Reclaiming McDonnell Douglas

Notre Dame Law Review

Volumn 83, Issue 1, 2007, Pages 109-184

  Katz, Martin J ^a  

^a UNIVERSITY OF DENVER   (United States)

Author keywords

[No Author keywords available]

Indexed keywords

EID: 38449086311     PISSN: 07453515     EISSN: None     Source Type: Journal    
DOI: None     Document Type: Review

References (279)

1. McDonnell Douglas Corp. v. Green, 411 U.S. 792 (1973). To be more precise, I come to defend the pretext method of proof set forth in the third stage of McDonnell Douglas. I see little use for the first two stages of McDonnell Douglas. See infra note 91.
2. There are two distinct theories of employment discrimination. Disparate treatment involves so-called intentional discrimination, in which the employer makes an adverse decision based on a protected characteristic, such as race or sex. 1 BARBARA LINDEMANN ET AL.., EMPLOYMENT DISCRIMINATION LAW 10-11 (4th ed. 2007). Disparate impact discriminalion involves decisions based on nonprotected characteristics, such as performance on a pre-employment test, which correlate with protected characteristics. See id. at 110-11 (listing theories of causation, including disparate treatment and disparate impact). This Article focuses exclusively on disparate treatment, the most common theory in employment discrimination litigation.
3. See GEORGE RUTHERGLEN, EMPLOYMENT DISCRIMINATION LAW 36 (2d ed. 2007) ("No decision in employment discrimination law has been cited more frequently than McDonnell Douglas Corp. v. Green.").
4. See Trans World Airlines, Inc. v. Thurston, 469 U.S. 111, 121 (1985) ("The shifting burdens of proof set forth in McDonnell Douglas are designed to assure that the 'plaintiff [has] his day in court despite the unavailability of direct evidence.'" (quoting Loeb v. Textron, Inc., 600 F.2d 1003, 1014 (1st Cir. 1979))); Tex. Dep't of Cmty. Affairs v. Burdine, 450 U.S. 248, 256-58, 255 n.8 (1981) (stating that the purpose of the framework is "to sharpen the inquiry into the elusive factual question" of discrimination); Tristin K. Green, Comment, Making Sense of the McDonnell Douglas Framework: Circumstantial Evidence and Proof of Disparate Treatment Under Title VII, 87 CAL. L. REV. 983, 998 (1999) (noting that McDonnell Douglas was designed as a response to the increasing rarity of "smoking gun" evidence in order to permit plaintiffs without such evidence to have their day in court).
5. See St. Mary's Honor Ctr. v. Hicks, 509 U.S. 502, 511 (1993) (holding that a plaintiff who proves pretext in the third stage does not necessarily win; a fact finder may still find that there was no discrimination); Burdine, 450 U.S. at 254-56 (holding that the defendant's burden in the second stage of McDonnell Douglas is only a burden of production, not a burden of persuasion); see also William R. Corbett, McDonnell-Douglas, 1973-2003: May You Rest in Peace?, 6 U. PA. J. LAB. & EMP. L. 199, 202-03 (2003) (arguing that McDonnell Douglas fell into disfavor largely as a result of later holdings that arguably limited the framework).
6. See William R. Corbett, The "Fall" of Summers, the Rise of "Pretext Plus, " and the-Escalating Subordination of Federal Employment Discrimination Law to Employment at Will: Lessons from McKennon and Hicks, 30 GA. L. REV. 305, 331-32 (1996) (noting that Hicks "substantially weakened plaintiffs' chances of winning cases at the third stage of the McDonnell Douglas analysis"); Kenneth R. Davis, Price -Fixing: Refining the Price Waterhouse Standard and Individual Disparate Treatment Law, 31 FLA. ST. U. L. REV. 859, 862 (2004) ("Hicks and Reeves robbed the [McDonnell Douglas] approach of its vitality.") ; Shannon R. Joseph, Note, Employment Discrimination: Shouldering the Burden of Proof After St. Mary's Honor Center v. Hicks, 29 WAKE FOREST L. REV. 963, 988 (1994); see also Deborah C. Malamud, The Last Minuet: Disparate Treatment After Hicks, 93 MICH. L. REV. 2229, 2235 & n.28 (1995) (discussing the outcry in commentary about Hicks).
7. See, e.g., Michael Z. Green, Addressing Race Discrimination Under Title VII After Forty Years: The Promise of ADR as Interest-Convergence, 48 How. LJ. 937, 953-54 (2005); Michael Selmi, Why Are Employment Discrimination Cases So Hard to Win?, 61 LA. L. REV. 555, 559-61 (2001).
8. See, e.g., Henry L. Chambers, Jr., The Effect of Eliminating Distinctions Among Title VII Disparate Treatment Cases, 57 SMU L. REV. 83, 99-100 (2004); Chad Derum & Karen Engle, The Rise of the Personal Animosity Presumption in Title VII and the Return to "No Cause." Employment, 81 TEX. L. REV. 1177, 1181 (2003); Tristin K. Green, Discrimination in Workplace Dynamics: Toward a Structural Account of Disparate Treatment Theory, 38 HARV. C.R.-C.L. L. REV. 91, 111 (2003); Melissa Hart, Subjective Decisionmaking and Unconscious Discrimination, 56 ALA. L. REV. 741, 758 (2005); Malamud, supra note 6, at 2236-38.
9. See Price Waterhouse v. Hopkins, 490 U.S. 228, 261-79 (1989) (O'Connor, J., concurring), superseded by statute. Civil Rights Act of 1991, Pub. L. No. 102-166, 105 Stat. 1071.
10. See Civil Rights Act of 1991 g 107, 42 U.S.C. §§ 2000e-2(m), 2000e-5(g) (2000).
11. See Corbett, supra note 5, at 200, 219 (suggesting that plaintiff's and employee rights advocates prefer the "mixed-motive" frameworks of Price Waterhouse and the 1991 Act, whereas defense lawyers prefer McDonnell Douglas). Some plaintiffs, however, prefer McDonnell Douglas, as they are willing to take on what they believe to be a higher burden in exchange for avoiding the "same decision "/"same action" defense. See Michael J. Zimmer, The New Discrimination Law: Price Waterhouse is Dead, Whither McDonnell Douglas?, 53 EMORY L.J. 1887, 1942-43 (2004). As I will show below, McDonnell Douglas does not in fact set a higher burden than the alternative frameworks. See infra Part I.C.3 (arguing that McDonnell Douglas does not prove or require "but for" causation). And using McDonnell Douglas does not always preclude a "same decision"/"same action" defense. See infra note 123 and accompanying text.
12. See, e.g., Tysinger v. Police Dep't, 463 F.3d 569, 577-78 (6th Cir. 2006) (discussing the different approaches courts have taken in applying McDonnell Douglas as opposed to the alternative frameworks); Sallis v. Univ. of Minn., 408 F.3d 470, 474-75 (8th Cir. 2005) (deliberating about which framework to apply and deciding to apply McDonnell Douglas rather than Price Waterhouse or the 1991 Act).
13. See infra Part I.A (discussing courts' methods of mandating McDonnell Douglas). There are actually two ways in which courts can-and do-mandate the use of McDonnell Douglas. First, a court might require the litigants to go through the three steps that comprise the framework. See, e.g., Stover v. Martinez, 382 F.3d 1064, 1071 (10th Cir. 2004) (requiring the plaintiff to "state a prima facie case of retaliation" under McDonnell Douglas). Second, a court might require the plaintiff to prove "pretext"-that is, to disprove the reason proffered by the defendant for its challenged action. See, e.g., Merillat v. Metal Spinners, Inc., 470 F.3d 685, 692-94 (7th Cir. 2006). This Article is primarily concerned with the second form of mandate (requiring the use of "pretext" proof), arguing that this method of proof should be optional. However, once the pretext method of proof is optional, it makes little sense to require litigants to go through the first two steps of the three-step framework (the first form of mandate). See infra note 91.
14. It is difficult to find more than a handful of writers who defend the framework. See Christopher R. Hedican et al., McDonnell Douglas: Alive and Well, 52 DRAKE L. REV. 383, 395-402, 425 (2004) (defending McDonnell Douglas as "a fair and appropriate way to ferret out discrimination"); Steven J. Kaminshine, Disparate Treatment as a Theory of Discrimination: The Need for a Restatement, Not a Revolution, 2 STAN. J. C.R. & C.L. 1, 38-62 (2005) (arguing for the retention of the McDonnell Douglas framework as one possible method for plaintiffs to present proof of disparate treatment); Zimmer, supra note 11, at 1933 (arguing that McDonnell Douglas should continue to play a limited role in disparate treatment cases because its "process of elimination is a fundamentally sound way of persuading the factfinder that discrimination was involved in the [defendant's] decision"). None of these defenders answers the critiques that have been leveled at McDonnell Douglas.
15. See, e.g., Kenneth R. Davis, The Stumbling Three-Step, Burden-Shifting Approach in Employment Discrimination Cases, 61 BROOK. L. REV. 703, 744-60 (1995); Derum & Engle, supra note 8, at 1188-90 (discussing unconscious bias); Linda Hamilton Krieger, The Content of Our Categories: A Cognitive Bias Approach to Discrimination and Equal Employment Opportunity, 47 STAN. L. REV. 1161, 1241 (1995); Malamud, supra note 6, at 2236-38 (criticizing the McDonnell Douglas framework and arguing for its abandonment); Marcia L. McCormick, The Allure and Danger of Practicing Law as Taxonomy, 58 ARK. L. REV. 159, 183 (2005); Lawrence D. Rosenthal, Motions for Summary Judgment when Employers Offer Multiple Justifications for Adverse Employment Actions: Why the Exceptions Should Swallow the Rule, 2002 UTAH L. REV. 335, 335; Stephen W. Smith, Title VII's National Anthem: Is There a Prima Facie Case for the Prima Facie Case?, 12 LAB. LAW. 371, 372-81 (1997); Jeffrey A. Van Delta, "Le Roi Est Mort; Vive le Roi": An Essay on the Quiet Demise of McDonnell-Douglas and the Transformation of Every Title VII Case After Desert Palace, Inc. v. Costa into a "Mixed Motives" Case, 52 DRAKE L. REV. 71, 90-108 (2003) [hereinafter Van Delta, Le Roi Est Mort]; Jeffrey A. Van Delta, Requiem for a Heavy weight: Costa as Countermonument to McDonnell Douglas-A Countermemory Reply to Instrumentalism, 67 ALB. L. REV. 965, 1011 (2004) (referring to McDonnell Douglas as a "creeping malaise . . . that evolved to euthanize Title VII cases"); Amy L. Wax, Discrimination as Accident, 74 IND. L.J. 1129, 1148 (1999); see also Corbett, supra note 5, at 202 (noting that McDonnell Douglas "has not been very popular among many scholars and employee rights advocates, and many have called for its abandonment").
16. See, e.g., Griffith v. City of Des Moines, 387 F.3d 733, 745 (8th Cir. 2004) (Magnuson, J., concurring specially) ("For thirty years, courts have been slaves to the McDonnell Douglas burden shifting paradigm that is inconsistent with Title VII."); Wells v. Colo. Dep't of Transp., 325 F.3d 1205, 1221 (10th Cir. 2003) (Hartz, J., concurring) (writing separately to express "displeasure" with the use of the McDonnell Douglas framework because "McDonnell Douglas has served its purpose and should be abandoned"); Dare v. Wal-Mart Stores, Inc., 267 F. Supp. 2d 987, 991-92 (D. Minn. 2003).
