in 1963
out of concern for its adverse effect on minorities; and in 1968 the NAACP
called for a moratorium on standardized tests.73 Thus, the Court in Griggs
confronted the same question as the Papermakers case: Were ostensibly neu-
tral practices that perpetuated intentional discrimination permissible under
the new statute?
Many commentators have analyzed the Supreme Court’s reasoning in
Griggs,74 but this Article focuses on those aspects of the case that best shed
light on the theory’s origins and future development. I have already noted
72. Id. at 432 (“The Company’s lack of discriminatory intent is suggested by special efforts
to help the undereducated employees through Company financing of two-thirds the cost of tuition
for high school training.”).
73. See Edmund W. Gordon & Tresmaine J. Rubain, Bias and Alternatives in Psychological
Testing, 49 J. NEGRO EDUC. 350, 359 (1980) (noting that the NAACP and other organizations
had called for a moratorium on standardized tests in 1968 due to their effect on minority test
takers); Neil Lawler, Developing New Employment Tests for Minorities, PUBLIC ADMIN. REV., July–
Aug. 1971, at 459, 460 (explaining that the U.S. Training and Employment Service determined
in 1963 that the General Aptitude Test Battery (GATB) was inappropriate for use on minorities).
One of the most important articles on discrimination in tests was also published at about this same
time. See T. Anne Cleary, Test Bias: Prediction of Grades of Negro and White Students in Integrated
Colleges, 5 J. EDUC. MEASUREMENT 115 (1968). It is worth reemphasizing that until the Civil
Rights Act of 1964, lawful discrimination preempted many of these concerns.
74. See, e.g., JACK GREENBERG, CRUSADERS IN THE COURTS: HOW A DEDICAT ED BAND
OF LAWYERS FOUGHT FOR THE CIVIL RIGHTS REVOLUTION 412–20 (1994); Herbert N.
Bernhardt, Griggs v. Duke Power Co.: The Implications for Private and Public Employers, 50 TEX. L.
REV. 901 (1972); Alfred W. Blumrosen, Strangers in Paradise: Griggs v. Duke Power Co. and the
Concept of Employment Discrimination, 71 MICH. L. REV. 59 (1972); Mack A. Player, Is Griggs
Dead? Reflecting (Fearfully) on Wards Cove Packing Co. v. Atonio, 17 FLA. ST. U. L. REV. 1 (1989);
Hugh Steven Wilson, A Second Look at Griggs v. Duke Power Company: Ruminations on Job
Testing, Discrimination, and the Role of the Federal Courts, 58 VA. L. REV. 844 (1972).
720 53 UCLA LAW REVIEW 701 (2006)
that Duke Power had a long history of segregated job lines and the results of
its practices were highly predictable. An equally important aspect of the
case was the historical timeframe in which it arose. By 1971, the Supreme
Court had confronted many evasive state and private practices in voting,
education, and, to a lesser extent, housing. The Court was well aware of the
vast and persistent means by which civil rights mandates could be frustrated.75
All of the briefs that were filed in support of the employees relied on those
earlier cases to emphasize that unvalidated tests could readily be used to
evade the purpose of Title VII.76 While Griggs was the Court’s first explora-
tion of potentially evasive practices in the employment context, the Court’s
past experience undeniably influenced its perspective in interpreting Title
VII. The Court referenced its decisions in the other contexts as support for
invalidating the employer’s practices.77 Equally revealing, during the same
75. See, e.g., Hunter v. Erickson, 393 U.S. 385 (1969) (invalidating a city charter that had
overridden fair housing ordinance); Green v. County Sch. Bd., 391 U.S. 430 (1968) (invalidating
freedom of choice plans); Reitman v. Mulkey, 387 U.S. 369 (1967) (invalidating a state housing
referendum designed to provide local control over fair housing); Harper v. Va. State Bd. of Elections,
383 U.S. 663 (1966) (invalidating a state poll tax); Griffin v. County Sch. Bd., 377 U.S. 218 (1964)
(holding unconstitutional a county’s decision to close the schools); Terry v. Adams, 345 U.S. 461
(1953) (holding unconstitutional the transfer of state primaries to a private discriminatory body);
Smith v. Allwright, 321 U.S. 649 (1944) (holding unconstitutional a state democratic party’s all-
white primaries).
76. For example, petitioner Griggs argued:
The use of tests and educational requirements is but one example of a new breed of racial
discrimination. While outright and open exclusion of Negroes is passé, the use of various
forms of neutral, objective criteria which systematically reduce Negro job opportunity are
producing much the same result. As this Court has long recognized in other contexts of
racial discrimination, those rules which are objective and neutral in form may well be
racially discriminatory in substance and effect.
Brief for Petitioner at 25, Griggs, 401 U.S. 424 (No. 70-124). In support of its argument, the
petitioner cited cases involving grandfather clauses, tuition grants, and gerrymandering. The
United States, as amicus curiae, cited the grandfather clause and literacy test cases early in its
brief. See Brief for the United States as Amicus Curiae at 12–13, Griggs, 401 U.S. 424 (No. 70-124);
see also Brief of the Attorney General of the State of New York as Amicus Curiae in Support of
Reversal at 9, Griggs, 401 U.S. 424 (No. 70-124) (“Duke’s transfer requirements are analogous in their
invidious effects upon Negroes to other practices in civil rights contexts which have been stricken
down by the Courts.”); Brief for United Steelworkers of America, AFL-CIO as Amicus Curiae,
supra note 69, at 4 (noting that the lower court’s interpretation would “cripple Title VII”).
77. The Court cited the voting rights case Gaston County v. United States, 395 U.S. 285 (1969),
explaining that “[t]here, because of the inferior education received by Negroes in North Carolina,
this Court barred the institution of a literacy test for voter registration on the ground that the test
would abridge the right to vote indirectly on account of race.” Griggs, 401 U.S. at 430. The
dissenting opinion in the court of appeals was even more explicit in connecting the past civil
rights cases to the employment setting, citing grandfather clauses, pupil transfer plans, and other
cases to support its claim that “[o]vert bias, when prohibited, has ofttimes been supplanted by
more cunning devices designed to impart the appearance of neutrality, but to operate with the
same invidious effect as before.” Griggs v. Duke Power Co., 420 F.2d 1225, 1238 (4th Cir. 1970)
(Sobeloff, C.J., concurring in part and dissenting in part).
Was the Disparate Impact Theory a Mistake? 721
term that Griggs was decided, the Supreme Court also approved busing as a
remedy for desegregation78 and addressed the difficult question of defining
legislative motive, a close cousin of intent, in the closing of swimming pools
in Jackson, Mississippi.79 In other words, while it was early in the
development of the Court’s employment discrimination doctrine, there was
a substantial history that informed the Court’s decision as to what kinds of
practices were discriminatory. Importantly, to the extent that the Court
had focused on the issue, the earlier civil rights cases were all decided under
a theory of intentional discrimination.80
Placed in context, the Court’s unanimous decision in Griggs was nei-
ther particularly difficult nor far reaching. Permitting the tests and degree
requirement without any justification other than a vague desire to improve
the quality of the workforce effectively would have preserved the segregated
job lines that Title VII was intended to eradicate. Any similar practice
likewise would have been insulated from challenge, and the Court was
unlikely to turn its back on the purpose of Title VII in the first case it con-
fronted on the merits of that statute. By the same measure, the Court
avoided broad proclamations by offering a short and undertheorized deci-
sion that traversed a middle ground.81 Rather than defining the employers’
practices as intentional discrimination, the Court allowed employers to use
selection methods despite their adverse impact so long as they were
demonstrated to be job related.82 The Chamber of Commerce had staked
78. Swann v. Charlotte-Mecklenburg Bd. of Ed., 402 U.S. 1 (1971).
79. Palmer v. Thompson, 403 U.S. 217 (1971). The Palmer case has always been a bit of a
curiosity, and while arguably consistent with Griggs in avoiding a focus on legislative motive, it proved
inconsistent with the Court’s later determination that intent was a necessary element of equal
protection claims. For an influential critique of the case, see Paul Brest, Palmer v. Thompson: An
Approach to the Problem of Unconstitutional Motive, 1971 SUP. CT. REV. 95. It is worth noting that
it would still be two years before the Court took up what would become the most common proof
structure in employment discrimination cases involving claims of disparate treatment. See
McDonnell Douglas Corp. v. Green, 411 U.S. 792 (1973) (establishing a three-part proof
structure to prove individual claims of intent).
80. This became clear in Washington v. Davis, where the Court held that intent was a
requirement of the Equal Protection Clause. Many of the earlier cases—and all of the education
cases—also were decided under that clause. See Washington v. Davis, 426 U.S. 229, 240–43 (1976).
81. This was a common feature of the Court’s race discrimination cases at the time, where
the Court wrote short, often unanimous, decisions to invalidate particular practices without
developing a broader theory. See White v. Regester, 412 U.S. 755 (1973) (invalidating a
multimember voting district based on the totality of the circumstances); Hunter v. Erickson, 393 U.S.
385 (1969) (invalidating a housing ordinance because it burdened racial minorities). I have
previously discussed these cases in Michael Selmi, Proving Intentional Discrimination: The Reality of
Supreme Court Rhetoric, 86 GEO. L.J. 279, 299–301 (1997).
82. Griggs, 401 U.S. at 431 (“The touchstone is business necessity. If an employment
practice which operates to exclude Negroes cannot be shown to be related to job performance, the
practice is prohibited.”).
722 53 UCLA LAW REVIEW 701 (2006)
out a largely similar position in support of Duke Power. The disagreement
between its position and the Court’s was over whether the employer had a
valid business justification for the challenged practices.83 Thus Griggs is
properly seen as a norms-reinforcing decision rather than a broad or different
interpretation of equality that challenged the status quo.84 The Court
upheld an employer’s right to establish efficient business practices even if
that meant excluding African Americans and, later, women, but it required
employers to come forward with some justification for doing so. At this
point the Court had not yet established the governing standards for what
those acceptable business justifications might be.85
Another important aspect of the Griggs case is that the proposition for
which the case is now best known—proof of intent is not necessary to
establish a violation under Title VII—was not a central part of the case.
All three published opinions, and all of the briefs filed in the Supreme
Court, included surprisingly little discussion regarding whether intent was a
required element of proof. There was no discussion of the legislative history
as it applied to the intent requirement. Rather, the Court extensively
explored Title VII’s legislative history to determine the meaning of
§ 703(h), the so-called Tower Amendment that insulated “professionally
developed” tests from challenge.86 The principal § 703(h) issue was
whether professionally developed meant “job related” or simply “any test,”
irrespective of whether the test provided valuable information regarding
83. See Brief Amicus Curiae on Behalf of the Chamber of Commerce of the United States
of America at 5, Griggs, 401 U.S. 424 (No. 70-124) (noting that a discriminatory “intent may
exist, where . . . there is an absence of ‘legitimate business needs’ which justified the employer’s
utilization of such educational or test requirements” but questioning what a “legitimate business
need” might constitute). Even the employer emphasized the validity of its justification rather
than arguing that it did not need a justification. See Brief for Respondent, supra note 69, at 13–17.
84. One reason the Court may have staked out the middle ground was the presence of a
deep irony lurking in the background of the case. Prior to the advent of written tests for the
workplace, subjective employment practices were widely thought to be discriminatory, and
objective practices were seen as a potential antidote. Several courts had, in fact, found subjective
practices to be inherently discriminatory. See Albert J. Rosenthal, Employment Discrimination and
the Law, 407 ANNALS AM. ACAD. POL. & SOC. SCI. 91, 95 (1973) (discussing cases).
85. The Court provided substance to those standards several years later in the case of
Albemarle Paper Co. v. Moody, 422 U.S. 405 (1975), discussed infra at notes 95–98.
86. See Griggs, 401 U.S. at 433 (“The Company contends that its general intelligence tests
are specifically permitted by § 703(h) of the Act. That section authorizes the use of ‘any
professionally developed ability test’ that is not ‘designed, intended, or used to discriminate
because of race . . . .’”); Griggs v. Duke Power Co., 420 F.2d 1225, 1233 (4th Cir. 1970) (“Next,
we consider the testing requirements to determine their validity and we conclude that they . . . are
valid under § 703(h) . . . .”); Griggs v. Duke Power Co., 292 F. Supp. 243, 249–50 (M.D.N.C.
1968) (upholding tests as consistent with § 703(h)).
Was the Disparate Impact Theory a Mistake? 723
one’s ability to perform a particular job.87 To the extent that there was
discussion regarding intent, particularly in the lower courts, unvalidated
tests were equated with intentional discrimination.88 This position was
consistent with the state of the doctrine at the time, as courts were primar-
ily seeking to determine what practices violated Title VII without thinking
more broadly about an underlying theory. In the words of Robert Belton,
one of the attorneys for Griggs, “It was all discrimination.”89
This interpretation of Griggs was apparent from the reactions to the
decision both within the media and among scholars. Immediately following
the decision, the commentary focused almost exclusively on the require-
ments for validating written tests—and there was no indication that the
case had spawned a new theory of liability.90 Similarly, the vast majority of
the cases that succeeded after Griggs followed the same pattern: Southern
employers with a history of intentional discrimination utilizing unvalidated
tests, most of which were implemented after the passage of Title VII, which
had a clear and predictable exclusionary effect on black employees and
87. See Griggs, 401 U.S. at 433–34 (upholding EEOC’s interpretation that § 703(h)
permits only job related tests). The district court, on the other hand, concluded that the tests
only had to be professionally developed: “A test which measures the level of general intelligence, but
is unrelated to the job to be performed is just as reasonably a prerequisite to hiring or promotion as
is a high school diploma.” Griggs, 292 F. Supp. at 250. The district court saw no basis for
challenging the high school degree requirement and, by analogy, upheld the testing requirement
as well. Id.
88. The only substantial discussion of intent came from the dissenting judge in the court of
appeals who wrote:
Distilled to its essence, the underpinning upon which my brethren posit their
argument is their expressed belief in the good faith of Duke Power. For them, the crucial
inquiry is not whether the Company can establish business need, but whether it has a
bad motive or has designed its tests with the conscious purpose to discriminate against
blacks . . . . But this is no[ ] answer.
A man who is turned down for a job does not care whether it was because the
employer did not like his skin color or because, although the employer professed
impartiality, procedures were used which had the effect of discriminating against the
applicant’s race.
Griggs, 420 F.2d at 1245–46 (Sobeloff, C.J., concurring in part and dissenting in part).
Interestingly, his language parallels what came to be the dominant interpretation of the disparate
impact provision.
89. Conversation with Robert Belton, Professor of Law, Vanderbilt Law School (Jan. 6, 2004).
90. See Bernhardt, supra note 74, at 918–20 (treating Griggs as a case about testing);
Wilson, supra note 74 (discussing tests and suggesting that courts should be sensitive to costs in the
business necessity calculation). For journalist reports, see Stanley Klein, Job Testing Comes Under
Fire: Too Many Disqualify Minority Applicants, N.Y. TIMES, Sept. 19, 1971, at F5 (discussing the
impact of the Griggs decision on employment testing); Job Tests Held in Violation of Rights Act,
WASH. POST, Mar. 9, 1971, at A1.
724 53 UCLA LAW REVIEW 701 (2006)
applicants.91 There were a number of exceptions to this rule that began to
stretch the disparate impact theory into other areas, including height and
weight requirements, which had an adverse impact against women and
Latinos, and arrest records, which tended to have an adverse impact on
blacks.92 Importantly, even at this early juncture, when employers had
legitimate justifications for their practices, courts did not hesitate to accept
them.93 Far more commonly, however, the challenged practices were
imposed, as in Griggs, without any significant validation or justification.94
91. See, e.g., Payne v. Travenol Labs., 565 F.2d 895 (5th Cir. 1978) (invalidating the
twelfth-grade education requirement of a Mississippi employer as not justified by business
necessity); Watkins v. Scott Paper Co., 530 F.2d 1159 (5th Cir. 1976) (invalidating education
and testing practices for an Alabama company with a history of discrimination); Rogers v. Int’l
Paper Co., 510 F.2d 1340 (8th Cir. 1975) (sustaining a disparate impact challenge for failure to
meet minimum validation standards for an Arkansas company where 1/160 supervisors were
black), vacated on other grounds, 423 U.S. 809 (1975), on remand, 526 F.2d 722 (8th Cir. 1975);
Pettway v. Am. Cast Iron Pipe Co., 494 F.2d 211 (5th Cir. 1974) (invalidating practices that
perpetuated the effects of past intentional discrimination); Sims v. Sheet Metal Workers Int’l
Ass’n, Local Union No. 65, 489 F.2d 1023 (6th Cir. 1973) (striking down a testing practice
instituted after Title VII became law); Head v. Timken Roller Bearing Co., 486 F.2d 870 (6th Cir.
1973) (holding that pre-Act discrimination rendered a seniority system discriminatory); United
States v. St. Louis-San Francisco Ry. Co., 464 F.2d 301 (8th Cir. 1972) (holding the defendant’s
past history of discrimination relevant to invalidating current seniority practices); Rowe v. Gen. Motors
Corp., 457 F.2d 348, 356 (5th Cir. 1972) (invalidating promotion and transfer practices by an
Atlanta plant with a history of discrimination where practices “froze” the status quo); United
States v. Jacksonville Terminal Co., 451 F.2d 418 (5th Cir. 1971) (invalidating seniority and
transfer practices for a Jacksonville, Florida company with a history of exclusionary practices);
Robinson v. Lorrillard Corp., 444 F.2d 791 (4th Cir. 1971) (invalidating a department seniority
practice because of perpetuated discrimination).
92. See Smith v. Olin Chem. Corp., 535 F.2d 862 (5th Cir. 1976) (challenging a discharge
for sickle cell anemia); Green v. Mo. Pacific R.R. Co., 523 F.2d 1290 (8th Cir. 1975)
(invalidating an employer’s practice of refusing to hire applicants with convictions other than
minor traffic offenses as inconsistent with business necessity); Wallace v. Debron Corp., 494 F.2d 674
(8th Cir. 1974) (challenging a company policy discharging employees whose wages were garnished
and remanding for determination of business necessity).
93. See, e.g., White v. Carolina Paperboard Corp., 564 F.2d 1073 (4th Cir. 1977) (finding
that an employer’s practices satisfied the business necessity test because the jobs required special
skills); Yuhas v. Libbey-Owens-Ford Co., 562 F.2d 496 (7th Cir. 1977) (upholding a no-spouse
rule as job related despite its adverse impact against women); Boyd v. Ozark Air Lines, Inc., 568 F.2d 50
(8th Cir. 1977) (upholding a pilot height requirement as relevant to operation within the
cockpit); Allen v. City of Mobile, 466 F.2d 122 (5th Cir. 1972) (police department tests);
Spurlock v. United Airlines, Inc., 330 F. Supp. 228 (D. Colo. 1971) (upholding a college degree
requirement as related to the training program).
94. See Walston v. County Sch. Bd., 492 F.2d 919 (4th Cir. 1974) (invalidating a teacher
examination in a Virginia school district as part of a desegregation order); United States v. Ga.
Power Co., 474 F.2d 906 (5th Cir. 1973) (invalidating tests that were implemented after the end
of formal segregation and where no validation was attempted until after suit was filed); Young v. Edgcomb
Steel Co., 363 F. Supp. 961 (M.D.N.C. 1973) (invalidating the use of the Wonderlic test as not
validated), aff’d in part, rev’d in part, 499 F.2d 97 (4th Cir. 1974).
Was the Disparate Impact Theory a Mistake? 725
D. Washington v. Davis and the Turn to Intent
The question of employment testing returned to the Supreme Court
just a few years later in a case that also arose from North Carolina and that
involved the same test that was at issue in Griggs.95 Albemarle Paper, now
known primarily as one of the most frequently misspelled cases,96 greatly
expanded the requirements for validating written examinations, and, in
many ways, was far more important than Griggs in defining the standards
employers had to meet to comply with Title VII.97 At the same time,
Albemarle Paper was strictly a testing case insofar as the validation require-
ments it established were applicable only to written examinations.98
The real watershed case arrived the following year in Washington v.
Davis,99 a case that in many ways paralleled the issues raised in Griggs, but
by a quirk of timing, was pursued as a constitutional rather than a statutory
claim. Davis involved a challenge to a written test developed by the federal
government for use by the civil service commission and administered by the
Washington, D.C. Police Department. The test had a significant adverse
impact on African Americans, who failed at a rate approximately four times
higher than that for whites.100 Despite the test’s adverse impact, the Department
had greater success in its actual hiring, in large part because in some years as
many as 70 percent of its applicants were African American.101 At the time
the case reached the Supreme Court, nearly half of the new police recruits,
95. Albemarle Paper Co. v. Moody, 422 U.S. 405 (1975).
96. For whatever reason, Albemarle is routinely spelled “Albermarle.” Runner-up in the mis-
spelling category is Atonio of the Wards Cove case, which spellcheckers frequently correct to “Antonio.”
97. Albemarle, 422 U.S. at 427–36. Some commentators interpreted Albemarle as a serious
restriction on employment testing. For a somewhat hyperbolic account, see James G. Johnson,
Albemarle Paper Company v. Moody: The Aftermath of Griggs and the Death of Employee Testing,
27 HASTINGS L.J. 1239 (1976) (arguing that the EEOC standards relied on by the Albemarle
Court were too stringent and unworkable).
98. The EEOC guidelines, adopted in Albemarle, established three means by which written
tests could be validated, and those guidelines continue to focus exclusively on written
examinations. See Uniform Guidelines on Employee Selection Procedures, 29 C.F.R. § 1607
(2004). Originally issued in 1966, the guidelines were last revised in 1978. For a thorough
analysis of the guidelines and the different means of validation, see Gillespie v. Wisconsin, 771 F.2d
1035 (7th Cir. 1985).
99. 426 U.S. 229 (1976).
100. Between 1968 and 1971, 57 percent of black applicants failed the test while only 13 percent
of whites failed. See Davis v. Washington, 512 F.2d 956, 958–59 (D.C. Cir. 1975).
101. Between 1968 and 1971, the pool fluctuated from a high of 70.3 percent black
applicants (1969) to 52.1 percent (1970), with an overall percentage of 58.4 percent. Brief for
Respondent at 5 n.8, Washington v. Davis, 426 U.S. 229 (1976) (No. 74-1492). During the same
time period, 42.5 percent of the hires were black. Id.; see also Davis, 512 F.2d at 961 n.32.
726 53 UCLA LAW REVIEW 701 (2006)
and 35 percent of the entire force, were African Americans.102
Foreshadowing a future debate, the Department defended its use of the test
by arguing that its aggressive recruitment efforts had resulted in a police
department that was representative of the relevant labor market, an issue
the court of appeals rejected and the Supreme Court ignored.103 It was also
significant that at the time the case arose, police departments were a
particular focus of integration efforts as a result of the Kerner Commission
Report, issued after the 1968 riots that had ravaged many cities, including
Washington, D.C.104 One of the central conclusions of the Kerner
Commission was that police departments required more minority members
in order to more effectively police urban cities.105 Adding to the complexity
of the case, Washington, D.C. was a majority black city with a progressive
black mayor, and the United States was a defendant in the case even while
it was the chief enforcer of Title VII.106
The challenge in Washington v. Davis was filed in 1970, but because
Title VII did not become applicable to public employers until 1972 the case was
brought under the Constitution, and the plaintiffs sought to import the
standards developed in Griggs into the Equal Protection Clause.107 The idea
of applying Title VII principles to the Equal Protection Clause may seem
radical today. However, at the time, it was the prevailing judicial approach.