17. See infra Part I.A.
18. See Costa v. Desert Palace, Inc., 299 F.3d 838, 851-54 (9th Cir. 2002) (noting that this area of law has been referred to as "a quagmire," a "morass," and "chaos"), aff'd, 539 U.S. 90 (2003); Benjamin C. Mizer, Note, Toward a Motivating Factor Test for Individual Disparate Treatment Claims, 100 MICH. L. REV. 234, 269 (2001) (using the term "morass"); Zimmer, supra note 11, at 1916 (using the term "morass" (quoting Desert Palace, 299 F.3d at 853)); see also Robert Belton, Mixed-Motive Cases in Employment Discrimination Law Revisited: A Brief Updated View of the Swamp, 51 MERCER L. REV. 651, 654 (2000) (referring to McDonnell Douglas and the other analytical frameworks created in Price Waterhouse and the 1991 Act as a "substantive and procedural swamp in employment discrimination law"); Paul N. Cox, Substance and Process in Employment Discrimination Lam: One View of the Swamp, 18 VAL. U. L. REV. 21, 22 (1983) (describing employment discrimination law as "incoherent").
19. See infra Part II.C.
20. 539 U.S. 90 (2003).
21. See Chambers, supra note 8, at 95-99; Corbett, supra note 5, at 200; Davis, supra note 6, at 907; Van Detta, Le Roi Est Mort, supra note 15, at 72 (describing McDonnell Douglas as "'dead as a doornail'" (quoting CHARLES DICKENS, A CHRISTMAS CAROL (John C. Winston Co. 1938) (1849)); see also Kaminshine, supra note 14, at 7 (using the term "dead"); Charles A. Sullivan, Disparate Impact: Looking Past the Desert Palace Mirage, 47 WM. & MARY L. REV. 911, 935 (2005) ("McDonnell Douglas may be either doctrinally or functionally dead. . . ."); Zimmer, supra note 11, at 1932 (arguing that McDonnell Douglas is mostly dead). But see Hedican et al., supra note 14, at 383 (arguing that McDonnell Douglas is "alive and well"); Matthew R. Scott & Russell D. Chapman, Much Ado About Nothing-Why Desert Palace Neither Murdered McDonnell Douglas nor Transformed All Employment Discrimination Cases to Mixed Motive, 36 ST. MARY'S L.J. 395, 405 (2005) ("[N]othing in Desert Palace hints at the death or even wounding of McDonnell Douglas.").
22. "Dead" critics frequently cite Dare v. Wal-Mart Stores, Inc., 267 F. Supp. 2d 987, 991-92 (D. Minn. 2003), as adopting this position. See Davis, supra note 6, at 890-91 (suggesting that the court in Dare found McDonnell Douglas to be dead in light of Desert Palace); Van Delta, Le Roi Est Mort, supra note 15, at 139-42 (discussing the court's reasoning in Dare as rejecting McDonnell Douglas in favor of Desert Palace). But Dare only declared that, under the 1991 Act, there was no such thing as a "single motive" case, in which, unlike "mixed motive" cases, the "direct evidence" distinction (and thus McDonnell Douglas) might continue to apply after Desert Palace. Dare, 267 F. Supp. 2d at 991-92. It did not purport to eradicate McDonnell Douglas in cases outside the 1991 Act. Moreover, despite Dare's hostility to McDonnell Douglas, to the extent that it purported to eradicate McDonnell Douglas even in 1991 Act cases, that case has been overruled. See Griffith v. City of Des Moines, 387 F.3d 733, 735-36 (8th Cir. 2004) (giving plaintiffs a choice about whether to use McDonnell Douglas in 1991 Act cases).
23. See George Loewenstein & Don A. Moore, When Ignorance is Bliss: Information Exchange and Inefficiency in Bargaining, 33 J. LEGAL STUD. 37, 43 (2004) ("In the presence of uncertainty, the expectations of the two parties are likely to diverge, and negotiators can easily fail to agree despite the potential for profitable settlement." (citing Kalyan Chatterjee & Larry Samuelson, Bargaining with Two-Sided Incomplete Information: An Infinite Horizon Model with Alternating Offers, 54 REV. ECON. STUD. 175, 175-92 (1987))); cf. Charles Silver, Does Civil Justice Cost Too Much?, 80 TEX. L. REV. 2073, 2107 (2002) ("When parties agree on expected trial results (as fully informed, rational parties always should), they should settle to minimize transaction costs.").
24. See Malamud, supra note 6, at 2236-38 (arguing that the McDonnell Douglas proof structure "constraints] fact finding" and should be abandoned).
25. See Van Delta, Le Roi Est Mort, supra note 15, at 138-39 & n.332 (suggesting that the shift from McDonnell Douglas in light, of Desert Palace will occur gradually over time rather than immediately, and noting several cases in which courts "have not fully grasped or discussed" the full implications of Desert Palace).
26. See infra note 50 and accompanying text.
27. Charles A. Sullivan, Circling Back to the Obvious: The Convergence of Traditional and Reverse Discrimination in Title VII Proof, 46 WM. & MARY L. REV. 1031, 1060 n.122 (2004) ("For more than a decade after it was decided, McDonnell Douglas was the only game in town for individual disparate treatment cases.").
28. Price Waterhouse v. Hopkins, 490 U.S. 228, 261-79 (1989) (O'Connor, J., concurring), superseded by statute, Civil Rights Act of 1991, Pub. L. No. 102-166, 105 Stat. 1071.
29. See McDonnell Douglas Corp. v. Green, 411 U.S. 792, 802-04 (1973). The word "pretext" is ambiguous, having been used in a number of different ways. For a discussion of the best understanding of this word, see infra note 76.
30. See Price Waterhouse, 490 U.S. at 276-77 (O'Connor, J., concurring). At some points, Justice O'Connor uses the term "substantial factor," rather than "motivating factor." See, e.g., id. at 265. However, there is no logical distinction between these two formulations. See Martin J. Katz, The Fundamental Incoherence of Title VII Making Sense of Causation in Disparate Treatment Law, 94 GEO. L.J. 489, 503-11 (2006).
31. See Price Waterhouse, 490 U.S. at 276 (O'Connor, J., concurring).
32. See Zimmer, supra note 11, at 1910; see also Griffith v. City of Des Moines, 387 F.3d 733, 743 (8th Cir. 2004) (Magnuson, J., concurring specially) (noting that after Price Waterhouse, courts follow Justice O'Connor's direct evidence distinction).
33. Price Waterhouse, 490 U.S. at 276 (O'Connor, J., concurring).
34. Id.
35. Id.
36. Id.
37. Id.
38. 42 U.S.C. § 2000e-2(m) (2000) (providing that a plaintiff must show that a protected characteristic was a "motivating factor" in the adverse decision); id. § 2000e-5(g)(2)(B) (providing that once a plaintiff has done so, the defendant may demonstrate that it would have taken the "same action" absent consideration of the protected characteristic).
39. Id. § 2000e-5(g)(2)(B).
40. See Desert Palace, Inc. v. Costa, 539 U.S. 90, 95, 97-98 (2003).
41. See H.R. REP. No. 102-40(I), at 45-49 (1991), as reprinted in 1991 U.S.C.C.A.N. 549, 583-87 (discussing the need to overturn Price Waterhouse).
42. See Desert Palace, 539 U.S. at 95 (indicating that circuits outside of the Ninth required 1991 Act plaintiffs without "direct evidence" to use McDonnell Douglas); Kaminshine, supra note 14, at 28-29 (same). Several circuits had ruled on the question at that point. See, e.g., Mohr v. Dustrol, Inc., 306 F.3d 636, 639-40 (8th Cir. 2002); Fernandes v. Costa Bros. Masonry, Inc., 199 F.3d 572, 580 (1st Cir. 1999); Trotter v. Bd. of Trs. of Univ. of Ala., 91 F.3d 1449, 1453 (11th Cir. 1996); Fuller v. Phipps, 67 F.3d 1137, 1141-42 (4th Cir. 1995); see also Green, supra note 4, at 992-93 (noting post-1991 Act courts that applied "direct evidence" distinction from Price Waterhouse).
43. See Desert Palace, 539 U.S. at 101-02. The portion of the 1991 Act that is relevant to the issues discussed in this Article amended section 703(a) of Title VII (codified at 42 U.S.C. § 2000e-2(m)), as well as section 706(g) of Title VII (codified at 42 U.S.C. § 2000e-5(g)(2)(B)). See Pub. L. No. 102-166, § 107, 105 Stat 1071, 1075-76 (1991). The amended part of section 706(g) relates specifically to cases brought under section 703(a). Accordingly, for purposes of this Article (and in disparate treatment law generally), we can refer to cases brought under section 703(a) as "1991 Act cases," and cases brought under any other disparate treatment statute as "non-1991 Act cases."
44. See, e.g., EEOC v. Warfield-Rohr Casket Co., 364 F.3d 160, 164 n.1 (4th Cir. 2004) (assuming that "direct evidence is still a prerequisite for a mixed-motive analysis in ADEA cases"); Mereish v. Walker, 359 F.3d 330, 340 (4th Cir. 2004) (discussing why the Price Waterhouse "direct evidence" distinction should remain applicable to ADEA claims after Desert Palace); Monaco v. Am. Gen. Assurance Co., 359 F.3d 296, 300 (3d Cir. 2004) (applying the "direct evidence" distinction in an ADEA case); Harp v. Se. Pa. Transp. Auth., No. Civ.A. 04-2205, 2006 WL 1517390, at *6-7 (E.D. Pa. May 31, 2006) (applying the "direct evidence" distinction in an ADA case). But see Richardson v. Monitronics Int'l, Inc., 434 F.3d 327, 334-35 (5th Cir. 2005) (applying Desert Palace in a FMLA case); Rachid v. Jack in the Box, Inc., 376 F.3d 305, 311 (5th Cir. 2004) (applying the reasoning of Desert Palace in an ADEA case).
45. 42 U.S.C. §§ 12101-12213 (2000 & Supp. IV 2004).
46. 29 U.S.C. §§ 621-634 (West 1999 & Supp. 2007).
47. See e.g., Richardson, 434 F.3d at 333; Bloomer v. United Parcel Serv., Inc., 94 F. App'x 820, 826 (10th Cir. 2004); Winter v. Bank of Am., No. Civ.A.3:02-CV-1591-L, 2003 WI. 23200278, at *3 (N.D. Tex. Dec. 12, 2003).
48. See, e.g., Keelan v. Majesco Software, Inc., 407 F.3d 332, 341 (5th Cir. 2005); Rachid, 376 F.3d at 312; Rishel v. Nationwide Mut. Ins. Co., 297 F. Supp. 2d 854, 864-65 (M.D.N.C. 2004); see also Brown v. Westaff (USA), Inc., 301 F. Supp. 2d 1011, 1016-18 & n.6 (D. Minn. 2003) (discussing the implications of Desert Palace); Dunbar v. Pepsi-Cola Gen. Bottlers of Iowa, Inc., 285 F. Supp. 2d 1180, 1196 (N.D. Iowa 2003) (same). But see Griffith v. City of Des Moines, 387 F.3d 733, 735-36 (8th Cir. 2004) (not adopting the modified version of McDonnell Douglas). For details on this modification, see infra note 136.
49. Of course, it would make sense to require McDonnell Douglas if the legislature so mandated. However, McDonnell Douglas is a creature of the common law; it is not mentioned in Title VII or any other major disparate treatment statute. See Sandra F. Sperino, Flying Without A Statutory Basis: Why McDonnell Douglas is Not Justified by Any Statutory Construction Methodology, 43 HOUS. L. REV. 743, 762-90 (2006) (arguing that the McDonnell Douglas framework should be abandoned because it is not supported by the text, legislative history or purpose of Title VII, or by any accepted principle of statutory construction). One might also argue that it makes "sense" for courts to require McDonnell Douglas as long as the Supreme Court has told them to. This would not meet my definition of making sense. A flawed interpretation of a statute by the Supreme Court is still a Hawed interpretation and should be changed. But, as I will argue in Part II.B, infra, the Supreme Court has fortunately given no such instructions.
50. See 42 U.S.C. § 2000e-2(a) (2000). Numerous other antidiscrimination laws use similar language. See, e.g., Age Discrimination in Employment Act (ADEA), 29 U.S.C.A § 623 (West 1998 & Supp. 2007); Americans with Disabilities Act (ADA), 42 U.S.C § 12112 (2000). As should be clear from the text, disparate treatment law also requires some kind of adverse employment action. See 42 U.S.C. § 2000e-2(a); 14A C.J.S. Civil Rights § 236 (2006). However, McDonnell Douglas does not implicate this requirement. See McDonnell Douglas Corp. v. Green, 411 U.S. 792, 802 (1973) (omitting an adverse employment action from analysis of pretext).
51. There are actually three available concepts in causal logic: necessity, sufficiency, and minimal causation. See Katz, supra note 30, at 503-04. However, only two of these concepts-necessity and minimal causation-appear in current disparate treatment law. Id. I have argued that this is a mistake; as a normative matter, disparate treatment law should be changed to impose sanctions based on the concept of sufficiency. Id. at 541-44.
52. See id. at 503-07.
53. More precisely, a factor is minimally causal if it has a tendency to bring about the decision (firing), but does not rise to the level of being necessary or sufficient to that decision. See id. at 506. In the text, I omit sufficiency to avoid confusion, because that concept does not appear in current disparate treatment law. See supra note 51.
54. See Katz, supra note 30, at 503-07. Some writers in disparate treatment use the phrase "substantial factor," presumably as a way of trying to describe a type of causation that is more restrictive than "motivating factor" causation and less restrictive than "but for" causation. See Price Waterhouse v. Hopkins, 490 U.S. 228, 276 (1989) (O'Connor, J., concurring), superseded by statute, Civil Rights Act of 1991, Pub. L. No. 102-166, 105 Stat. 1071. However, as a matter of causal logic, there is no type of causation between minimal ("motivating factor" causation) and necessity ("but for" causation). Thus, "substantial factor" causation is best understood as being equivalent to minimal-or "motivating factor"-causation. See Katz, supra note 30, at 503-07.
55. See 42 U.S.C. § 2000e-2(m); Price Waterhouse, 490 U.S. at 276 (O'Connor, J., concurring). There is an important difference between the two alternative frameworks in terms of the consequence of proving "motivating factor" causation. Under Price Waterhouse, such a showing only transfers the burden of proof to the defendant; if there is no "but for" causation (i.e., the defendant proves a "same decision" defense), there is no liability. See id. Under the 1991 Act, such a showing triggers liability and at least minimal damages. 42 U.S.C. § 2000e-5(g)(2) (B) (making the "same action" defense only a partial defense). However, the premise for requiring McDonnell Douglas is that it requires more of plaintiffs than either of the alternative frameworks. Thus, for our purposes, it suffices to say simply that the alternative frameworks require "motivating factor" causation-that, upon such a showing, the plaintiff gets some benefit (in Price Waterhouse, a burden shift; in the 1991 Act, liability and a burden shift).
56. There are actually three distinct concepts at work here. First, McDonnell Douglas might be required as a substantive standard if that framework imposed a higher substantive standard than that imposed by the two alternative frameworks. That is, it would make sense to require McDonnell Douglas if it demanded a "but for" standard of causation, while the alternative frameworks demanded only a "motivating factor" standard of causation. Second, McDonnell Douglas might be required as a method of proof if that framework proved a higher standard than that proved by the two alternative frameworks. That is, it would make sense to require McDonnell Douglas if it proved "but for" causation, while the alternative frameworks proved only "motivating factor" causation. Third, McDonnell Douglas might be required as a procedural requirement if it contained a more stringent burden of proof than the two alternative frameworks. That is, it would make sense to require McDonnell Douglas if it placed the burden for proving a particular standard (such as "but for" causation) on the plaintiff, while the alternative frameworks placed that burden on the defendant. The formulation in the text condenses all three of these concepts for the sake of simplicity. I will distinguish these concepts below in Part II.A.
57. I will question this assumption below in Part II.C.1 (arguing that plaintiffs should never be required to bear the burden of proving "but for"). But for purposes of the instant argument, I will assume that, in some cases, plaintiffs might be required to meet this burden.
58. See, e.g., Griffith v. City of Des Moines, 387 F.3d 733, 745 (8th Cir. 2004) (Magnuson, J., concurring specially) (noting that McDonnell Douglas "focuses on the but-for cause of the employment decision"); Stover v. Martinez, 382 F.3d 1064, 1076 (10th Cir. 2004); Rowland v. Am. Gen. Fin., Inc., 340 F.3d 187, 192 n.3 (4th Cir. 2003); Miller v. Cigna Corp., 47 F.3d 586, 597 (3d Cir. 1995) (en banc) (adopting the determinative-factor test, which is equivalent to the "but for" test); Ostrowski v. Atl. Mut. Ins. Cos., 968 F.2d 171, 185 (2d Cir. 1992) (adopting the determinative-factor test); Konowitz v. Schnadig Corp., 965 F.2d 230, 232 (7th Cir. 1992) (adopting the "but for" test); William R. Corbett, An Allegory of the Cave and the Desert Palace, 41 HOUS. L. REV. 1549, 1567 & n.107 (2005) ("It is often stated that the McDonnell Douglas pretext analysis adopted a but-for standard of causation."); Davis, supra note 6, at 895 & n. 197; Kaminshine, supra note 14, at 5, 18; Robert A. Kearney, The High Price of Price Waterhouse : Dealing with Direct Evidence of Discrimination, 5 U. PA. J. LAB. & EMP. L. 303, 310 (2003); Stan Pietrusiak, Labor and Employment Law, 38 TEX. TECH. L. REV. 911, 924 (2006); Zimmer, supra note 11, at 1930 (noting that in McDonnell Douglas cases, courts have typically required plaintiff to prove that the discriminatory motivation was a "but for" cause of the employer's decision). The idea that McDonnell Douglas requires "but for" causation seems to come from three places. First, it may have come from the apparent association in certain Supreme Court cases between "but for" and McDonnell Douglas. See, e.g., Hazen Paper Co. v. Biggins, 507 U.S. 604, 610 (1993) (stating that an ADEA plaintiff must prove that age was a "determinative influence," which likely means "but for" causation, and indicating that the plaintiff might be able to do so using McDonnell Douglas); see also Davis, supra note 6, at 895 (noting that many writers believe that McDonnell Douglas requires "but for" causation based on Hazen Paper). However, just because the Supreme Court might believe that McDonnell Douglas proves "but for" causation does not make it so. Second, the idea that McDonnell Douglas requires "but for" causation may come from the (also flawed) idea that this framework works by a process of elimination. If McDonnell Douglas eliminated all possible nondiscriminatory reasons for the challenged action, it would in fact prove "but for" causation. See infra Part I.C.4. However, contrary to popular belief, McDonnell Douglas does not work this way. See infra Part I.C.2. Third, the idea that McDonnell Douglas requires "but for" causation may come from the (also flawed) observation that McDonnell Douglas presents an either-or dichotomy, in which a factfinder must find either that the employer was motivated by a nondiscriminatory reason or that the employer was motivated by discrimination. It is true that, in such an either-or world, there would always be "but for" causation. See infra Part I.C.4. However, as will be seen below, this is a flawed understanding of McDonnell Douglas, which does not in fact posit such a dichotomous world. See infra Part 111.D.I.
59. McDonnell Douglas Corp. v. Green, 41 1 U.S. 792, 804 (1973).
60. E.g., 42 U.S.C. § 2000e-2(a); Age Discrimination in Employment Act, 29 U.S.C. § 623(a) (2000); Americans with Disabilities Act, 42 U.S.C. § 12112(a) (2000).
61. See Tex. Dep't of Cmty. Affairs v. Burdine, 450 U.S. 248, 255 n.8 (1980) (noting that proof of intentional discrimination can be "elusive").
62. See id.
63. The employer is required to explain its action only after the plaintiff makes out a prima facie case. See McDonnell Douglas, 411 U.S. at 802. However, the prima facie case is designed to set a fairly low threshold, which most plaintiffs are able to clear. See Smith, supra note 15, at 377 ("The 'burden' of showing a prima facie case is commonly acknowledged to be extremely light."); see also Denny Chin & Jodi Golinsky, Moving Beyond McDonnell Douglas : A Simplified Method for Assessing Evidence in Discrimination Cases, 64 BROOK, L. REV. 659, 668 (1998) (noting that the "prima facie case has evolved into something of a formality" and "many courts simply presume that the plaintiff has made out a prima facie case").
64. Students of McDonnell Douglas-and those who have grown accustomed to reading articles and cases with prolonged discussions of the three stages of McDonnell Douglas-will note that I have skipped rather quickly to the third stage of the framework, the "pretext" stage. The point I am making focuses on the third stage. As discussed in the text, it is by proving pretext that the plaintiff actually uses the McDonnell Douglas framework to prove causation. See Mark S. Brodin, The Demise of Circumstantial Proof in Employment Discrimination Litigation: St. Mary's Honor Center v. Hicks, Pretext, and the "Personality" Excuse, 18 BERKELEY J. EMP. & LAB. L. 183, 191 (1997) (noting that "[t]he crux of an individual disparate treatment lawsuit is... the pretext stage"); Hart, supra note 8, at 753 (remarking that "the third stage of the McDonnell-Douglas framework is where most of the action. . . seems to be"); Sam Stonefield, Non-Determinative Discrimination, Mixed Motives, and the Inner Boundary of Discrimination Law, 35 BUFF. L. REV. 85, 111 & n.89 (1986) ("The pretext issue is where the action is, where most disparate treatment cases are won or lost."). In fact, I argue that that primary purpose-perhaps the only purpose-of the prima facie case (the first stage) is to trigger the defendant's obligation to proffer a reason for the challenged action (the second stage), and that the only reason for this is to provide the plaintiff with a target to attack in order to prove causation. See infra note 91.