Both of the lower courts had, in fact, applied the Title VII standards with-
out significant analysis, as had been true of many lower courts, and the issue
102. Brief for Respondent, supra note 101, at 22 n.29 (noting that 36.5 percent of the police
force was black). The Supreme Court noted that “[s]ince August 1969, 44% of the new police
force recruits had been black. . . .” Davis, 426 U.S. at 235. The court of appeals stated that 55 percent
of all new officers reporting for the academy were black. Davis, 512 F.2d at 961 n.32.
103. See Brief for Petitioners at 14–18, Davis, 426 U.S. 229 (No. 74-1492) (arguing that the
test did not have an adverse impact because the department’s overall numbers were consistent
with representation from the recruitment pool). This emphasis on the “bottom line” was
subsequently rejected by the Supreme Court. See Connecticut v. Teal, 457 U.S. 440 (1982).
104. See REPORT OF THE NATIONAL ADVISORY COMMISSION ON CIVIL DISORDERS (1968).
The Kerner Commission Report has been the subject of extensive scholarly commentary. See,
e.g., John O. Calmore, Spatial Equality and the Kerner Commission Report: A Back-to-the-Future
Essay, 71 N.C. L. REV. 1487 (1993).
105. See Paul Frymer & John D. Skrentny, The Rise of Instrumental Affirmative Action: Law
and the New Significance of Race in America, 36 CONN. L. REV. 677, 690 (2004) (noting that the
Kerner Commission “[R]eport urged more efforts to recruit more African Americans, and those
officers ‘should be so assigned as to ensure that the police department is fully and visibly
integrated’”). The National Advisory Commission on Criminal Justice Standards and Goals
issued a report in 1973 calling for explicit hiring goals and suggesting elimination of all entrance
requirements other than those necessary to police work. See id. at 690–91.
106. For a discussion of Washington, D.C. and its politics at the time, see Philip G. Schrag, By
the People: The Political Dynamics of a Constitutional Convention, 72 GEO. L.J. 819, 829–31 (1984).
107. Davis, 426 U.S. at 232–33, 236 n.6.
Was the Disparate Impact Theory a Mistake? 727
was barely addressed in the Supreme Court briefs.108 Nevertheless, by 1976,
the disparate impact theory had acquired a different meaning than origi-
nally articulated in Griggs, and the Supreme Court now faced a potentially
far more expansive theory that could have prompted a significant amount of
litigation and social restructuring. Many lower courts and some scholars
had seized on the theory as a broad tool for social reform in housing, munici-
pal services, and other areas.109 Civil rights issues also had taken a different turn.
The busing controversy had erupted throughout the country, and the
images of the Boston melee were still fresh in everyone’s mind.110 And
while the Court had approved busing as a desegregation remedy during the
same term as its Griggs decision, Washington v. Davis followed shortly after
the Court’s decision in Milliken v. Bradley prohibiting desegregation efforts
that crossed district lines, a decision many advocates viewed as effectively
ending serious hope of meaningful desegregation.111 Between Griggs and
Davis, the Court also had its first taste of affirmative action, rejected a
challenge to unequal funding of schools, addressed school desegregation in a
Northern school system, and balanced seniority rights against the
108. The respondent’s brief addressed the issue of the constitutional standard in a footnote.
Brief for Respondent, supra note 101, at 23 n.35. Equal protection cases that applied the statutory
standard included Castro v. Beecher, 459 F.2d 725 (1st Cir. 1972) (holding that the Constitution
requires no less than Title VII); Chance v. Board of Examiners, 458 F.2d 1167 (2d Cir. 1972) (a
challenge to a competitive examination for school supervisory positions); Carter v. Gallagher, 452 F.2d
315 (8th Cir. 1971) (Minneapolis fire department challenge); Harper v. Mayor of Baltimore, 359 F.
Supp. 1187 (D. Md. 1973) (§ 1981 and equal protection challenges to a fire department test), modified
and aff’d sub nom. Harper v. Kloster, 486 F.2d 1134 (4th Cir. 1973); Pennsylvania v. O’Neill, 348 F.
Supp. 1084 (E.D. Pa. 1972) (challenge to a Philadelphia police test), aff’d in part, vacated in part, 473 F.3d
1029 (3d Cir. 1973); and Arrington v. Massachusetts Bay Transportation Authority, 306 F. Supp.
1355 (D. Mass. 1969) (a constitutional challenge to the GATB test).
109. See, e.g., Coal. for Educ. v. Bd. of Elections, 370 F. Supp. 42 (S.D.N.Y. 1974) (a voting
rights challenge); Cuyahoga Metro. Hous. Auth. v. City of Cleveland, 342 F. Supp. 250 (N.D.
Ohio 1972) (a challenge to a zoning requirement), aff’d, 474 F.2d 1102 (6th Cir. 1973); see also
William Silverman, Equal Protection, Economic Legislation, and Racial Discrimination, 25 VAND. L.
REV. 1183 (1972) (arguing that the disparate impact theory could invalidate minimum wage and
usury laws and preferential licensing).
110. The riots that broke out over the school desegregation order requiring busing in Boston
occurred in 1974–75. See U.S. COMM’N ON CIVIL RIGHTS, DESEGREGATING THE BOSTON
PUBLIC SCHOOLS: A CRISIS IN CIVIC RESPONSIBILITY (1975). The episode has been chronicled
wonderfully in Eyes on the Prize: American’s Civil Rights Years (PBS television broadcast 1987) and
in J. ANTHONY LUKAS, COMMON GROUND (1985).
111. See Milliken v. Bradley, 418 U.S. 717 (1974). The Milliken case has been widely
denounced as one of the most restrictive cases involving school desegregation. See, e.g., GARY
ORFIELD & SUSAN E. EATON, DISMANTLING DESEGREGATION: THE QUIET REVERSAL OF
BROWN V. BOARD OF EDUCATION 10–11 (1996) (claiming that school desegregation “hit a stone
wall” with Milliken); Tomiko Brown-Nagin, Elites, Social Movements, and the Law: The Case of
Affirmative Action, 105 COLUM. L. REV. 1436, 1518 (2005) (noting that Milliken “effectively
foreclosed the possibility of integrated schooling in the central cities”).
728 53 UCLA LAW REVIEW 701 (2006)
antidiscrimination mandate, all of which may have awakened the Court to
the complexities of equality under an expansive interpretation of the
disparate impact theory.112 Moreover, in the particular context of the Davis
case, the constitutional question was wholly unnecessary, as any challenge
to public employment practices could now be brought under Title VII.
These factors surely contributed to the Court’s decision to exclude effects
claims from the scope of the Equal Protection Clause.
Although Washington v. Davis has been canonized for holding that
intent is an element of the Fourteenth Amendment, the Court’s discussion
of the test, and the way that discussion diverged from its analysis in Griggs,
is equally important to understanding the meaning of the disparate impact
theory, and this point too often is missed by those who cling to the power of
the theory. Central to the Court’s holding was a normative judgment that
the administration of the test challenged in Davis was not properly defined
as discrimination—intentional or otherwise. In the last part of its opinion,
the Court, in fact, upheld the test under the standards applicable to Title
VII, and did so in a way that was remarkably different in tone and substance
from the Court’s earlier decision in Griggs.113 Although the test at issue in
Davis was similar to that in Griggs in its structure, primarily an SAT-style
test, it had been created by the federal government and validated for use
based on the Washington, D.C. Police Department’s training program.114
Moreover, although Washington, D.C. was nominally a Southern city, it
was one without a deep history of civil rights resistance, and in light of the
city’s recruitment efforts and the presence of a black mayor, it would have
112. See Franks v. Bowman Transp. Co., Inc., 424 U.S. 747 (1976) (considering seniority
rights issued under Title VII); DeFunis v. Odegaard, 416 U.S. 312 (1974) (dismissing an
affirmative action case as moot); Keyes v. Sch. Dist. No. 1, 413 U.S. 189 (1973) (discussing de facto
and de jure discrimination in the Denver school system); San Antonio Indep. Sch. Dist. v.
Rodriguez, 411 U.S. 1 (1973) (rejecting an equal protection challenge to a state education
funding system).
113. See Washington v. Davis, 426 U.S. 229, 249–50 (1976) (noting that the district court
“assumed that Title VII standards were to control the case” and upholding the job-relatedness of
the test).
114. Id. at 251 (noting that “[t]he District Court’s . . . conclusion that Test 21 was in fact
directly related to the requirements of the police training program was supported by a validation
study, as well as by other evidence of record . . . .”). A copy of the test was attached as an
appendix to the appellate court decision by the dissenting judge. See Davis v. Washington, 512 F.2d
956, 967 (D.C. Cir. 1975). There were eighty questions and a score of forty was necessary to pass.
The test was designed to measure verbal ability, and most of the questions asked variations on
reading comprehension or vocabulary. To give one example, Question Eighty asked: “The saying
‘Anger dies quickly with a great man’ means most nearly A) A good man is slow to anger. B) Nothing
ruffles a good disposition. C) One can forgive but not forget. D) Strong passions cannot last. E)
To continue to bear malice is petty.” Id. at 976.
Was the Disparate Impact Theory a Mistake? 729
been difficult to suggest that the city was using the test with an intent to
exclude African Americans from its force. 115 As a result, and in direct con-
trast to its analysis in Griggs, the Court did not see the administration of the
test as discriminatory, specifically commenting: “[W]e have difficulty under-
standing how a law establishing a racially neutral qualification for employ-
ment is nevertheless racially discriminatory . . . simply because a greater
proportion of Negroes fail to qualify than members of other racial or ethnic
groups.”116 The Court concluded: “[I]t is untenable that the Constitution
prevents the Government from seeking modestly to upgrade the communi-
cative abilities of its employees rather than to be satisfied with some lower
level of competence, particularly where the job requires special ability to
communicate orally and in writing.”117
These statements were directly contrary to those made in Griggs.
Although it might be tempting to see the distinction in the case as turning
on the difference in jobs—police compared to power plant workers—there
is little basis for this distinction in the case.118 In Davis, there were no find-
ings on the importance of the verbal abilities of police officers, nor was
there any indication that this particular test sought the requisite verbal
abilities necessary to read manuals or communicate effectively with the
public, presumably the primary basis for the requirement. Rather, the test
had been validated against a written examination administered at the end
of the training academy, a process the appellate court had dismissed as
demonstrating little more than a correlation between success on written
examinations.119 Whether the test had been properly validated was the
115. Davis, 426 U.S. at 235. The Court later referenced the Department’s successful
recruiting efforts in summarily rejecting an intentional discrimination allegation:
[W]e think the District Court correctly held that the affirmative efforts of the Metropolitan
Police Department to recruit black officers, the changing racial composition of the recruit
classes and of the force in general, and the relationship of the test to the training program
negated any inference that the Department discriminated on the basis of race . . . .
Id. at 246. In their brief in the Supreme Court, the plaintiffs disputed the city’s recruitment
efforts, noting that the city’s nationwide recruitment led to a sharp increase in white applicants.
See Brief for Respondent, supra note 101, at 4–5. The district court, however, had concluded,
“The Metropolitan Police Department is a model nationwide for its success in bridging racial
barriers.” Davis v. Washington, 348 F. Supp. 15, 18 (D.D.C. 1972).
116. Davis, 426 U.S. at 245.
117. Id. at 245–46.
118. For an interpretation that relies heavily on a distinction in the job duties, see Barbara
Lerner, Washington v. Davis: Quantity, Quality and Equality in Employment Testing, 1976 SUP. CT.
REV. 263, 279–82.
119. See Davis, 512 F.2d at 962–64. One problem with the validation effort is that the test
administered at the end of the training program had not been shown to be related to the actual
qualities of a successful police officer, and without more, there was a possibility that the test
simply indicated which applicants performed well on written examinations.