65. See Reeves v. Sanderson Plumbing Prods., Inc., 530 U.S. 133, 147 (2000) ("'[R]ejection of the defendant's proffered reasons will permit the trier of fact to infer the ultimate fact of intentional discrimination.'" (emphasis omitted) (quoting St. Mary's Honor Ctr. v. Hicks, 509 U.S. 502, 511 (1993)); Burdine, 450 U.S. at 256 (stating that a plaintiff may satisfy her "ultimate burden of persuading the court that she has been a victim of intentional discrimination... by showing that an employer's proffered explanation is unworthy of credence"); see also Kenneth R. Davis, Age Discrimination and Disparate Impact: A New Look at an Age-Old Problem, 70 BROOK. L. REV. 361, 370 n.64 (2004) ("[A] plaintiff [may] prove discrimination inferentially by proving that the defendant's alleged justification for the challenged action was a pretext for discrimination."); Michael Selmi, Subtle Discrimination: A Matter of Perspective Rather Than Intent, 34 COLUM. HUM. RTS. L. REV. 657, 669 (2003) (noting that the Court in Hicks held that "proof of pretext leads to a permissive. . . inference of discrimination").
66. A few courts and commentators occasionally seem to grasp the fact that McDonnell Douglas relies on a chain of inferences. See, e.g., William J. Vollmer, Note, Pretext in Employment Discrimination Litigation: Mandatory Instructions for Permissible Inferences?
67. WASH. & LEE L. REV. 407, 410 (2004) (referring to a "chain of inferences"); see also Reeves, 530 U.S. at 147 ("In appropriate circumstances, the trier of fact can reasonably infer from the falsity of the explanation that the employer is dissembling to cover up a discriminator)' purpose."). However, no court or commentator has analyzed each step in the chain, as necessary to truly understand McDonnell Douglas. (And almost all go on-mistakenly-to talk about McDonnell Douglas as working by a process of elimination, as opposed to by a chain of inferences. See infra Part I.C.2, especially note 81.)
68. The fact that a mistake might be an honest one-and not a cover-up for discrimination-demonstrates that good faith should be a defense in a pretext case. Most courts have so held. See, e.g., Forrester v. Rauland-Borg Corp., 453 F.3d 416, 419 (7th Cir. 2006) ("An honest mistake, however dumb, is not [a pretext]."); Exum v. U.S. Olympic Comm., 389 F.3d 1130, 1137-38 (10th Cir. 2004) ("To show pretext, the plaintiff must call into question the honesty or good faith of the [employer's] assessment of his abilities."); cf. Rivas Rosado v. Radio Shack, Inc., 312 F.3d 532, 535 (1st Cir. 2002) ("Title VII . . . does not ensure against inaccuracy by an employer, only against gender-based discrimination."). But see ROBERT BELTON ET AL., EMPLOYMENT DISCRIMINATION LAW 103-04 (7th ed. 2004) (discussing divergent positions on good faith defense).
69. This conclusion might be based on the fact that the employer's reason was mistaken (which, as noted above, might support such an inference), or from the fact of mistake in addition to other facts which might suggest that the employer knew the claim of theft was wrong.
70. Saying the employee was a thief is unlikely to support an inference that the employer was trying to protect the employee's feelings. But other stated reasons which turn out to be incorrect, such as a statement that employee's job functions were "no longer required," might support an inference of a benign lie. See, e.g., Rhodes v. Guiberson Oil Tools, 39 F.3d 537, 545-46 (5th Cir. 1994) (Zagel, J., concurring specially) (observing that employers, when firing workers, often give polite, dishonest explanations to "soften the blow"), rev'd en banc, 75 F.3d 989 (5th Cir. 1996); Sigal Constr. Co. v. Stanbury, 586 A.2d 1204, 1206 (D.C. 1991) (finding that an employer who fired her employee for poor performance told him that he was being laid off in order to protect his feelings); see also McCormick, supra note 15, at 179 n.76 (noting strategic but nondiscriminatory reasons why an employer might provide a false reason for its action).
71. Some writers have argued that such a conclusion (that a lie was for benign reasons) is unlikely. See, e.g., Melissa A. Essary, The Dismantling of McDonnell Douglas v. Green : The High Court Muddies the Evidentiary Waters in Circumstantial Discrimination Cases, 21 PEPP. L. REV. 385, 441 (1994) ("[I]f a plaintiff proves that an employer's proffered reason for its actions is false, most jurors will logically infer that the proffered reason is a cover-up for discrimination."); Malamud, supra note 6, at 2243 n.49 ("Rarely can it be envisioned that a jury, as factfinder, will hold for a lying defendantemployer, except in that exceptional case where it is established at trial that the pretextual reason is a cover-up for say an embarrassing one, rather than a discriminatory one." (citation and internal quotation marks omitted)). The point is not which conclusion the factfinder would be most likely to draw. The point is merely that a factfinder might logically make either inference.
72. There are, of course, other types of evidence that might let a factfinder infer that the defendant is engaged in a cover-up. For example, if the employer proffers a reason for its action that is so nonsensical that no reasonable employer would utilize such a criteria in decisionmaking (such as claiming to have fired the plaintiff for being a stamp collector or because of the plaintiffs zodiac sign), the factfinder might conclude that the proffered reason, even if true, was a cover-up. Or if the employer does not apply its rule uniformly (such as firing only some of the employees who were caught stealing), the factfinder may conclude that the proffered reason, even if true, was a cover-up. However, these two forms of evidence would not be pretext evidence in the true sense of the word, as they do not involve inferences of dishonesty based on the falsity of the proffered reason. See infra note 76 (discussing the proper use of the term "pretext").
73. This type of firing would likely be actionable under most states' doctrine of wrongful discharge in violation of public policy. See, e.g., Martin Marietta Corp. v. Lorenz., 823 P.2d 100, 108 (Colo. 1992) (holding that claims for wrongful discharge in violation of public policy are cognizable in Colorado and cover situations where an employee is terminated for refusing to follow a superior's order to make a false statement to a federal agency); Sheets v. Teddy's Frosted Foods, Inc., 427 A.2d 385, 388-89 (Conn. 1980) (holding that an employee who was fired for insisting his employer comply with a state statute had a cause of action in tort for wrongful discharge); see also Pierce v. Ortho Pharm. Corp., 417 A.2d 505, 512 (N.J. 1980) (stating that "[a]n employer's fight to discharge an employee at will carries a correlative duty not to discharge an employee who declines to perform an act that would require a violation of a clear mandate of public policy"); Brockmeyer v. Dun & Bradstreet, 335 N.W.2d 834, 840 (Wis. 1983) (holding that "[a] wrongful discharge is actionable when the termination clearly contravenes the public welfare and gravely violates paramount requirements of public interest").
74. See, e.g., Hazen Paper Co. v. Biggins, 507 U.S. 604, 612 (1993) (holding that a decision to fire the plaintiff to prevent his pension from vesting, while reprehensible, and possibly illegal under other laws, was not age discrimination).
75. There is some debate over whether instructions on pretext should be mandatory1 or whether such instructions should be left to trial judges' discretion. See Vollmer, supra note 66, at 410-13. Given the confusion that seems to surround this chain of inferences, an instruction on pretext that mirrors the discussion in this subpart would likely be helpful. Note that this issue-whether juries should be given pretext instructions-is distinct from the issue of whether juries should be instructed about the three stages of McDonnell Douglas. Most courts have held that juries should never be instructed about the three stages of McDonnell Douglas. See William R. Corbett, Of Babies, Bathwater, and Throwing Out Proof Structures: It Is Not Time to Jettison McDonnell Douglas, 2 EMP. RTS. & EMP. POL'Y J. 361, 381 & n.91 (1998) (noting that courts are divided on whether to instruct juries on McDonnell Douglas burden-shifting); Sandra F. Sperino, Recreating Diversity in Employment Law by Debunking the Myth of the McDonnell Douglas Monolith, 44 Hous. L. REV. 349, 376-77 & nn.132-33 (2007) (same).
76. It might be argued that McDonnell Douglas' chain of inferences runs afoul of the rule against piling "inference upon inference" in order to prove a fact. See, e.g., Interlake Iron Corp. v. NLRB, 131 F.2d 129, 133 (7th Cir. 1942). However, this rule does not preclude the three sequential inferences in the McDonnell Douglas chain. No one appears to have criticized McDonnell Douglas on these grounds. And most modern courts and writers accept that inference may be piled upon inference if all of the inferences in the chain are sound. See, e.g., NLRB v. Cameo, Inc., 340 F.2d 803, 811 (5th Cir. 1965) ("The so-called rule against pyramiding inferences, if there really is such a 'rule' and if it is anything more than an empty pejorative, is simply legalese fustian to cover a clumsy exclusion of evidence having little or no probative value." (citing 1 JOHN HENRY WIGMORE, EVIDENCE IN TRIALS AT COMMON LAW § 41, at 434-36 (3d ed. 1940))). In most modern cases that purport to apply the "inference upon inference" prohibition, there was no clear evidence that the defendant's proffered reason was wrong-i.e., no evidence of pretext. See, e.g., Cline v. BWXT-Y12, L.L.C., No. 304-CV-588, 2007 WL 1227482, at *8 (E.D. Tenn. Apr. 24, 2007); Schmidt v. Chao, No. 04-892(RMC), 2006 WL 1663389, at *5 (D.D.C. June 13, 2006); Anderson v. Nat'1 R.R. Passenger Corp., No. 03C7589, 2006 WL 931699, at *9, *12 (N.B. 111. Apr. 6, 2006); see also infra note 76 (discussing the meaning of pretext). No court appears to have dismissed a claim under the "inference upon inference" rule where the plaintiff has offered evidence that the proffered reason is wrong.
77. Note that this use of the word "pretext" differs from three common ways in which this word tends to be used. First, several writers use the word to describe one of the links in the chain of inferences, as opposed to the entire chain. For example, some writers define pretext as a lie; others define it as a cover-up. See McDonnell Douglas Corp. v. Green, 411 U.S. 792, 805 (1973) ("coverup"); Turnery. Tex. Instruments, Inc., 555 F.2d 1251, 1255 n.3 (5th Cir. 1977) (describing discrimination as "the use, by employers, of legitimate reasons for action to hide racial animus in decision making"), overruled by Burdine v. Tex. Dep't of Cmty. Affairs, 647 F.2d 513 (5th Cir. 1981); McCormick, supra note 15, at 177 n.69 ("The courts usually define pretext as a lie."). In terms of describing a method of proof, these writers' definitions are flawed. As we have seen, none of the links in the chain of inferences, standing by itself, proves discrimination (except for the final link, discrimination-but defining pretext as discrimination would result in a tautology: the plaintiff could prove discrimination by proving discrimination). See supra Part I.C.I. It is the entire chain that proves discrimination. Thus, "pretext" as a method of proof is best understood as referring to the entire chain. Second, some writers use the word "pretext" to include a reason that, while true, was not the employer's "real" or "actual" reason for its action. See Kaminshine, supra note 14, at 42; Zimmer, supra note 11, at 1925. This usage does not make sense in the context of McDonnell Douglas. While such a post hoc search for a true-but-not-"real" justification might be a "pretext" in the colloquial sense of the word, it would not be the type of pretext which would permit a plaintiff to prove discrimination using the McDonnell Douglas pretext method. This is because if the employer's stated reason is true (irrespective of whether it was the employer's "real" reason)-that is, if there is no error in the stated reason-then the factfmder cannot find a lie based on the error, and therefore cannot proceed down the McDonnell Douglas chain of inferences. This is not to say that the plaintiff might not still be able to prove that the employer's stated reason is not its real reason and that its real reason is discriminatory. It is only to say that the plaintiff will not be able to make this showing by proving that the employer's stated reason is false-that is, by the pretext method of proof. See infra Part III.D.1. Finally, several writers-and occasionally the Court-have used the word "pretext" to refer to the entire third stage of McDonnell Douglas. See, e.g., Tex. Dep't of Cmty. Affairs v. Burdine, 450 U.S. 248, 256 (1981); Chambers, supra note 8, at 85; Catherine ). Lanctot, The Defendant Lies and the Plaintiff Loses: The Fallacy of the "Pretext-Plus" Rule in Employment Discrimination Cases, 43 HASTINGS L.J. 57, 65-66 (1991); Malamud, supra note 6, at 2234. This is actually a misnomer. At the third stage, there are two ways that the plaintiff can proceed, only one of which involves "pretext." At that stage, the plaintiff can proceed either by (1) proving that the defendant's proffered reason is "unworthy of credence" (the pretext method), or by (2) showing that discrimination was the "real" reason (presumably using some method of proof other than the "pretext" method). See Burdine, 450 U.S. at 256; see also Green, supra note 4, at 984-85, 988 (noting that there are two ways to proceed at the third stage).
78. 530 U.S. 133 (2000).
79. See Id. at 147 (internal quotation marks omitted) (citing Wilson v. United States, 162 U.S. 613, 620-21 (1896); 2 JOHN HENRY WIGMORE, EVIDENCE IN TRIALS AT COMMON LAW § 278(2), at 133 (J. Chadbourn ed., 1979)).
80. See Binder v. Long Island Lighting Co., 57 F.3d 193, 200 (2d Cir. 1995) ("Resort to a pretextual explanation is, like flight from the scene of a crime, evidence indicating consciousness of guilt, which is, of course, evidence of illegal conduct."); Brodin, supra note 64, at 202 ("The concept of pretext. . . reflects our experience with evasive explanations offered by those accused of wrongdoing, such as the frequent use of 'consciousness of guilt" evidence in criminal cases and 'consciousness of liability' evidence in civil cases."); see also Fisher v. Vassar Coll., 114 F.3d 1332, 1390 (2d Cir. 1997) (Winter, J., dissenting) (referencing the "vast body of law allowing inferences of consciousness of guilt to be drawn from dishonest behavior concerning facts material to litigation" (citing United States v. Sureff, 15 F.3d 225, 227 (2d Cir. 1994); 1 EDWARD J. DEVITT ET AL., FEDERAL JURY PRACTICE AND INSTRUCTIONS § 14.06 (4th ed. 1987); 1 LEONARD B. SAND ET AL., MODERN FEDERAL JURY INSTRUCTIONS § 6.05 (Instruction 6-11) (1996))). Other similar examples of the general principle include factfinders' ability to infer guilt from: use of a false name, fabrication of an alibi, use of disguised handwriting, falsification of evidence, intimidation of witnesses, and engaging in clandestine behavior. See id. McDonnell Douglas' pretext analysis also finds support in a slightly different principle regarding veracity: where a party has lied about one issue (such as why the plaintiff was fired), a factfinder may (but need not) choose to disbelieve that party's testimony on other issues (such as whether he discriminated against the plaintiff). See id.
81. This understanding of McDonnell Douglas as a chain of permissive inferences should serve to resolve two longstanding debates in disparate treatment law. First, this understanding resolves the so-called "pretext-plus" debate. This debate involves the effect of proving pretext. Once a plaintiff has proven pretext, has she created a jury question on the issue of causation? Or do plaintiffs also need to produce some evidence in addition to evidence of pretext-"plus" evidence-to create a jury question? The Court has now held that, most of the time, pretext evidence is sufficient; but sometimes additional evidence might be required. See Reeves, 530 U.S. at 147-48. Because Reeves did not offer much guidance on when additional "plus" evidence is required, lower courts have (predictably) split on the issue. See Steven H. Adelman et al., Summary Judgment Standards Following Reeves v. Sanderson Plumbing Products and Its Progeny, in EMPLOYMENT DISCRIMINATION AND CIVIL RIGHTS ACTIONS IN FEDERAL AND STATE COURTS, 301, 317 (ALI-ABA Course of Study Materials, Sept. 15-17,2005) available at WLSL021 ALI-ABA 301 (noting post-Reeves split on "pretext plus" issue); Catherine J. Lanctot, Secrets and Lies: The Need for a Definitive Rule of Law in Pretext Cases, 61 LA. L. REV. 539, 547-48 (2001) (same); Audrey J. Lee, Note, Unconscious Bias Theory in Employment Discrimination Litigation, 40 HARV. C.R.-C.L. L. REV. 481, 496 n.91 (2005) (same). From the analysis in the text, we can see that, as a general matter, proof that the defendant's proffered reason was wrong is sufficient to permit an inference of discrimination-and thus to survive summary judgment. From the evidence of error, the factfinder can infer a lie; from a lie, a cover-up; and from a cover-up, discrimination. There might be cases in which a defendant offers evidence that might tend to push the factfinder off this chain of inferences. For example, the defendant might offer evidence that its error was a good faith mistake. However, in most cases, this would simply present a factual question for the jury: was the error a good faith mistake or a lie? A plaintiff could offer additional evidence that the defendant lied. But the plaintiff would not need to do so to create a question of fact on the issue. Of course, if the defendant offered evidence of good faith mistake (or a lie for a benign reason, or a cover-up for a nondiscriminatory reason) that no reasonable factfinder could reject, then McDonnell Douglas would not permit an inference of discrimination. In such a case, to survive summary judgment, the plaintiff would need to offer evidence of discrimination other than McDonnell Douglas evidence. While this could be called "plus" evidence, that would be a misnomer. In such a case, this non-McDonnell Douglas evidence would be the plaintiffs only evidence of discrimination. The second debate that should be cleared up by understanding McDonnell Douglas as a chain of permissive inferences is the ongoing academic debate over whether St. Mary's Honor Center v. Hicks was correctly decided. In Hicks, the Court held that proof of pretext does not compel a verdict for the plaintiff. See St. Mary's Honor Ctr. v. Hicks, 509 U.S. 502, 511 (1993). This holding gave rise to a firestorm of criticism, with one of the primary complaints being that a finding of "pretext" logically required a finding of discrimination. See, e.g., Krieger, supra note 15, at 1209-24; Michael Selmi, Proving Intentional Discrimination: The Reality of Supreme Court Rhetoric, 86 GEO. L.J. 279, 329-34 (1997). See generally Corbett, supra note 6, at 342-58 (discussing the burdens on the plaintiff). The flaw in this reasoning is that, depending on what one means by "pretext," a finding of pretext does not necessarily require a finding of discrimination. If, by "pretext," the critics of Hicks mean only one of the earlier links in the chain (e.g., error, a lie, or a cover-up), such a finding does not logically require a finding of discrimination. Each successive link is permissive. A reasonable factfinder can, for example, find error, or even a lie or cover-up, and still not find discrimination. If, on the other hand, the critics define "pretext" as the last link in the chain of inferences (discrimination), then their argument is merely a tautology: where a plaintiff proves discrimination, a factfinder must find discrimination. In other words, despite the firestorm of criticism that has been leveled against it, Hicks was correctly decided-at least as a doctrinal matter. For other, normative criticisms of Hicks, see, e.g., Derum & Engle, supra note 8, at 1224-25 (discussing Hicks as a "personal animosity" case); Donna G. Coldian, New Reason to Lie
82. See, e.g., Kaminshine, supra note 14, at 10-11 & n.45; Lanctot, supra note 76, at 117 (noting that if the factfinder disbelieves a proffered reason, the only remaining reason is discrimination); McCormick, supra note 15, at 162; Sullivan, supra note 21, at 934-35 ("Getting to a single discriminatory motive by process of elimination is the core of McDonnett Douglas ...."); Zimmer, supra note 11, at 1933; Roberts. Whitman, Note, Clearing the Mixed-Motive Smokescreen: An Approach to Disparate Treatment Under Title VII, 87 MICH. L. REV. 863, 884 (1989) ("The three-stage inferential inquiry is a process of elimination ...."). In this, these writers have been aided in their confusion by the Court, which has at times suggested that McDonnell Douglas works by process of elimination. See, e.g., Reeves, 530 U.S. at 147 ("[O]nce the employer's justification has been eliminated, discrimination may well be the most likely alternative explanation . . . ."); Furnco Constr. Corp. v. Waters, 438 U.S. 567, 577 (1978) ("[W] hen all legitimate reasons for rejecting an applicant have been eliminated as possible reasons for the employer's actions, it is more likely than not the employer, who we generally assume acts only with some reason, based his decision on an impermissible consideration such as race." (emphasis omitted)).
83. This description-in which discrimination is the only possible remaining reason-is the strong version of the process-of-elimination argument. This strong version of the process-of-elimination argument is the only version relevant to the issue of "but for" causation. There is also a weaker version of the process-of-elimination argument, in which many-but not all-nondiscriminatory reasons are eliminated. Under the weak version, discrimination is not the only possible remaining reason; but depending on one's assumptions it might be the most likely remaining reason. This weak version of the process-of-elimination argument will be addressed below in Part III.C.
84. See supra note 81.
85. See infra Part I.C.3.
86. See McDonnell Douglas Corp. v. Green, 411 U.S. 792, 802 (1973) (noting that in a failure-to-hire case, a prima facie case includes proof that plaintiff was objectively qualified for the job and that the job remained open).
87. See id. at 804-05.
88. See supra Part I.C.I.
89. See infra Part I.C.3.
90. I make no claim about the likelihood of such a scenario. It may be that factfinders are inclined completely to believe or completely to disbelieve a defendant. Thus, if the factfinder found one proffered reason (performance) to be false, it might well be inclined to find the other proffered reason (tardiness) to be false. Or if the factfinder found one proffered reason (tardiness) to be true, it might be inclined to find the other proffered reason (performance) to be true-or, at the very least, be disinclined to find any error in this reason to be a lie or a cover-up. My point, rather, is that it is possible that the factfinder might believe one proffered reason and disbelieve another; the fact that a plaintiff proves one proffered reason to be a pretext for discrimination does not necessarily eliminate all potential nondiscriminatory reasons for the challenged action.
91. Below, I will address the possibility of a strong version of McDonnell Douglas, in which the plaintiff disproves not just one proffered reason, but all proffered reasons. See infra Part I.C.4. This strong version would, in fact, eliminate all potential nondiscriminatory reasons. However, even here, it is not the elimination of all nondiscriminatory reasons that gives rise to the inference of discrimination. Rather, it is the fact that the factfinder concludes that at least one reason was erroneous, and that the error was a lie and a cover-up for discrimination. A factfinder could, for example, conclude that all of the defendant's multiple proffered reasons were wrong and still conclude that all of those errors were good faith mistakes. Moreover, as I will discuss below, current law could never require such a strong version of McDonnell Douglas. See infra Part I.C.4.