730 53 UCLA LAW REVIEW 701 (2006)
primary issue raised in the Supreme Court briefs.120 Moreover, some of the
power plant jobs at issue in Griggs also involved potentially dangerous and
complex work, and the employer had argued that its tests were necessary to
ensure its employees had the ability to progress to those higher level jobs.121
Rather than focusing on the distinction in job duties, something else
appeared to have shifted on the Court, as within the Washington, D.C.
police force, the Court did not see the use of the test as discriminatory.
The factual difference in the cases also highlights an important
unasked question: Why should Washington, D.C. have been liable for
administering the test? This is a question that has rarely been addressed by
those advocating an expansion of disparate impact liability, but one that
goes to the core of the theory while also exposing its limits. In Griggs, it was
relatively easy to make the moral case for liability given the company’s
history of discrimination and the way the tests perpetuated that past dis-
crimination without providing clear information relevant to the employer’s
business interests. But in Davis, the link was far less clear. The test at issue
in Davis was used throughout the civil service system, and the city’s recruit-
ment efforts did not suggest a desire to exclude African Americans.122 The
success of the Department’s hiring practices also made this a difficult case,
because, for practical purposes, the plaintiffs were arguing that 70 percent of
the new recruits should have been black rather than 50 percent. By any measure
this was a difficult argument to sell.123 As in Griggs, there was a high
probability that the Department’s test would have an adverse effect on
black applicants, but there was also reason to believe the city would take
steps to mitigate that harm. Just the opposite appeared true with Duke
120. The questions presented involved the test’s adverse impact and whether the test had
been properly validated. Testing organizations also weighed in on the case, whereas they were
absent from Griggs. See Brief of American Society for Personnel Administration as Amicus
Curiae, Davis, 426 U.S. 229 (No. 74-1492); Brief of the Executive Committee of the Division of
Industrial Organizational Psychology (Division 14) of the American Psychological Ass’n as Amicus
Curiae, Davis, 426 U.S. 229 (No. 74-1492); Brief of Educational Testing Service as Amicus Curiae,
Davis, 426 U.S. 229 (No. 74-1492).
121. The Fourth Circuit explained: “Duke claims that the policy was instituted because its
business was becoming more complex, it had employees who were unable to grasp situations, to read,
to reason, and who did not have an intelligence level high enough to enable them to progress
upward through the company’s line of advancement.” Griggs v. Duke Power Co., 420 F.2d 1225,
1231 (4th Cir. 1970).
122. See Lerner, supra note 118, at 271–72 (emphasizing the city’s recruitment efforts).
123. The court of appeals signed on, however, noting specifically: “Although the
Department, quite commendably, has succeeded in increasing the proportion of black officers
through vigorous efforts, it is self-evident that use of selection procedures that do not have a disparate
effect on blacks would have resulted in an even greater percentage of black officers than exists
today.” Davis, 512 F.2d at 961 (footnote omitted).
Was the Disparate Impact Theory a Mistake? 731
Power, which did not appear opposed to the results of its employment prac-
tices.124 One might have advanced the theory articulated by Professor
Blumrosen and others that liability should attach based on the torts
standard of knowledge of the probable effects of the city’s acts, but that
would have been a significant expansion into a negligence theory of liabil-
ity, an expansion rejected, at least implicitly, in Griggs.
Two other liability rationales were possible. Assuming the differential
success rates on the test were attributable to the deficient education that
Washington, D.C. provided to African Americans in its segregated schools,
one might argue that the city was now required to take remedial steps so
that its prior discrimination in the schools did not lead to a legacy of workplace
discrimination in the future.125 To my mind, this theory presents the
strongest basis for holding the city liable, but it was not a theory advanced
or even referenced by any of the parties that filed briefs in the case. One
might also assert a more general theory independent of discrimination
within the schools to suggest that the city had an obligation to remedy past
discrimination that rendered African Americans less prepared for written
tests. If developed, this theory would have imposed an affirmative action
obligation on the city to remedy what has come to be defined as societal
discrimination, and although also a solid basis for holding the city liable,
this argument would have extended the disparate impact theory well past its
moorings.126 In the end, no adequate theory justifying liability against the
city of Washington, D.C. was advanced; indeed, no theory was advanced at
all. Instead, the parties focused on whether the Department had satisfied
the statutory requirements of Title VII.127
124. The city emphasized this point in its brief: “[I]t is indeed manifest that Test 21 does not
operate to lock in a prior practice of discrimination, to freeze a racially unacceptable status quo, or
to perpetuate or carry forward a racially tainted hiring practice of yesteryear.” Brief for Petitioner,
supra note 103, at 16. The brief also sought to distinguish the situation in Griggs by noting that
Duke Power had instituted its test the day Title VII became applicable and did so without
meaningful study. Id. at 17.
125. While the majority of black applicants were from Washington, D.C., many were from
outside the city. In fact, one of the issues raised in the briefs involved what the proper labor
market should be for comparison purposes. The city argued that it recruited primarily from a fifty-
mile radius of the city, where the population of young black men was 36.5 percent, nearly exactly
the percentage of black officers on the force. See Reply Brief of the United States at 4, Washington
v. Davis, 426 U.S. 229 (1976) (No. 74-1492).
126. I have discussed the link between the disparate impact theory and societal discrimination
previously. See Michael Selmi, Remedying Societal Discrimination Through the Spending Power,
80 N.C. L. REV. 1575, 1601–03 (2002).
127. Subsequently, in critiquing the Court’s decision, a number of scholars provided strong
rationales. See Kenneth L. Karst, Foreword: Equal Citizenship Under the Fourteenth Amendment, 91 HARV.
L. REV. 1, 51–52 (1977) (arguing that tests with disparate impact reinforce the “stigma of caste”);
732 53 UCLA LAW REVIEW 701 (2006)
The Supreme Court, however, needed more than a statutory argument
to impose liability in this particular circumstance, and I suggest below that
one reason the disparate impact theory has failed to produce greater results
is precisely because no argument was developed to explain why the theory
was consistent with the commitment to equality—why Washington, D.C.
should have been held liable for the discriminatory results of the test rather
than whether proof of intent was necessary either under the Constitution or
Title VII. This failure to develop a theory, grounded in continuing dis-
crimination, helps explain more generally why antidiscrimination law has
had such a limited scope.128
Any doubt that Washington v. Davis was a limitation on the disparate
impact theory was soon put to rest in the cases that followed. In Dothard v.
Rawlinson,129 the Supreme Court permitted a disparate impact challenge, brought
Michael J. Perry, The Disproportionate Impact Theory of Racial Discrimination, 125 U. PA. L. REV.
540, 558 (1977) (“The underlying cause of disproportionate racial impact, the especially
disadvantaged social position of black Americans, is one for which American society and
government bear a heavy moral responsibility.”). In an earlier article, Owen Fiss had set forth a
theory including disparate effects in his equal achievement perspective. See Owen M. Fiss, A
Theory of Fair Employment Laws, 38 U. CHI. L. REV. 235 (1971).
128. At least two other theories or arguments were available. Consistent with the doctrine
developed in the Albemarle Paper case, the plaintiffs might have offered alternative selection
devices that would satisfy the department’s needs while having a less adverse impact. As the case
proceeded, no alternative was offered, but a typical alternative would be to lower the cut-off score.
When a reasonable alternative is offered, and the employer declines to adopt it, it seems
reasonable to infer discrimination from its acts because there would no longer be any business
justification. This is an inference that is now part of the formal proof structure arising from the
Civil Rights Act of 1991.
The plaintiffs also might have sought to demonstrate that the test was culturally biased in
that the test questions favored white applicants and were less likely to be familiar to black
applicants. My sense is that this is the most common objection to written tests, but it turns out to
be a difficult claim to establish. Depending on the test, it might be possible to show that certain
test questions favor white applicants either due to cultural or educational differences, and that
might have been true at the time. These claims, however, frequently run aground in
professionally developed tests, particularly with more recent tests that are designed with an eye to
avoiding cultural bias. A more difficult issue is that there is little empirical support for the notion
that the tests underpredict the performance of minority candidates, which would be a strong sign
of test bias. See Sackett et al., supra note 15, at 303 (“An extensive body of research in both the
employment and education literatures has demonstrated that these tests generally do not exhibit
predictive bias.”). If the test were biased, one would expect African American employees to do
better on the job than their test scores predict, but in most studies, the opposite has proved true. In the
last decade, psychologist Claude Steele and others have developed a theory known as stereotype
threat which suggests that a concern with confirming stereotypes leads minority students to
underperform on certain examinations. See, e.g., Claude M. Steele, A Threat in the Air: How
Stereotypes Shape Intellectual Identity and Performance, 52 AM. PSYCHOLOGIST 613 (1997); Claude M.
Steele, Thin Ice: “Stereotype Threat” and Black College Students, ATLANTIC MONTHLY, Aug. 1999,
at 44, 48. This is a complicated and controversial theory that has not yet been shown to apply
outside of the academic context.
129. 433 U.S. 321 (1977).
Was the Disparate Impact Theory a Mistake? 733
by women, to height and weight requirements for correctional officer
positions in a male prison.130 Yet, while permitting the challenge, the
Supreme Court also upheld the specific gender restrictions as a bona fide
occupational qualification, negating the disparate impact finding and cre-
ating the curious result that what the employer was prohibited from doing
indirectly it could accomplish directly.131 Two years later the Court rejected
a challenge to the New York City Transit Authority’s policy of refusing to
employ individuals who were receiving methadone treatment, which the
plaintiffs alleged had a disparate impact against African Americans, who
were disproportionately represented among the affected class.132 In a cursory
analysis, the Court accepted the agency’s safety justification as satisfying the
business necessity test.133 Then in quick succession, the Supreme Court
declined to extend the theory to contracting claims filed under the civil rights
statute known as § 1981, voting rights under the Fifteenth Amendment,
and a challenge to a road closing.134 The only application of the theory the
Court permitted was a complicated decision allowing disparate impact
challenges under the regulations implementing Title VI, the governmental
funding statute requiring nondiscrimination, even while holding that the
statute itself only precluded intentional discrimination.135
By the end of the theory’s first decade, the Court had rejected more
challenges than it had accepted, and it had largely limited the theory to its
origins—namely testing claims and perhaps some other objective proce-
dures capable of formal validation. At this point, there was little reason to
think the disparate impact theory would mark a radical doctrinal shift. A theory
that burst onto the scene in 1971 ended its first decade with a whimper. As
130. See id.
131. See id. at 335 (“A woman’s relative ability to maintain order in a male, maximum-
security, unclassified penitentiary of the type Alabama now runs could be directly reduced by her
womanhood.”).
132. See New York City Transit Auth. v. Beazer, 440 U.S. 568 (1979).
133. Id. at 587 n.31. The Court accepted the Transit Authority’s policies that methadone
was a narcotic that created many problems associated with other narcotics such as drowsiness and
insomnia. Id. at 588 n.32.
134. See Gen. Bldg. Contractors Ass’n, Inc. v. Pennsylvania, 458 U.S. 375 (1982) (holding
that § 1981 prohibiting discrimination in contracting only reaches intentional conduct); City of
Memphis v. Greene, 451 U.S. 100 (1981) (rejecting an intentional discrimination challenge to a road
closing in Memphis that adversely affected African Americans); City of Mobile v. Bolden, 446 U.S. 55
(1980) (holding that the Fifteenth Amendment only precludes intentional discrimination).
135. See Guardians Ass’n v. Civil Serv. Comm’n, 463 U.S. 582 (1983). The Guardians case
is notoriously difficult, complicated by the various opinions issued in the case. In the way the
claim unfolded, the case was primarily about the government’s power to issue regulations
proscribing disparate impact. The Supreme Court has since reversed that part of the Guardians
decision. See Alexander v. Sandoval, 532 U.S. 275 (2001).
734 53 UCLA LAW REVIEW 701 (2006)
described in the next section, the two ensuing decades simply confirmed the
theory’s limited reach, even within employment discrimination, despite the
theory’s unmistakable allure among academics and advocates.
II. ASSESSING THE THEORY IN THE COURTS
To this point, I have focused on the origins of the disparate impact
theory in its particular historical context, as well as the development of the
theory in the Supreme Court. In this section, I shift focus to the way the
theory has developed in the lower courts, including an empirical assessment
of the success litigants have had under the theory. In addition to its
extended reach, one of the central attractions to disparate impact claims is the
perception that they are easier to prove than claims of intentional dis-
crimination, given that intent is often difficult to establish through circum-
stantial evidence.136 In reality, however, the opposite is true: Disparate
impact claims are more difficult to prove than standard intentional dis-
crimination claims. This is particularly significant given that employment
discrimination claims themselves are notoriously difficult to prove.137
A. The Scope of the Study
As part of this project, I have reviewed all of the disparate impact cases
from select years in both district courts and courts of appeals. As indicated
in Table A, I reviewed all of the reported appellate cases for 1984–1985,
1994–1995, and 1999–2001 for a total of six years for which there were
130 reported cases. By reported cases, I mean those cases available on
LEXIS/NEXIS, including unpublished decisions.138 In addition, I have reviewed
136. See sources cited supra notes 10, 12.
137. This principle has been documented repeatedly. For an early articulation, see
Theodore Eisenberg, Litigation Models and Trial Outcomes in Civil Rights and Prisoner Cases, 77 GEO.
L.J. 1567 (1989) (establishing that employment discrimination cases have a success rate superior
only to prisoner cases). See also Michael Selmi, Why Are Employment Discrimination Cases So Hard
to Win?, 61 LA. L. REV. 555 (2001) (exploring reasons for the difficulty of proving cases); sources
cited infra notes 151–153.
138. After collecting the cases, I excluded from the counts those that may have stated a
claim of disparate impact theory without any discussion or ruling, or where the theory was simply
mentioned, or where they were purely procedural claims including class certification and remedial
orders. There were a significant number of these cases, and it appears that many plaintiffs allege a
disparate impact theory without ever attempting to develop the theory. The search was a basic
one: date specified and “disparate impact” or “disparate w/1 impact” (there was no difference
between the two). This search picked up a significant number of nonemployment cases that were
also excluded from the analysis but which were reviewed to determine the effect the theory has
had in other areas, such as housing or environmental claims.
Was the Disparate Impact Theory a Mistake? 735
all of the cases within the district courts for six years (1983, 1987, 1991,
1996, 1999, 2002), with a total of 171 reported cases. Combining both
samples produces a total of 301 analyzed cases. In selecting the years to
study, I sought to avoid substantial overlap, but I also wanted to capture
trends that might have emerged over time, trends which turned out most
apparent in the district court cases. I coded the cases based on the court’s
determination, whether the appellate court affirmed or reversed a judgment,
either with respect to summary judgment or a trial, or reversed and
remanded for further proceedings. In the district court, decisions on the
merits were recorded for both trials and motions to dismiss, and when
plaintiffs survived a motion for summary judgment (or a motion to dismiss) the
case was categorized as a success for the plaintiff. I also noted the nature of the
claim (for example, race, sex, or age) and sought to determine the basis for
the challenge, whether, for example, it was a challenge to a test, a subjective
employment practice, or some other practice.139 Each of the cases was
checked for subsequent history but that history itself did not affect the
empirical analysis unless the case also appeared in one of the other years
under review.140
In addition to this systematic study of cases, I analyzed various seg-
ments of cases that seemed potentially revealing. I reviewed all of the
appellate court cases decided between Griggs and Washington v. Davis, as
well as many of the district court cases decided during that period, and all
cases that arose after the passage of the Civil Rights Act of 1991 that con-
tained any substantial doctrinal discussion. The latter group included fewer
than a dozen cases; one unmistakable trend is the waning importance of the
disparate impact theory after the Civil Rights Act of 1991.141 Undoubtedly,
the addition of damages for intentional discrimination claims provided by
the 1991 Act, while withholding them from disparate impact claims, has
substantially altered the incentives for defining claims as intentional dis-
crimination.142 In addition, as discussed more fully below, I analyzed certain
139. In several instances, the plaintiff prevailed on one claim but lost on others. In those
circumstances, of which there were very few, the case would be categorized as a win and a loss.
140. In other words, if a plaintiff succeeded in a district court case in 1991 but that case was
later reversed in 1993 (a year not included in the study), that reversal would be noted, but the
1991 victory would be counted as a plaintiff’s victory.
141. Ironically, the disparate impact theory was considered the most important and controversial
provision of the Act. See Adam Clymer, Senate Democrats Back a Compromise on Civil Rights Bill, N.Y.
TIMES, Oct. 26, 1991, at A1 (noting that the Administration and Congress “have fought over when
businesses can impose hiring conditions that appear fair but discriminate in practice”).
142. The Civil Rights Act of 1991 made damages available for claims of intentional
discrimination, whereas prior to that Act only equitable relief was available. Damages, however,
are not available for disparate impact claims. See Civil Rights Act of 1991, 42 U.S.C. § 1981(a)
736 53 UCLA LAW REVIEW 701 (2006)
cases and areas that have received particular attention by scholars, includ-
ing age discrimination cases and those involving subjective employment
practices, pregnancy discrimination cases, and a set of cases involving men
who wear beards. My goal was to offer a more detailed, but not fully exhaus-
tive, portrait of how the disparate impact theory has fared in the courts.
Much of the empirical analysis that follows is based on cases that are
available on the LEXIS/NEXIS database and therefore provides only a lim-
ited picture of the universe of cases. To the extent that it is a representa-
tive sample, relying on published cases can offer substantial insight into the
structure of litigation.143 This issue has been played out extensively in the
literature. Recent years also have seen a substantial increase in studies
evaluating published decisions.144 Published decisions are even more likely
to present a substantial picture of the litigation landscape for disparate
impact claims than other kinds of cases because they are typically class
actions, at least for most successful claims.145 By their nature, class action
claims have more at stake, not just in monetary terms but also in terms of
(2000). Many of the recent large class action claims have proceeded under an intentional discrimination
theory, even though many of their core allegations sound in traditional disparate impact language.
143. It is frequently noted that only about 25 percent of cases are represented in published
opinions. See Peter Siegelman & John J. Donohue, III, Studying the Iceberg From Its Tip: A
Comparison of Published and Unpublished Employment Discrimination Cases, 24 LAW & SOC’Y REV.
1133 (1990). This figure is now likely higher given that today more technically unpublished cases
are available on electronic databases.
144. See, e.g., Edward K. Cheng & Albert H. Yoon, Does Frye or Daubert Matter? A Study of
Scientific Admissibility Standards, 91 VA. L. REV. 471 (2005) (evaluating cases on removal
standards); Ruth Colker, The Americans With Disabilities Act: A Windfall for Defendants, 34 HARV.
C.R.-C.L. L. REV. 99 (1999) (analyzing published Americans with Disabilities Act (ADA) cases);
Ann Juliano & Stewart J. Schwab, The Sweep of Sexual Harassment Cases, 86 CORNELL L. REV. 548
(2001) (analyzing published sexual harassment cases); David Benjamin Oppenheimer, Verdicts
Matter: An Empirical Study of California Employment Discrimination and Wrongful Discharge Jury
Verdicts Reveals Low Success Rates for Women and Minorities, 37 U.C. DAVIS L. REV. 511 (2003)
(analyzing reported California employment law cases); Richard L. Revesz, Environmental
Regulation, Ideology, and the D.C. Circuit, 83 VA. L. REV. 1717 (1997) (reviewing published
environmental cases to determine judicial voting patterns); Vicki Schultz, Telling Stories About
Women and Work: Judicial Interpretations of Sex Segregation in the Workplace in Title VII Cases
Raising the Lack of Interest Argument, 103 HARV. L. REV. 1749 (1990) (analyzing cases where
employers assert lack of interest as a defense to sex disparities); John A. Swain & Edwin E.
Aguilar, Piercing the Veil to Assert Personal Jurisdiction Over Corporate Affiliates: An Empirical Study
of the Cannon Doctrine, 84 B.U. L. REV. 445 (2004) (reviewing published decisions).
145. There are a surprising number of individual claims, almost all of which fail. In the
context of this study, it is not particularly relevant if successful claims are overrepresented among
published cases because that would simply bias the study in favor of the disparate impact theory,
and mean that the strength of the theory would be overrepresented in the empirical study.
Similarly, among the settled cases, it is only an issue if those cases differ substantially from the
published cases, for example, if there is a higher percentage of strong plaintiff cases among the
settled cases. Without some basis, there is no particular reason to believe that would be the case,
and it is more likely that strong plaintiff and defendant cases are among the settled claims.
Was the Disparate Impact Theory a Mistake? 737
publicity and potential injunctive relief. The size of the cases, and the
prospect of costly injunctive relief, suggest that these claims are likely to be
litigated at some level rather than quickly settled, although the potential
for adverse publicity might influence some companies to settle quickly.146
And because so much is at stake, adverse decisions are more likely to be
appealed than in a typical case.
A difference may exist between published and unpublished cases in the
testing claims. Assuming that cases are settled in the shadow of the law,147
employers may have settled testing cases more readily because the law was
better established, somewhat favorable to plaintiffs, and potentially costly
to litigate. Yet, as discussed more below, the published cases also demon-
strate significant success for plaintiffs in the testing cases, so there is noth-
ing necessarily lost by excluding unpublished cases.148 At the same time, the
success rate of nontesting cases is so low that it seems inconceivable that
there is a substantial segment of such claims that defendants readily settle
without litigating in a way that would produce an opinion available in an
electronic database.149 Finally, even if the data set is not fully representative
of the universe of disparate impact claims, it remains a valuable compara-
tive resource for understanding the power and limits of the theory. Within
law, we too often focus on a very limited set of leading cases, such as the
Griggs decision, which is sometimes the only disparate impact case taught in
a course on employment law. This survey represents the most comprehen-
sive overview of the theory.
146. This is what arguably occurred in the race discrimination cases involving Texaco and
Coca-Cola. Indeed, the case against Coca-Cola was settled without any substantial motions
having been decided. In contrast, Home Depot, Denny’s, and now Wal-Mart have all engaged in
substantial litigation despite the adverse publicity. For a discussion of these cases, see Michael
Selmi, The Price of Discrimination: The Nature of Class Action Employment Discrimination Litigation
and Its Effects, 81 TEX. L. REV. 1249 (2003).
147. See Robert H. Mnookin & Lewis Kornhauser, Bargaining in the Shadow of the Law: The
Case of Divorce, 88 YALE L.J. 950 (1979) (arguing that cases are settled within the shadows of
existing law).
148. Conceivably, if there are a substantial number of unpublished but successful testing
cases, the success rates of the empirical study could be skewed if those testing cases far outnumber
unpublished unsuccessful disparate impact cases. However, as discussed in the text, because the
success rate of other kinds of disparate impact claims is so low, the claims ought to offset one
another so that the published cases reflect the broader class of claims.
149. As noted previously, cases might be settled for their nuisance value, particularly given
that litigating adverse impact claims can be quite expensive. If this is true, it is difficult to see
how this could be treated as a benefit, rather than an unintended consequence of the theory, as
presumably no one but the most zealous plaintiff advocates would countenance the creation of a
theory solely for the purpose of creating nuisance value.
738 53 UCLA LAW REVIEW 701 (2006)
B. The Success of the Disparate Impact Theory
1. Empirical Assessment
There were a total of 130 appellate cases analyzed in the manner
described above, and plaintiffs prevailed in only 19.2 percent of the cases, and
60 percent of these cases (15 of 25) were remands rather than outright
victories.150 In contrast, the majority of the defendants’ victories affirmed the
granting of summary judgment motions (62 of the 105 determinations
favorable to defendants, or 59 percent), and another 38 determinations (36.1
percent) preserved defendants’ trial verdicts. The figures are even more
dramatic when the years 1984–85 are excluded because those years accounted
for 56 percent of the plaintiffs’ successful claims (14 of 25). The success rates
were substantially lower for 1994–95 and 1999–2001, during which time
period only two successful trial verdicts were preserved on appeal.