92. The fact that McDonnell Douglas does not work through a process of elimination casts the first stage of the framework-the plaintiffs prima facie case-in a new, and diminished, light. The prima facie case eliminates some common reasons for adverse employment decisions, such as the plaintiffs lack of objective qualifications or the employer's lack of need to fill the position. See Tex. Dep't of Cmty. Affairs v. Burdine, 450 U.S. 248, 253-54 (1981); see also McDonnell Douglas Corp. v. Green, 411 U.S. 792, 802 (1973) (noting that a Title VII plaintiff may establish a prima facie case of racial discrimination in hiring "by showing (i) that he belongs to a racial minority; (ii) that he applied and was qualified for a job for which the employer was seeking applicants; (iii) that, despite his qualifications, he was rejected; and (iv) that, after his rejection, the position remained open"). It is therefore not surprising that many writers have seen this stage as a critical first step in a process of elimination. See, e.g., Sheila R. Foster, Causation in Antidiscrimination Law: Beyond Intent Versus Impact, 41 HOUS. L. REV. 1469, 1504-05 (2005); Michael J. Zimmer, Leading by Example: An Holistic Approach to Individual Disparate Treatment Law, 11 KAN. J.L. & PUB. POL'Y 177, 177 (2001). But once we understand that McDonnell Douglas does not work by a process of elimination, the significance of the prima facie case is severely diminished. Eliminating a few common potential reasons for the challenged action does little, if anything, to prove discrimination. If one of the common reasons was a reason that the defendant professed to rely upon, then the defendant would presumably proffer that reason and the factfinder would need to determine if that reason was a true reason. If the common reason was not a reason that the defendant professed to rely upon, it would be irrelevant to the pretext analysis. So eliminating common reasons is only relevant to the pretext method if one of those reasons also happens to be one of the defendant's proffered reasons. See Michael J. Hayes, That Pernicious Pop-Up, the Prima Facie Case, 39 SUFFOLK U. L. REV. 343, 375-76 (2006) (noting that once the employer identifies the motivating reasons, " speculative" reasons become unnecessary). And in such a case, the validity of that reason would be tested at the pretext stage. The prima facie case would merely serve as a needless complication. See Malamud, supra note 6, at 2243-45 (questioning the evidentiary value of the prima facie case). If the prima facie case has any value at all, it is as a trigger for forcing an at-will employer to proffer a reason for its action (the second stage of McDonnell Douglas). See Henry L. Chambers, Jr., Discrimination, Plain and Simple, 36 TULSA L.J. 557, 561 (2001) ("The mandatory presumption of discrimination that accompanies proof of a prima facie case is merely a vehicle to coax an LNR [legitimate nondiscriminatory reason] from the employer, and meant nothing once the employer articulated the LNR."); see also Burdine, 450 U.S. at 255-56 (explaining that the point of the second stage of the framework-and thus that of the prima facie case that triggers the second stage-is to present a reason for the challenged action and frame the issue in terms of the veracity of the reason). Absent such compulsion, an at-will employer might simply stand mute as to the reasons for its actions. In such a case, the plaintiff would not be able to use the pretext method. If the plaintiff had no other proof of discrimination (such as statements or comparative evidence), the court might dismiss the claim without making the defendant provide a reason-and without discovery-thus short-circuiting the plaintiffs ability to use the pretext method. So the issue-and the only issue-addressed by the prima facie case is simply when a court should compel an atwill employer to proffer a reason for its actions, which might enable the plaintiff to use the pretext method against an otherwise silent defendant. Whether it would be proper to dismiss a claim prior to discovery, and thus prior to any opportunity to ascertain the defendant's reason absent McDonnell Douglas, probably depends on one's view of pleading requirements. Such an assessment is beyond the scope of this Article. Cf. Swierkiewicz v. Sorema N.A., 534 U.S. 506, 515 (2002) (holding that notice pleading applies in antidiscrimination cases); RUTHERGLEN, supra note 3, at 36 ("Legal doctrine does not require employers to offer good reasons for their decisions, but the practicalities of litigation often compel them to do so."). A corollary of the insignificance of the prima facie case is that courts should never dismiss cases for failure to state a prima facie case where the defendant nonetheless proffers a reason for its action and the plaintiff has evidence tending to show that the proffered reason is pretext.
93. See supra note 58.
94. See supra notes 71-75 and accompanying text.
95. See 42 U.S.C. § 2000e-2(m) (2000) (indicating that liability attaches upon a "motivating factor" showing).
96. See Price Waterhouse v. Hopkins, 490 U.S. 228, 244-45 (1989) (plurality opinion), superseded by statute, Civil Rights Act of 1991, Pub. L. No. 102-166, 105 Stat. 1074; see also supra note 30 (noting that the "substantial factor" test for burden-shifting in Justice O'Connor's concurrence most likely refers to "motivating factor" causation).
97. See supra notes 51-55 and accompanying text.
98. As I will demonstrate below, the same can be said for almost all other methods of proof. See infra notes 183-91 and accompanying text.
99. See supra Part I.C.2.
100. See supra Part I.C.2.
101. Professor Davis tries to argue that McDonnell Douglas does not require "but for" causation on very different grounds. He attempts to argue that (1) many of the writers who believe that McDonnell Douglas requires "but for" causation base their reasoning on the language of Hazen Paper v. Biggins, 507 U.S. 604, 610 (1992), which states that plaintiffs must show "determinative factor" causation, and (2) the Justices who wrote Hazen Paper, despite using the phrase "determinative factor," did not really intend to require "but for" causation. See Davis, supra note 6, at 895-900. Though Professor Davis reaches the correct result (that McDonnell Douglas does not require "but for" causation), his argument is not completely persuasive. First, whatever the Justices who wrote Hazen Paper may have intended by the phrase "determinative factor," this phrase means "but for." See Katz, supra note 30, at 501-03 (noting that "determinative factor" means "but for" causation); Stonefield, supra note 64, at 115-16 (noting that "but for" causation in tort law takes the form of a "determinative factor" requirement in discrimination cases). Second, Hazen Paper did not need to address-and thus should not be seen as addressing-causal standards. The case did not involve claims of multiple independent reasons, such as tardiness and poor performance on a particular project; rather, the plaintiffs claim was that the single reason asserted by the defendant (time until retirement) was in fact dependent on-i.e., influenced by-age. See Hazen Paper, 507 U.S. at 611-12. Thus, there was no need, at least at that stage of the case, to decide the appropriate standard of causation. Finally, many of those who believe that McDonnell Douglas requires "but for" causation believe this independently of Hazen Paper. For example, many writers believe that McDonnell Douglas requires "but for" causation because they believe that it is based on a process of elimination or because they believe that it presents an either-or paradigm. See supra note 81; infra note 248. Although both of these beliefs turn out to be incorrect, see supra Part I.C.2; infra Part III.D.1, trying to recast Hazen Paper is not a productive way to refute writers whose beliefs are not based on that case.
102. See supra Part I.C.1.
103. The set of proffered reasons can easily be limited by the plaintiffs counsel asking that all-important question: "Is there any other reason for your decision?"
104. Dean Kaininshine asserts that McDonnell Douglas does prove "but for" causation. Kaminshine, supra note 14, at 18. His analysis assumes this strong version of McDonnell Douglas. See id. However, as will be discussed below, no court could require this strong version under current law.
105. Actually, there could be an independently sufficient set of factors. For simplicity, I will assume only a single independently sufficient factor. The analysis is the same.
106. As a practical matter, it seems unlikely that a factfinder would find that (1) the first proffered reason was a cover-up for discrimination, and (2) an additional proffered reason was erroneous, and then go on to find that this error was not a cover-up for discrimination (e.g., that the defendant made a good faith error regarding the second proffered reason). However, it is possible. And this possibility introduces the possibility of a second, independent factor-and thus the possibility of less-than-"but for" causation. This is why the strong version of McDonnell Douglas requires that each proffered reason be shown to be a pretext for discrimination, rather than merely wrong.
107. Note that this strong version of McDonnell Douglas uses an elimination technique to prove "but for" causation. That is, it proves "but for" causation by eliminating other potential reasons besides discrimination for the defendant's action. However, the elimination does not prove that discrimination was a reason. As noted above, this is proven by the chain of permissive inferences flowing from proof that one of the employer's proffered reasons was wrong. See supra Part I.C.2. The elimination only proves the lack of an independently sufficient reason-and thus "but for" causation once discrimination is established.
108. See supra Part I.B.
109. See supra notes 51 and 53.
110. See Hawkins v. Dir., Office of Workers' Comp. Programs, 907 F.2d 697, 704 (7th Cir. 1990) (noting that "necessary and . . . sufficient" is equivalent to "sole cause"). The logical explanation for this is as follows: If a factor is sufficient, it is capable of triggering the outcome (here, the adverse employment decision). If a factor is necessary, that means that there are no other factors that would be sufficient to cause the outcome. See Katz, supra note 30, at 512-14. So if a factor is necessary and sufficient, it will be the only factor capable of triggering the outcome-the sole cause.
111. See 110 CONG. REC. 13,837, 13,837-38 (1964); see also Price Waterhouse v. Hopkins, 490 U.S. 228, 241 n.7 (1989) (plurality opinion) ("Congress specifically rejected an amendment that would have placed the word 'solely' in front of the words 'because of.' " (citing 110 CONG. REC. 13,837) ), superseded by statute, Civil Rights Act of 1991, Pub. L. No. 102-166, 105 Stat. 1071; Griffith v. City of Des Moines, 387 F.3d 733, 740 & 11.4 (8th Cir. 2004) (Magnuson, J., concurring specially) (citing 110 CONG. REC. 13,837-38) (noting the rejection by both the House and Senate of a proposal to add the word "solely" to the statute); see also Cheryl L. Anderson, What is "Because of the Disability " Under the Americans with Disabilities Act? Reasonable Accommodation, Causation, and the Windfall Doctrine, 27 BERKELEY J. EMP. & LAB. L. 323, 340-41 (2006) (noting that Congress rejected the "sole cause" requirement in the ADA).
112. This is not to say that a plaintiff could not choose to use the strong version of McDonnell Douglas-that is, to try to prove that all of the reasons proffered by the defendant were pretextual. If the plaintiff does so successfully, she would prove "but for" causation (actually, "sole" causation). Such a showing would preclude the defendant from proving lack of "but for" causation through a "same decision"/"same action" defense. See infra Part II.A.
113. See supra Part I.C.3.
114. See 42 U.S.C. § 2000e-2(m) (2000).
115. See supra note 56.
116. Many courts and commentators have recognized that McDonnell Douglas is a method of proof. See, e.g., EEOC v. PVNF, L.L.C., 487 F.3d 790, 800-02 (10th Cir. 2007); Brewer v. Bd. of Trs. of Univ. of Ill., 479 F.3d 908, 915 (7th Cir. 2007); Kaminshine, supra note 14, at 32; Sullivan, supra note 21, at 931. The problem is that these courts and commentators have generally seen it as a method of proving "but for" causation. See supra note 58 and accompanying text. From there, it has been too easy to see it as requiring "but for" causation.
117. See supra note 58 and accompanying text.
118. See supra Part I.C.3.
119. It is possible, I suppose, to use McDonnell Douglas to designate a "motivating factor" standard of causation or an allocation of the burden of proof to the plaintiff to prove "motivating factor" causation. However, even this use of McDonnell Douglas would not make much sense. Using a method of proof to denote the standard that it proves is unduly narrow. While McDonnell Douglas certainly proves "motivating factor" causation, there are several other ways to prove the same standard. See infra Part III.E. So if we want to say that plaintiffs must prove "motivating factor" causation, it is unduly restrictive to say that they must use McDonnell Douglas. Moreover, using McDonnell Douglas as a way to require plaintiffs to prove "motivating factor" causation would fail to distinguish it from Price Waterhouse or the 1991 Act. It is also possible-even likely-that courts and commentators have intended McDonnell Douglas to denote a requirement that plaintiffs prove "but for" causation. See, e.g., Price Waterhouse v. Hopkins, 490 U.S. 228, 278-79 (1989) (O'Connor, J., concurring), superseded by statute, Civil Rights Act of 1991, Pub. L. No. 102-166, 105 Stat. 1071; see also infra Part II.B.2.C (discussing Justice O'Connor's likely intent in Price Waterhouse). The analysis above shows that it does not make sense to use McDonnell Douglas in this way. This is not to say that the law might not still require certain plaintiffs to prove "but for" causation. It is merely to say that the law cannot do so by requiring those plaintiffs to use McDonnell Douglas. (Below, I will argue that the law should not and does not require any plaintiffs to prove "but for" causation. See infra Part II.C.1, particularly text accompanying notes 182-91.)