TABLE A
DISPARATE IMPACT DETERMINATIONS
COURT OF APPEALS DECISIONS
Plaintiff
Defendant
Aff Rev Rem Total SJ Trial Rev Rem Total
1984–85 5 3 6 14 12 20 3 0 35
1994–95 1 0 2 3 14 11 0 2 27
1999–01 1 0 7 8 36 7 0 0 43
Total 7 3 15 25 62 38 3 2 105
Source: Lexis/Nexis
Plaintiff’s Success Rate:
1984–85 28.5%
1994–95 10.0%
1999–01 15.6%
Total 19.2%
150. Because of the nature of this study, I will avoid providing citations to select cases, and
where I provide citations, it will typically be for an entire category of claims, or to illustrate some
particular proposition. One of the issues one becomes acutely aware of when conducting a study
like this is that it is easy to find a single case to support a particular argument, as most lawyers
readily know. In this study, however, I want to provide a more comprehensive picture rather than
focusing on leading or illustrative cases.
Was the Disparate Impact Theory a Mistake? 739
The figures for the district court decisions are quite similar but have
some important variations. There were 171 cases in the six analyzed years,
and the plaintiffs’ succeeded in 25.1 percent of the cases. As was true with
the appellate court cases, a substantial number of what I define as successful
claims involved surviving summary judgment, and restricting the cases to
decisions on the merits lowers the success rate to 16.9 percent. Plaintiffs
fared substantially better during 1983, a year that accounts for 27.9 percent
of the successful cases, and 42.3 percent of the successful trial victories.
Excluding 1983, the plaintiffs obtained only 15 outright victories, for a suc-
cess rate on the merits of 13 percent (15 of 115).
TABLE B
DISPARATE IMPACT DETERMINATIONS
DISTRICT COURT DECISIONS
Plaintiff
Defendant
Prevails Forward Total SJ Trial Total
1983 11 1 12 3 10 13
1987 7 2 9 13 9 22
1991 4 6 10 20 5 25
1996 1 2 3 21 2 23
1999 1 4 5 18 1 19
2002 2 2 4 26 0 26
Total 26 17 43 101 27 128
Source: Lexis/Nexis
Plaintiff’s Success Rate:
1983 48.0%
1987 29.0%
1991 28.6%
1996 11.5%
1999 20.8%
2002 13.3%
Total 25.1%
Although the statistics are imperfect, they plainly reflect the difficulty of
proving disparate impact cases. Numerous studies have shown that employment
discrimination cases tend to have a success rate in federal court of
approximately 35 percent, while civil cases more broadly tend to have success
740 53 UCLA LAW REVIEW 701 (2006)
rates that approximate 50 percent.151 There are some variations within
these studies, and I have not sought to determine whether the success rate of
disparate impact cases is significantly lower statistically than employment
discrimination cases more generally,152 but no evidence seems to suggest that
disparate impact claims are easier to prove than disparate treatment claims.
Going behind the numbers also reveals the limits of the disparate impact
theory, including providing insight into why so many of the cases fail.
The cases can generally be divided into three broad categories: (1) pure
disparate impact claims; (2) mixed claims of disparate treatment and impact;
and (3) add-on claims where the disparate impact claim is added on but
never properly developed. Claims in this latter group never succeed, although
there might be some litigation advantages to asserting a disparate impact
claim, perhaps for its settlement value.153 The second category of claim is
more interesting. One important finding of this analysis is the high
percentage of successful disparate impact cases that also succeed on dispa-
rate treatment theories. Among the successful appellate court cases, nearly
one-third also succeeded on a disparate treatment claim, and one-half of
the district court cases (21 of 43) included successful disparate treatment
claims. The numbers were even higher during the early years. In 1983,
151. In a recent study, defendants obtained appellate reversals in nearly 44 percent of their
employment discrimination appeals. See Kevin M. Clermont & Theodore Eisenberg, Plaintiphobia
in the Appellate Courts: Civil Rights Really Do Differ From Negotiable Instruments, 2002 U. ILL. L.
REV. 947, 957. Plaintiffs also had an extremely low success rate following defendant trial verdicts.
Id. (noting that plaintiffs succeeded in reversing only 6 percent of defendant trial victories); see
also Kevin M. Clermont & Stewart J. Schwab, How Employment Discrimination Plaintiffs Fare in
Federal Court, 1 J. EMPIRICAL LEGAL STUD. 429 (2004) (reviewing Administrative Office of Courts
Case Data); Wendy Parker, Lessons in Losing: Employment Discrimination Cases in Federal District
Courts, 81 NOTRE DAME L. REV. (forthcoming 2006) (analyzing case files for district courts).
152. Most of the studies involve diverse databases and thus limit the prospect of simple
comparisons. I have not sought to determine whether a meta-analysis would be feasible. The success
rates also vary by time and by the nature of the claim with some evidence that age discrimination
plaintiffs fare better than other employment discrimination plaintiffs, while disability plaintiffs
have frequently fared worse. See Colker, supra note 144, at 107–10 (demonstrating success rates as
low as 5 percent for ADA plaintiffs in years following passage of the Act); George Rutherglen,
From Race to Age: The Expanding Scope of Employment Discrimination Law, 24 J. LEGAL STUD. 491, 512
(1995) (documenting 47 percent defendant success rate in non−Age Discrimination in
Employment Act (ADEA) cases but only 26.3 percent in ADEA cases, taking into account settlements).
153. A claim that never succeeds should have no settlement value or certainly nothing
above a nuisance value. However, there does seem to be widespread misperception regarding the
viability of disparate impact claims and this misperception may add some value to the claims.
One might also suggest that these cases should be excluded from the empirical analysis, which
would raise the overall success rates significantly. But it would also dampen the comparative
effect because many disparate treatment claims have little merit, and those claims are part of the
calculations in other studies. The analysis that follows focuses on substantive cases, and within
those cases the success rate is still very low.
Was the Disparate Impact Theory a Mistake? 741
three-fourths (8 of 12) of successful disparate impact claims in the district
courts also succeeded under a theory of disparate treatment. The high per-
centage of cases in 1983 supports the notion that the early cases were more
closely associated with theories of intentional discrimination, which
undoubtedly accounts for much of the early success.154
In the remaining category, pure claims, plaintiffs have achieved some
success, although that success has clearly decreased over time, particularly
after the mid-1980s. Many challenges to employment tests were successful
in the 1970s, and like Griggs, these challenges typically involved defendants
who had made little, or no, effort to validate the examinations at issue.155
With time, these challenges have become increasingly difficult for plaintiffs
as tests have become more sophisticated and professionally developed for
their particular use, and as courts have become less concerned about the
disparate impact of the tests. Courts appear far more willing to accept vali-
dation efforts today than they were a decade or two ago, and they are also
far less likely to find that a test has adverse impact than they once were.156
The latter finding is partly attributable to the change in the nature of the
cases. Adverse impact was relatively easy to identify when the tests were
administered with a large group of diverse applicants, such as with urban
police departments, but it is far more difficult to establish with small
154. See, e.g., Kilgo v. Bowman Transp., Inc., 570 F. Supp. 1509 (N.D. Ga. 1983) (holding
that required truck driver experience had disparate impact on women in a company that had long
excluded women), aff’d, 789 F.2d 859 (11th Cir. 1986) (finding the requirement was a pretext for
discrimination); Veazie v. Greyhound Lines Inc., No. 72-2729, 1983 U.S. Dist. LEXIS 11889
(E.D. La. Nov. 8, 1983) (holding that a seniority system had its genesis in intentional
discrimination); Catlett v. Mo. Highway & Transp. Comm’n, 589 F. Supp. 929 (W.D. Mo. 1983)
(holding that word of mouth recruiting caused disparate impact on women and finding disparate
treatment based in part on historical practices), aff’d, 828 F.2d 1260 (8th Cir. 1987) (focusing on
the disparate treatment claim).
155. See, e.g., Johnson v. Goodyear Tire & Rubber Co., 491 F.2d 1364, 1372–73 (5th Cir. 1974)
(enjoining the use of written employment tests until the employer made an effort to validate the
tests with the EEOC); Walston v. County Sch. Bd., 492 F.2d 919 (4th Cir. 1974) (finding the cut-
off score on a teacher test arbitrary and unvalidated); W. Addition Cmty. Org. v. Alioto, 360 F.
Supp. 733 (N.D. Cal. 1973) (upholding the plaintiffs’ successful challenge to a San Francisco
Firefighters examination for which the defendants offered no validation study); Harper v. Mayor
of Baltimore, 359 F. Supp. 1187 (D. Md. 1973) (holding that a firefighter examination was
discriminatory when only “questionable” validation effort was offered), modified and aff’d sub nom.
Harper v. Kloster, 486 F.2d 1134 (4th Cir. 1973).
156. See, e.g., Allen v. City of Chicago, 351 F.3d 306 (7th Cir. 2003) (upholding a
promotion test by rejecting the plaintiffs’ alternative practice); Lanning v. Se. Pa. Transp. Auth.,
308 F.3d 286 (3rd Cir. 2002) (upholding a physical agility examination); Firefighters’ Inst. for
Racial Equality v. City of St. Louis, 220 F.3d 898 (8th Cir. 2000) (upholding a fire department
promotional exam despite disparate impact); Williams v. Ford Motor Co., 187 F.3d 533 (6th Cir.
1999) (granting summary judgment on validity grounds); see additional cases cited infra note 225.
742 53 UCLA LAW REVIEW 701 (2006)
numbers of applicants as increasingly occurs today.157 In the early 1970s,
many jurisdictions also decided not to defend the adverse impact of their
tests, either because their validation efforts had been inadequate or for
political reasons. As I will discuss below, many governmental entities
appeared to have welcomed, at least to an extent, disparate impact
challenges as a means to diversify their workforces.158
Looking solely at the cases with successful disparate impact claims provides
additional support for the limitations of the theory. As noted previously,
many successful disparate impact claims also succeeded under a disparate
treatment approach, thus rendering the disparate impact theory largely
superfluous. Another substantial set of cases involved remands for further
determinations without any indication of what occurred on remand,
although the most likely prospect is that the cases settled and could there-
fore be treated as successful cases for the plaintiffs. There were also a small
set of successful district court cases reversed on appeal. At the same time,
among the successful cases, there was a surprising dearth of testing cases in
the later years, most likely for two reasons. A significant portion of testing
cases may have been resolved along the way because the law was most set-
tled on issues relating to testing and the standards for validation. For
employers that had not done any validation, their likelihood of success in
most jurisdictions was quite poor. As employers began to validate their exami-
nations, these cases likely migrated from successful plaintiff cases to successful
defendant cases, substantially restricting the prospect of prevailing.159
Outside of the testing cases, the successful cases are not easy to catego-
rize and instead are best described as a miscellaneous set of cases. In the
sample, there were successful challenges to various ranking systems, referral
policies, and severance pay;160 there were also several cases that mirrored
157. See, e.g., Boyd v. Borg-Warner Protective Servs., Inc., No. 98-14072-CIV-Roettger,
1999 U.S. Dist. LEXIS 13974 (S.D. Fla. July 21, 1999) (holding that the plaintiffs failed to establish
the disparate impact of a fitness test); Mems v. City of St. Paul, 73 F. Supp. 2d 1031 (D. Minn.
1999), aff’d, 224 F.3d 735 (8th Cir. 2000) (holding that statistical evidence based on a ten-person
sample was insufficient to prove disparate impact).
158. See infra Part III.A.2.
159. Another reason the cases may have receded is that many of the initial cases were filed
by the Department of Justice, which has since mostly abandoned testing cases for political reasons.
Civil rights groups also brought many challenges in the early years, and those groups have
significantly restricted their litigation over the last decade.
160. See EEOC v. Steamship Clerks Union, Local 1066, 48 F.3d 594 (1st Cir. 1995), cert.
denied, 516 U.S. 814 (1995) (invalidating a sponsorship requirement for a union with no minority
members); Crawford v. W. Elec. Co., Inc., 745 F.2d 1373 (11th Cir. 1984) (invalidating a defendant-
employer’s index review system); Caviale v. Wisconsin, 744 F.2d 1289 (7th Cir. 1984) (invalidating
participation in a career program as a prerequisite to promotion); Walker v. Jefferson County Home,
726 F.2d 1554 (11th Cir. 1984) (invalidating a requirement of prior supervisory experience).
Was the Disparate Impact Theory a Mistake? 743
some of the early claims involving patronage and degree requirements.161
These cases, however, are isolated and no obvious pattern emerged. Novel
claims were also few and far between, with the only distinctive claim
involving a challenge brought by female police officers to a police depart-
ment’s choice of gun for its officers.162 One area where the theory may have
made a difference is with residency requirements; the established principle
today is that residency requirements are permissible for employees so long as
there is a reasonable move-in period, but they are impermissible for applicants,
at least in a city where the population is homogeneous.163 A surprising area
within the sample involved several cases challenging English-only policies
where the district court allowed the claims to survive summary judgment
even though appellate courts have proved uniformly hostile to such claims.164
161. See Alexander v. Local 496, Laborers’ Int’l Union, 177 F.3d 394 (6th Cir. 1999)
(upholding a race discrimination claim against a union for failure to refer jobs); Nash v. Consol.
City of Jacksonville, 763 F.2d 1393 (11th Cir. 1985) (reversing the district court in a testing claim
for applying the wrong legal standard), rev’d, 905 F.2d 355 (11th Cir. 1990); Cooper v. Rosenberg,
694 F. Supp. 1377 (E.D. Mo. 1987) (involving patronage); Baranek v. Kelly, No. 85-0376-C, 1987
U.S. Dist. LEXIS 8338, at 17–20 (D. Mass. Sept. 9, 1987) (invalidating a bachelor’s degree
requirement that employer conceded had no business necessity).
162. See Pumphrey v. City of Coeur D’Alene, No. 92-36748, 1994 U.S. App. LEXIS 3892
(9th Cir. Feb. 24, 1994). Two other claims might also be considered novel. In Murphy v.
Derwinski, 776 F. Supp. 1466 (D. Colo. 1991), the Veteran’s Administration (VA) sought a Catholic
chaplain, a position for which a woman applied. The district court found the VA’s requirement
that the chaplain have the church’s endorsement as discriminatory under a disparate impact
approach, and the Tenth Circuit affirmed. See Murphy v. Derwinski, 990 F.2d 540 (10th Cir.
1993). This seems an unusual case for the disparate impact theory, and the religious dimensions
to the case also make it difficult to classify as a pure Title VII case. In another case, the plaintiff
survived a motion to dismiss in her challenge to the employer’s policy of refusing to grant a leave
of absence for periods of incarceration. See Butler v. Elwyn Inst., 765 F. Supp. 243 (E.D. Pa. 1991).
163. Both the Department of Justice and the NAACP filed a series of challenges to
municipal residency requirements, and when the cases were litigated, the plaintiffs typically
prevailed. See Newark Branch, NAACP v. Township of Harrison, 940 F.2d 792 (3rd Cir. 1991)
(invalidating a residency requirement); United States v. City of Warren, 759 F. Supp. 355 (E.D.
Mich. 1991) (same). A recent case, however, has demonstrated that removing residency
requirements does not necessarily increase the number of minority employees. The NAACP sued
the city of Bayonne, arguing that its residency requirement was discriminatory, and the parties
settled. During the course of the settlement, the City found that the number of minority
employees actually decreased, most likely as a result of a civil service examination administered
statewide. The Third Circuit subsequently upheld the city’s determination to reimpose its
residency requirement. See Newark Branch, NAACP v. City of Bayonne, 134 F.3d 113 (3rd Cir. 1998).
164. See EEOC v. Synchro-Start Products, Inc., 29 F. Supp. 2d 911 (N.D. Ill. 1999)
(holding that an English-only policy survived the defendant’s summary judgment motion); EEOC
v. Premier Operator Servs., Inc., 75 F. Supp. 2d 550 (N.D. Tex. 1999) (finding a prima facie case
of disparate impact and treatment for an English-only policy). In contrast, the courts of appeals
have rejected the challenges that have come before them. See Garcia v. Spun Steak Co., 998 F.2d
1480 (9th Cir. 1993); Jurado v. Eleven-Fifty Corp., 813 F.2d 1406 (9th Cir. 1987); Garcia v. Gloor,
618 F.2d 264 (5th Cir. 1980), cert. denied, 449 U.S. 1113 (1981). An exception arose in the
Ninth Circuit, which upheld a challenge to an English-only policy instituted by the Los Angeles
744 53 UCLA LAW REVIEW 701 (2006)
2. Subjective Employment Practices and Age Cases
For many years, advocates sought to extend the disparate impact the-
ory to cases involving subjective employment practices and age dis-
crimination, under the express idea that the theory would uncover more
discrimination than could be rooted out by the more common intentional
discrimination framework. In both instances, the efforts ultimately succeeded
in that the Supreme Court extended the theory to such cases. However,
these claims have been almost uniformly unsuccessful; indeed, successful
claims are almost always pursued as claims of intentional discrimination. As a
result, these two areas illustrate both the allure and the limits of the theory.
Subjective employment practices have long been the focus of dis-
crimination claims, in large part because the discretion inherent in
subjective practices can be a slippery vehicle for discrimination. These
claims can be difficult to establish as intentional discrimination because
they commonly rely on circumstantial evidence. In contrast, to the extent
that subjective practices favor a preferred group, it might be possible to
establish a significant disparate impact that would then require employers
to justify those practices. In Watson v. Fort Worth Bank and Trust,165 the
Supreme Court ultimately recognized the utility of applying the disparate
impact theory to subjective employment practices, but at the same time, the plu-
rality opinion authored by Justice O’Connor expressed concern that
employers might have difficulty justifying their practices under the business
necessity test.166 The case has been a mixed blessing for plaintiffs; the real-
ity is that subjective employment practices are almost always more success-
ful as intentional discrimination claims.
The reason has to do with the nature of subjective employment prac-
tices. There is nothing accidental about those practices; rather, employers
use such practices and make such decisions intentionally, and I would add,
consciously. Although it is true that intent can be difficult to prove, it is
Municipal Court. See Gutierrez v. Mun. Court, 861 F.2d 1187 (9th Cir. 1988). That case,
however, was subsequently vacated as moot by the Supreme Court. See Mun. Court v. Gutierrez,
490 U.S. 1016 (1989). Depending on the rationale for the policy, English-only cases can be litigated
as intentional discrimination cases, and as the appellate court cases demonstrate, when the
employer has a reasonable justification, the policy is typically upheld.
165. 487 U.S. 977 (1988) (plurality opinion).
166. Id. While permitting disparate impact challenges to subjective employment practices,
Justice O’Connor’s plurality opinion likewise noted that the burden of proof remains with the
plaintiff at all times, id. at 997, and further required the plaintiff to identify the specific
employment practice causing the disparity. Id. at 996–97. In ratcheting up the proof standards,
Justice O’Connor specifically noted the prospect that employers might “adopt surreptitious quota
systems in order to ensure that no plaintiff can establish a statistical prima facie case.” Id. at 992.
Was the Disparate Impact Theory a Mistake? 745
certainly not impossible, and often is easier to prove than through a theory
of disparate impact. As Justice O’Connor hinted, there is no ready means
to validate subjective employment practices under the disparate impact
theory, leaving courts to apply their own normative judgments regarding
whether the practices are discriminatory.167 As discussed in more detail in
the next section, the judicial inquiry turns on how the subjectivity is exer-
cised, an inquiry that lends itself to a disparate treatment analysis. For
example, if managers apply subjective criteria in a way that favors men over
women, then there is likely something about the way the criteria are used
that results in women being treated differently, and there is no reason to
avail oneself of the disparate impact theory.168 Under a disparate impact
approach, the employer would have to justify its use of subjective practices,
but short of requiring employers to prove that subjective practices are the
best employment practice, a standard the Supreme Court long ago
rejected,169 there is no basis for evaluating subjective practices other than in
how they are applied. An employer’s differential treatment often will be
the product of stereotyping rather than a demonstrable overt practice, but
contrary to the view of some commentators,170 stereotyping fits better
167. Id. at 991 (discussing the difficulty of validating subjective employment criteria such as
“common sense, good judgment, originality, ambition, loyalty, and tact”). In his concurring
opinion, Justice Blackmun referenced the possibility of validating subjective employment
practices, based on an amicus submission by the American Psychological Association. See id. at
1007 n.5 (Blackmun, J., concurring). While this may be theoretically possible, it is not at all
common and may be the product of a self-interested industry rather than something one might
reasonably expect in a workplace. More commonly, subjective practices are validated by
rendering them more objective. This process, however, is intended to severely limit discretion,
which is quite different from validating a subjective process where discretion is at the core.
168. In a recent challenge to subjective promotion practices, the Eighth Circuit explained:
It is difficult to understand this claim as one of disparate impact. Plaintiffs’ claim as to the
subjective decisionmaking process is not that this facially race-neutral process has an adverse
impact on blacks and the process cannot be justified by business necessity. Rather,
Plaintiffs claim the subjective decisionmaking resulted in blacks remaining in center-
manager positions longer than whites before they were promoted to the division-manager
level. We read Plaintiffs’ argument as alleging disparate treatment through the
subjective decisionmaking process; that is, that the subjective selection process provided
the opportunity for UPS to choose not to promote some employees because they were
black—to discriminate on account of race.
Morgan v. UPS, Inc., 380 F.3d 459, 465 n.2 (8th Cir. 2004), cert. denied, 125 S. Ct. 1933 (2005).
The appellate court applied the disparate treatment model, rejected the plaintiffs’ statistical
analysis, and upheld the lower court’s grant of summary judgment to defendants. Id. at 468–72.
169. See Furnco Constr. Corp. v. Waters, 438 U.S. 567 (1978).
170. In her plurality opinion, Justice O’ Connor also noted the importance of the
disparate impact theory for rooting out “subconscious stereotypes and prejudices . . . .”
Watson, 487 U.S. at 990. As discussed below, many academics have suggested applying disparate
impact theory, or a form of it, to capture stereotyping. See, e.g., Jody D. Armour, Race
746 53 UCLA LAW REVIEW 701 (2006)
within an intentional discrimination framework than within the business
necessity model applicable to impact claims. After all, there is no permissi-
ble business rationale for relying on negative stereotypes, and the most dif-
ficult part of the proof is establishing that stereotyping factored into the
decisionmaking process. The disparate impact model has nothing to do
with that proof, and if stereotyping can be proved, a finding of intentional
discrimination should follow.171
The recent spate of class action cases confirms the propriety of challenging
subjective practices as intentional discrimination. In the last decade, class
action attorneys have filed suits against many large employers, including
Wal-Mart and Home Depot, for using subjective employment practices that
disadvantage women.172 Although the cases frequently include allegations
to support a disparate impact claim, the cases have all proceeded primarily
as claims of intentional discrimination under the statistical pattern or
practice theory, and in each instance, the plaintiffs have sought to prove
intent through detailed statistical analyses.173 To date, the cases have all
settled, leaving little precedential trail. Yet, there is no question that the
intentional discrimination framework can serve to challenge subjective
employment practices, and there is very little to gain, and much to lose, by
resorting to the disparate impact framework.174
The same proves true for age discrimination cases, though for some-
what different reasons. Prior to the Supreme Court’s recent acceptance of
the disparate impact theory under the age discrimination statute, a number
of appellate courts had held that the theory was unavailable for age
Ipsa Loquitur: Of Reasonable Racists, Intelligent Bayesians, and Involuntary Negrophobes, 46 STAN.