120. See Katz, supra note 30, at 501-11 (explaining these frameworks in causal terms). The difference between the two alternative frameworks is in the effect of satisfying these burdens. Under Price Waterhouse, liability attaches only at the "but for" level. Id. at 528. Under the 1991 Act, liability attaches at the "motivating factor" level, while full damages attach only at the "but for" level. Id. at 530.
121. Arguably, Price Waterhouse does specify how causation must be proven-at least for some plaintiffs. Justice O'Connor's concurrence states that plaintiffs who do not have "direct evidence" must use McDonnell Douglas. See Price Waterhouse, 490 U.S. at 278-79 (O'Connor, J., concurring). However, this statement should be understood as being about burdens of proof, not methods of proof. Justice O'Connor almost certainly wanted to ensure that plaintiffs without "direct evidence" would bear the full burden of proving "but for" causation. See id. (focusing on issue of when a plaintiff can transfer the burden of proof). She believed that McDonnell Douglas did this. But, as we have seen above, McDonnell Douglas does not prove "but for" causation. See supra Part I.C.3. It seems highly unlikely that Justice O'Connor wanted to make plaintiffs use a method of proof which only proved "motivating factor" causation. A better understanding is that she wanted to make plaintiffs prove "but for" and did not care what method of proof they used to do so. See also infra Part II.B.2.C (discussing Justice O'Connor's likely intent in Price Waterhouse).
122. See Curley v. St. John's Univ., 19 F. Supp. 2d 181, 188 (S.D.N.Y. 1998) (suggesting that a plaintiff might use the "pretext" analysis to show that discrimination was a "motivating factor"); Davis, supra note 15, at 752 ("Discrediting defendant's articulated legitimate reason is indirect proof that discriminatory intent motivated the defendant. Proof of pretext is, therefore, relevant to and may satisfy the 'motivating factor' test of the 1991 Act."); see also Fogg v. Gonzales, 492 F.3d 447, 451 n.* (D.C. Cir. 2007) ("A plaintiff may also, of course, use evidence of pretext and the McDonnell Douglas framework to prove a mixed-motive [i.e., "motivating factor"] case."). But see Davis, supra note 15, at 752 (suggesting-erroneously-that pretext does not necessarily show motivating factor causation).
123. See 42 U.S.C. § 2000e-2(m) (2000); Price Waterhouse, 490 U.S. at 261-79; see also Katz, supra note 30, at 501-11 (explaining these frameworks in causal terms).
124. For example, Dean Kaminshine states that the "same action" defense is not appropriate when a plaintiff uses McDonnell Douglas because he believes that McDonnell Douglas proves "but for" causation, which would make it impossible for a defendant to prove "same action" (a lack of "but for" causation). See Kaminshine, supra note 14, at 29; see also Fogg, 492 F.3d at 454 (concluding that the "same action" defense is inapplicable to McDonnell Douglas); RUTHERGLEN, supra note 3, at 45-47 (suggesting that a plaintiff who proves pretext need not worry about a "same action" defense). This is true if the plaintiff uses the strong version of McDonnell Douglas (which proves "sole factor" causation, and thus "but for" causation). See supra Part I.C.4. However, it is not true if the plaintiff uses the basic version of McDonnell Douglas, which proves only "motivating factor" causation. See supra Part I.C.3. Dean Kaminshine's analysis is correct as to the strong version, which is the one on which he focuses.
125. See supra Part I.C.3.
126. See supra Part I.C.4.
127. By successfully using the strong version of McDonnell Douglas, the plaintiff would establish "sole causation," which by definition includes "but for" causation (since "sole factor" means a factor is both sufficient and necessary-a "but for" cause). See supra Part. I.C.4.
128. At least one court has understood this point. See Fogg v. Gonzales, 492 F.3d 447, 451 & n.* (D.C. Cir. 2007) (noting that a plaintiff can use McDonnell Douglas as one way of proving that discrimination "played a 'motivating part' " in the defendant-employer's decision (citing Price Waterhouse, 490 U.S. at 250)); see also Herawi v. Ala. Dep't of Forensic Scis., 311 F. Supp. 2d 1335, 1345-46 (M.D. Ala. 2004) (recognizing that McDonnell Douglas provides "one methodology for establishing liability through circumstantial evidence" in either "single- or mixed-motive cases").
129. In the text, I have discussed only two causal standards and allocations of burden: Price Waterhouse and the 1991 Act. These each require the plaintiff to prove only "motivating factor" causation, placing the burden of proving "but for" causation (or lack of it) on the defendant. See supra notes 51-55. It is possible that the law contains a third causal standard and allocation of burden; one in which the plaintiff bears the full burden of proving "but for" causation. I will argue below that current law should not-and does not-include this third causal standard and allocation of burden. See infra Part II.C.I, particularly text accompanying notes 182-191. However, in the event that the law did require some plaintiffs to bear the burden of proving "but for" causation, this subpart shows that they could elect to do so by using the strong version of McDonnell Douglas.
130. See, e.g., Tysinger v. Police Dep't, 463 F.3d 569, 577-78 (6th Cir. 2006) (discussing the different approaches courts have taken in deciding which framework to apply after Desert Palace); Sallis v. Univ. of Minn., 408 F.3d 470, 474-75 (8th Cir. 2005) (deliberating about which framework to apply and deciding to apply McDonnell Douglas rather than Price Waterhouse or Desert Palace); Price Waterhouse v. Hopkins, 490 U.S. 228, 278-79 (1989) (O'Connor, J., concurring), superseded by statute, Civil Rights Act of 1991, Pub. L. No. 102-166, 105 Stat. 1071 (indicating that "the court should determine whether the McDonnell Douglas or Price Waterhouse framework properly applies . . . ."); supra note 21 (listing scholars who believe that McDonnell Douglas is "dead" because of the availability of alternative frameworks after Desert Palace).
131. See supra Part I.A.
132. As noted above, McDonnell Douglas is a creature of the courts, not Congress. See supra note 49. Thus, there is no need for congressional action; the Court could adopt the proper understanding of McDonnell Douglas.
133. See supra note 32.
134. See Price Waterhouse, 490 U.S. at 278-79 (1989).
135. See Desert Palace, Inc. v. Costa, 539 U.S. 90, 101-02 (2003) (O'Connor, J., concurring).
136. See supra note 47 and accompanying text; see also Scott & Chapman, supra note 21, at 405 (arguing that 1991 Act framework should only apply in "mixed motive" cases). The issue of how to define a "mixed motive" case is significant. See infra note 142 (noting that there is no good definition for a "mixed motive" case for purposes of mandating the use of McDonnell Douglas).
137. See supra note 48 and accompanying text. This modified version mimics the standard version of McDonnell Douglas at the first two stages (the plaintiffs prima facie case and the defendant's proffered reason). But at the third stage, the modified version offers the plaintiff a choice: the plaintiff can prove either (1) that the proffered reason was pretextual (the "pretext alternative," just like the standard version of McDonnell Douglas), or (2) "that the defendant's reason, while true, is only one of the reasons for its conduct, and another 'motivating factor' is the plaintiff's protected characteristic" (the "mixed motive alternative"). See Rachid v. Jack in the Box, Inc., 376 F.3d 305, 312 (5th Cir. 2004) (citing Rishel v. Nationwide Mut. Ins. Co., 297 F. Supp. 2d 854, 865 (M.D.N.C. 2003)). If the plaintiff chooses the mixed motive alternative, the defendant can try to prove a "same action" defense to limit damages. See id.
138. Proponents of the "modified" McDonnell Douglas approach might argue that it is harmless, since at the third stage it lets plaintiffs choose between the traditional McDonnell Douglas pretext method and the 1991 Act framework. However, there are two flaws with this choice. First, it does not give litigants any choice about the first two stages. Under the "modified" framework, all plaintiffs must prove a prima facie case and all defendants must proffer a nondiscriminatory reason. While neither of these steps is onerous, see supra note 63 and accompanying text, they may well present unnecessary hurdles. See Davis, supra note 15, at 751 n.262 (citing cases where plaintiffs with pretext evidence have had their cases dismissed for failure to present a prima facie case); Ann C. McGinley, Credulous Courts and the Tortured Trilogy: The Improper Use of Summary Judgment in Title VII and ADEA Cases, 34 B.C. L. REV. 203, 228-37 (1993) (giving examples of plaintiff's burden); see also Malamud, supra note 6, at 2282-301 (noting that some courts require an excessive showing to make a prima facie case). Second, and perhaps more problematic, at the third stage of the "modified" McDonnell Douglas, it is a serious overstatement to say that plaintiffs have a choice; that they are truly free to decline to use the traditional McDonnell Douglas method. This is because plaintiffs can only exercise this option by paying a significant price: a plaintiff who elects this option must prove "that the defendant's reason, while true, is only one of the reasons for its conduct." See, e.g., Rachid, 376 F.3d at 312. That is, to opt out of the traditional McDonnell Douglas approach, the plaintiff must admit that the defendant's proffered reason is true-a significant admission. See Kaminshine, supra note 14, at 20 n.99 (noting that plaintiffs would generally not want to admit that the defendant had a nondiscriminatory motive). So the plaintiffs "choice" at the third stage of the "modified" McDonnell Douglas framework is somewhat illusory, at least as the "modified" McDonnell Douglas is currently constructed.
139. In a cryptic footnote, Desert, Palace reserved the issue of whether a "direct evidence" requirement might apply in "single motive" cases. See Desert Palace, 539 U.S. at 94 n.1. However, that case certainly did not mandate the use of McDonnell Douglas in all "single motive" cases. Moreover, as demonstrated above, it does not make sense to require the use of McDonnell Douglas in any case-even a "single motive" case without "direct evidence" (whatever that might be). See supra, Part I.