L. REV. 781, 811–12 (1994) (calling for a new model directed at unconscious bias).
171. In the district court, Watson lost her disparate treatment claim. See Watson v. Fort
Worth Bank & Trust, 798 F.2d 791, 799 (5th Cir. 1986).
172. Sex discrimination lawsuits also have been filed against Costco and many grocers and
securities firms. These claims are discussed in Michael Selmi, Sex Discrimination in the Nineties, Seventies
Style: Case Studies in the Preservation of Male Workplace Norms, 9 EMP. RTS. & EMP. POL’Y J. 1 (2005).
173. Id. at 18.
174. Within the sample, I was able to identify two cases involving subjective practices that
prevailed on an impact theory but not on an intentional discrimination cause of action. In
Bullington v. United Air Lines, Inc., 186 F.3d 1301 (10th Cir. 1999), the court allowed a disparate
impact challenge to the company’s interview process to go forward while ruling against the
plaintiff’s treatment claim. The business necessity issue, however, was not addressed. Id. at 1315
n.10. And Lewis v. Bloomsburg Mills, Inc., 773 F.2d 561 (4th Cir. 1985), was principally a
disparate impact case where the plaintiffs succeeded in their challenge to assignment and
promotion policies, although it could also be read as a disparate treatment claim because the court
moved between the two standards.
Was the Disparate Impact Theory a Mistake? 747
claims.175 Although these courts typically applied some statutory analysis,
their real concern was normative: Age claims might prove too disruptive to
standard business practices, many of which frequently have a disparate
effect on older workers. In these courts, age claims involving disparate
impact theories were often seen as the discrimination equivalent of
securities cases, where plaintiffs’ attorneys are sometimes accused of filing
suit any time there is a significant stock price drop. Age discrimination
claims were perceived as routinely following mass layoffs or reductions in
force, which frequently would adversely impact older workers who were
often more expensive than their younger counterparts even when their
experience and productivity were taken into account. As Judge Posner
once noted, allowing disparate impact claims in these situations is highly
problematic and might “as a practical matter forbid all firms to reduce wages
or fringe benefits in periods of adversity.”176 Judge Posner’s statement cap-
tures the prevailing sentiment that the disparate impact theory simply
proved too much, unless the plaintiffs could show that older employees had
been targeted because of their age rather than because of their expense.177
Of course, if the plaintiffs could make that showing, they would succeed on
an intentional discrimination claim.
175. The cases are catalogued and discussed in Kenneth R. Davis, Age Discrimination and
Disparate Impact: A New Look at an Age-Old Problem, 70 BROOK. L. REV. 361 (2004).
176. Finnegan v. Trans World Airlines, Inc., 967 F.2d 1161, 1164 (7th Cir. 1992). Judge
Posner’s influential opinion, in a case that challenged TWA’s decision to cap vacation time as
part of a bankruptcy proceeding, is worth quoting at length:
There is something wrong with an interpretation of the Age Discrimination in
Employment Act that forbids a bankrupt corporation to adopt a companywide policy of
limiting paid vacations to 4 weeks a year and that would as a practical matter forbid all
firms to reduce wages or fringe benefits in periods of adversity.
Id. Earlier in the opinion Judge Posner had noted:
A company that for legitimate business reasons decides to cut wages across the board, or to cut
out dental insurance, or to curtail the use of company cars is not required to conduct a study
to determine the impact of the measure on employees grouped by age and if it is
nonrandom to prove that the same amount of money could not have been saved in some
different fashion.
Id. at 1163.
177. See, e.g., Allen v. Diebold, Inc., 33 F.3d 674, 677 (6th Cir. 1994) (“The ADEA was not
intended to protect older workers from the often harsh economic realities of common business
decisions and the hardships associated with corporate reorganizations, downsizing, plant closings
and relocations.”); Barnes v. GenCorp, Inc., 896 F.2d 1457, 1466–67 (6th Cir. 1990) (suggesting
that in a reduction in force case plaintiffs must show they were singled out because of age). As an
example of the road to which the disparate impact theory can lead, in Stutts v. Sears, Roebuck and
Co., 855 F. Supp. 1574 (N.D. Ala. 1994), the plaintiffs challenged a companywide compensation
plan and proposed as an alternative an entirely different wage structure, much like a management
consultant might do. The court rejected this effort. Id. at 1581.
748 53 UCLA LAW REVIEW 701 (2006)
The Supreme Court’s recent decision in Smith v. City of Jackson178 con-
firms the limited thrust of the disparate impact theory for age dis-
crimination claims. In Smith, the Court upheld the applicability of the
disparate impact theory to age discrimination claims largely borrowing from
its prior cases, including Watson.179 Yet, the Court also upheld the
employer’s practice of providing larger raises to younger employees to help
retain them as a reasonable practice that satisfied the business necessity
prong of the age discrimination inquiry.180 This step was both illustrative
and surprising for at least two reasons: It was issued by the liberal wing of
the Court in an opinion authored by Justice Stevens, and the question of
the reasonableness of the policy had not been briefed by any of the parties
or addressed by the lower courts.181 Rather, the Court summarily approved
of the practice, removing any doubts that the disparate impact theory would
prove anything but ephemeral under the age discrimination statute. Again,
the empirical sample supports this idea because there were no successful age
discrimination claims on the merits, and challenges to reductions in force
have regularly been rejected.182
178. 125 S. Ct. 1536 (2005).
179. See id.
180. Id. at 1546. The age discrimination statute is structured differently from Title VII in
that it allows employers to make decisions based on “reasonable factors other than age.” 29 U.S.C.
§ 623(f)(1) (2000). The Supreme Court interpreted this provision as part of the employer’s
defense to a disparate impact inquiry and afforded what appears to be a lower threshold for the
employer to meet. In the context of the age discrimination statute, an important unresolved
question which goes to the core of class claims is whether it is permissible for employers to adopt
cost saving measures that will have a greater effect on older workers, an inquiry that can be
pursued under either an intent or impact framework. Compare Allen, 33 F.3d at 677 (“[P]laintiffs
must allege that Diebold discriminated against them because they were old, not because they were
expensive.”) with Leftwich v. Harris-Stowe State Coll., 702 F.2d 686, 691–92 (8th Cir. 1983)
(holding that a cost savings justification was insufficient). See also Hazen Paper Co. v. Biggins, 507 U.S.
604 (1993) (holding that a plaintiff must show more than factors correlated with age to establish
intentional discrimination).
181. The only surprise was that the Court upheld the theory with Justice Scalia deferring to
the EEOC’s regulations. See Smith, 125 S. Ct. at 1547 (Scalia, J., concurring in part and
concurring in the judgment). Justices O’Connor, Kennedy, and Thomas concurred in the
judgment but under the rationale that the disparate impact theory was not available under the
ADEA. Id. at 1549 (O’Connor, J., concurring in the judgment).
182. For cases dismissing challenges to reductions in force, see Smith v. Xerox Corp., 196 F.3d
358 (2d Cir. 1999); Allen v. Entergy Corp., 181 F.3d 902 (8th Cir. 1999); Graffam v. Scott Paper
Co., No. 95-1046, 1995 U.S. App. LEXIS 17120 (1st Cir. July 14, 1995); Cronin v. Aetna Life Ins.
Co., 46 F.3d 196 (2d Cir. 1995); Armbruster v. Unisys Corp., No. 91-5948, 1996 U.S. Dist. LEXIS
1340 (E.D. Pa. Feb. 8, 1996). There were, however, several cases in which the plaintiffs survived
summary judgment. See Arnett v. Cal. Pub. Employees Ret. Sys., 179 F.3d 690 (9th Cir. 1999)
(permitting a challenge to pension benefits proceeds on both disparate treatment and impact
theories), vacated, 528 U.S. 1111 (2000); Houghton v. SIPCO, Inc., 38 F.3d 953 (8th Cir. 1994)
(reversing the district court for applying the wrong legal standard); Monroe v. United Air Lines, Inc.,
Was the Disparate Impact Theory a Mistake? 749
The pursuit of the disparate impact theory for subjective practices and
age discrimination cases also reveals what is perhaps the most common mis-
take underlying the disparate impact theory. The expectation that these
claims would be easier to establish than intentional discrimination claims
rests entirely on the first part of the theory regarding the prima facie case of
discrimination, but ignores the business necessity prong, which has always
proved the greater hurdle.
3. Pregnant Women and Bearded Men
Scholars have highlighted two particular issues and associated cases to
demonstrate the power of the disparate impact theory: cases involving
claims brought by pregnant women and challenges to policies prohibiting
employees from wearing beards. While these two particular issues are unre-
lated, I want to explore them together to illustrate how limited the theory is
even when it proves successful, as well as the thin reeds on which the myth
of the disparate impact theory persists.
Many scholars have advocated the use of the disparate impact theory
to address workplace structures that disadvantage women, and they rely on
a series of cases, mostly arising in the district courts, challenging various
leave policies as disparately impacting women.183 A leading case is EEOC v.
Warshawsky & Co., where the district court granted summary judgment for a
plaintiff who challenged an employer’s policy of prohibiting new employees from
taking sick leave during the first year of employment, a policy the court found
significantly disadvantaged pregnant women without an adequate employer
736 F.2d 394 (7th Cir. 1984) (noting that the plaintiffs survived summary judgment in a challenge
to a sixty-year requirement for pilots on both disparate treatment and disparate impact); Camacho v.
Sears Roebuck, 939 F. Supp. 113 (D.P.R. 1996) (allowing a salary structure challenge to go forward).
I have opted not to provide numerical totals for age discrimination claims because a significant
number of the cases were dismissed in light of circuit court determinations that the disparate impact
theory was unavailable for age discrimination cases. Such determinations have now been reversed,
but as evident in the Smith case, that does not mean the cases would have proved successful.
183. See, e.g., Deborah A. Calloway, Accommodating Pregnancy in the Workplace, 25 STETSON L.
REV. 1, 42 (1995) (“Disparate impact analysis can be used to resolve many of the accommodation
problems faced by pregnant women.”); Michelle A. Travis, Recapturing the Transformative Potential
of Employment Discrimination Law, 62 WASH. & LEE L. REV. 3, 84 (2005) (identifying the
disparate impact theory as a “tool for transforming the workplace”); 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, 134–36 (2003) (“If, for example, an employer does not permit
employees with medical needs to take leave, request light duty, take bathroom breaks, or work a flexible
schedule, a pregnant woman may be able to challenge that policy on the ground that it has a
disparate impact on women.”). Admittedly, most of those who tout the disparate impact theory in
the context of pregnancy-related claims acknowledge what Christine Jolls defines as “unrest, and
even some outright conflict, in the case law.” Jolls, supra note 14, at 663.
750 53 UCLA LAW REVIEW 701 (2006)
justification.184 The Court of Appeals for the D.C. Circuit also invalidated
a union policy that permitted a maximum of ten days of leave because of its
adverse effect on pregnant women.185 These cases, however, turn out to be
rather isolated examples of success, and they are countered by a far larger
array of unsuccessful cases. The Seventh Circuit, out of which the
Warshawsky case arose, has in fact upheld restrictive leave policies
instituted by other school districts, and specifically disclaimed any need to
treat pregnancy differently from other disabilities.186 Courts have also
routinely denied challenges to part-time work, light duty requests, and
disability policies when the requests were made to accommodate
pregnancy,187 and two appellate courts have questioned the application of
the disparate impact theory to pregnancy claims at all.188
184. See EEOC v. Warshawsky & Co., 768 F. Supp. 647 (N.D. Ill. 1991). This case also can
be seen as a form of intentional discrimination. The employer offered no meaningful justification
for its policy, and the justifications it did offer easily could be satisfied from a more reasonable
probationary policy of sixty or ninety days. See id. at 655.
185. Abraham v. Graphic Arts Int’l Union, 660 F.2d 811 (D.C. Cir. 1981). Another cas