140. See supra notes 47-48 and accompanying text.
141. See supra notes 47-48 (noting that the Fifth Circuit, in separate opinions, adopts both approaches).
142. See supra Part I.
143. Moreover, each of these two approaches suffers from its own specific flaws. The "modified" McDonnell Douglas approach presents plaintiffs with a false dichotomy: at the third stage, the "modified" McDonnell Douglas requires plaintiffs to choose between a method of proof (pretext) and a standard of proof (the 1991 Act standard). See infra Part II.C.3, especially text accompanying notes 203 and 207 (discussing the false dichotomy posed by choice between McDonnell Douglas and alternative frameworks). The "mixed motive" approach (requiring plaintiffs without "mixed motive" cases to use McDonnell Douglas) also suffers from a serious flaw specific to that approach: there is no good definition of what constitutes a "mixed motive" case. See Fukete v. Aetna, Inc., 308 F.3d 335, 337-38 n.2 (3d Cir. 2002) (explaining confusion surrounding "mixed motive"/"single motive" distinction); Miller v. CIGNA Corp., 47 F.3d 586, 597 n.9 (3d Cir. 1995) (same). Prior to Desert Palace, most courts and commentators used the phrase to denote cases in which plaintiffs were required to use McDonnell Douglas-that is, cases in which the plaintiff did not have "direct evidence." See, e.g., Haddon v. Executive Residence at the White House, 313 F.3d 1352, 1357 (Fed. Cir. 2002). However, after Desert Palace, this conclusory definition is not available. Most of the post-Desert Palace courts that have adopted the "mixed motive" approach have defined the phrase by reference to an admission by the plaintiff: a "mixed motive" case is one in which the plaintiff admits that the defendant has legitimate, as well as illegitimate, motives. See, e.g., Winter v. Bank of Am., N.A., No. Civ.A.3:02-CV-1591-L, 2003 WL 23200278, at *3 (N.D. Tex. Dec. 12, 2003); see also Scott & Chapman, supra note 21, at 405 (providing alternative definition, in which one alternative requires the plaintiff to "prove" that the defendant had two motives, one of which was legitimate). However, this is a silly definition. As noted above, no plaintiff in his or her right mind would want to concede this point. See supra note 137. Moreover-and more importantly-there is no reason to think that Congress intended to limit the 1991 Act framework to plaintiffs who were willing to make such a damaging concession. Other writers have tried to define "mixed motive" cases by reference to an admission by the defendant: a "mixed motive" cases is one in which the defendant concedes that it had an illegal motive, as well as a legal one. See Scott & Chapman, supra note 21, at 405 (providing alternative definition, in which one alternative involves a concession by the defendant that it had an illegal motive). However, this definition is also silly. Again, few defendants would seem likely to want to make such a concession. And again, there is no reason to think that Congress intended to limit the 1991 Act framework to cases involving defendants who were foolish, particularly honest, or had been caught red-handed and so had no alternative but to admit that they had an illegal motive. Assuming that one needed a definition of "mixed motive" for some purpose-which is far from clear-the best definition is probably this: a case in which the factfinder might reasonably conclude that more than one factor influenced the defendant's decision. Presumably, the plaintiff would provide evidence of illegal factors and the defendant would provide evidence of legal factors. While this definition is workable, it could not serve as a way to mandate the use of McDonnell Douglas. As noted above, McDonnell Douglas presents only one of many ways that the plaintiff might try to show that an illegal factor influenced the defendant's decision. See supra Part H.A. The corollary of the workable definition of "mixed motive" is that a "single motive" case would be one in which a reasonable factfinder could conclude that only one factor, either legitimate or illegitimate, influenced the defendant's decision. Some McDonnell Douglas cases-those in which the defendant offers only one legitimate reason for its decision, which the factfinder must either believe or disbelieve-may result in such a record. But the fact that some McDonnell Douglas cases may present "single motive" cases is hardly a reason to require McDonnell Douglas in all "single motive" cases.
144. See Desert Palace, Inc. v. Costa, 539 U.S. 90, 98-102 (2003).
145. See supra Part I.A.
146. 29 U.S.C.A. §§ 621-634 (West 1999 & Supp. 2007).
147. 42 U.S.C. §§ 12101-12213 (2000 & Supp. IV 2004).
148. See Desert Palace, Inc., 539 U.S. at 101-02 (2003).
149. See, e.g., Richardson v. Monitronics Int'l, Inc., 434 F.3d 327, 334-35 (5th Cir. 2005) (finding that the district court erred in using the McDonell Douglas framework to analyze the plaintiffs FMLA claim); Rachid v. Jack in the Box, Inc., 376 F.3d 305, 311-12 (5th Cir. 2004) (holding direct evidence of discrimination unnecessary to ADEA claims); Estades-Negroni v. Assocs. Corp. of N. Am., 345 F.3d 25, 31 (1st Cir. 2003) (ADEA); Strauch v. Am. Coll. of Surgeons, 301 F. Supp. 2d 839, 844 (N.D. Ill. 2004) (ADEA); Warren v. Terex Corp., 328 F. Supp. 2d 641, 646 (N.D. Miss. 2004) (Title VII retaliation claim); Knutson v. Ag Processing, Inc., 273 F. Supp. 2d 961, 995 (N.D. Iowa 2003) (ADEA), rev'd on other grounds, 394 F.3d 1047 (8th Cir. 2005); Skomsky v. Speedway SuperAmerica, L.L.C., 267 F. Supp. 2d 995, 998-1000 (D. Minn. 2003) (ADA); Myers v. AT&T Corp., 882 A.2d 961, 971 (N.J. Super. Ct. App. Div. 2005) (state law). These courts have not gone on to assign McDonnell Douglas its proper role. Rachid, for example, required all plaintiffs to use a modified version of McDonnell Douglas. Rachid, 376 F.3d at 312. But these courts have at least eradicated the "direct evidence" distinction.
150. See Price Waterhouse, 490 U.S. at 261-62 (O'Connor, J., concurring). In one pre-Price Waterhouse case, the Court indicated that an ADEA plaintiff with "direct evidence" did not need to rely on McDonnett Douglas. See Trans World Airlines, Inc. v. Thurston, 469 U.S. 111, 121 (1985). But the Court did not hold that an ADEA plaintiff without "direct evidence" would have to use McDonnell Douglas. (And Justice O'Connor was the only Justice in Price Waterhouse-four years later-who was inclined to so hold. See supra notes 132-33 and accompanying text.)
151. See Desert Palace, 539 U.S. at 99-101.
152. Because the relevant portion of the 1991 Act amended section 703(a), we can think of all cases other than those under section 703(a) as non-1991 Act cases. See supra note 43.
153. See, e.g., EEOC v. Warfield-Rohr Casket Co., 364 F.3d 160, 163 n.1 (4th Cir. 2004) (assuming that "direct evidence is still a prerequisite for a mixed-motive analysis in ADEA cases"); Monaco v. Am. Gen. Assurance Co., 359 F.3d 296, 300 (3d Cir. 2004) (applying the "direct evidence" distinction in an ADEA case); Mereish v. Walker, 359 F.3d 330, 340 (4th Cir. 2004) (discussing why the Price Waterhouse "direct evidence" distinction should remain applicable to ADEA claims after Desert Palace); Harp v. Se. Pa. Transp. Auth., No. Civ.A. 04-2205, 2006 WL 1517390, at *6-7 (E.D. Pa. May 31, 2006) ("[I] n an ADA action, if direct evidence of discrimination is produced, the Price Waterhouse framework applies; if circumstantial evidence of discrimination is produced, the McDonnell Douglas framework applies. . . ."). But see Richardson v. Monitronics Int'l, Inc., 434 F.3d 327, 334-35 (5th Cir. 2005) (applying Desert Palace in a FMLA case); Rachid, 376 F.3d at 311 (applying the reasoning of Desert Palace in an ADEA case).
154. See Jamie Darin Prenkert, Bizarro Statutory Stare Decisis, 28 BERKELEY J. EMP. & LAB. L. 217 (2007). Professor Prenkert's article focuses on a different application of Bizarro statutory stare decisis: the Court's contention that a particular disparate impact case which, like Price Waterhouse, was overruled by the 1991 Act, should somehow apply in non-1991 Act cases. But the basics of the argument are the same and clearly apply to Price Waterhouse. See id. at 256-63 (discussing the potential application of the Bizarro argument to Price Waterhouse).
155. See Martin J. Katz, Unifying Disparate Treatment (Really), 59 HASTINGS L.J. (forthcoming Feb. 2008) (manuscript at 17-18), available, at http://ssrn.com/sol3/ papers.cfm?abstract_id=982417 (describing the assumption of uniformity and why it makes sense).
156. See id. (explaining why the assumption of uniformity justified the expansion of Price Waterhouse to cases outside of Section 703(a)).
157. See id. (explaining why the assumption of uniformity now suggests that Congress' corrected interpretation in the 1991 Act, and not Price Waterhouse, should apply in non-1991 Act cases).
158. See id. (manuscript at 18) (explaining "limited amendment" argument-or, what Professor Prenkert calls the Bizarro argument). The courts seem split over whether to accept the Bizarro argument. Compare id. (manuscript at 17 n.61) (citing cases that apply Price Waterhouse to non-1991 Act claims), with id. (manuscript at 19 n.66) (citing cases that do not apply Price Waterhouse to non-1991 Act claims).
159. See id.
160. See supra note 32; see also Marks v. United States, 430 U.S. 188, 193 (1997) ("When a fragmented Court decides a case and no single rationale explaining the result enjoys the assent of five Justices, 'the holding of the Court may be viewed as that position taken by those members who concurred in the judgments on the narrowest grounds. . . .'" (quoting Gregg v. Georgia, 428 U.S. 153, 169 n.15 (1976))).
161. Price Waterhouse v. Hopkins, 490 U.S. 228, 261-79 (1989) (plurality opinion), superseded by statute, Civil Rights Act of 1991, Pub. L. No. 102-166, 105 Stat. 1071; id. at 259-60 (White, J., concurring).
162. See id. at 276 (O'Connor, )., concurring).
163. See supra Part I.C.3.
164. Price Waterhouse, 490 U.S. at 257 (plurality opinion); id. at 258-60 (White, J., concurring).
165. Id. at 249-50 (plurality opinion); id. at 259 (White, J., concurring). While Justice White's concurrence is a little cryptic on this matter, he states at the outset of his concurrence that the correct approach is found in Mt. Healthy City School District Board of Education v. Doyle, 429 U.S. 274 (1977). See Price Waterhouse, 490 U.S. at 258-59 (White, J., concurring). The Mt. Healthy approach is essentially identical to that of the plurality: if a plaintiff proves "motivating factor" causation, the burden shifts to the defendant to prove a lack of "but for" causation (and, if the defendant succeeds in this, there is no liability). Mt. Healthy, 429 U.S. at 287. The only point on which he seems to disagree with the plurality is that he believed the plurality wanted defendants to provide "objective" evidence of their reasons-which does not seem to have been an issue in any post-Price Waterhouse cases (or in Price Waterhouse). See Price Waterhouse, 490 U.S. at 261. But see Garcia v. City of Houston, 291 F.3d 672, 677 (5th Cir. 2000) (discussing, but not requiring, an objective evidence requirement and finding defendant produced sufficient objective evidence); Foster v. Univ. of Ark., 938 F.2d 111, 114 (8th Cir. 1991) (noting the objective evidence requirement set forth by the plurality in Price. Waterhouse, but not deciding on that basis); Ford v. St. Elizabeth Hosp., No. 92-CV-511, 1993 WL 330036, at *5 (N.D.N.Y. Aug. 20, 1993) (considering objective evidence but not specifically basing the court's decision on it).
166. See, e.g., Monaco v. Am. Gen. Assurance Co., 359 F.3d 296, 300 (3d Cir. 2004).
167. See Price Waterhouse, 490 U.S. at 278-79 (O'Connor, J., concurring). I have put "deserving" in quotes to make clear that I do not necessarily subscribe to this view of plaintiffs' merits. But for purposes of this argument, I will accept Justice O'Connor's view on the matter.
168. See supra Part I.C.3.
169. Remember, even if she wanted to, Justice O'Connor could not have required plaintiffs without "direct evidence" to use the strong version of McDonnell Douglasthat is, to prove "sole factor" causation. See supra Part I.C.4. Moreover, it is fairly clear that Justice O'Connor did not intend to require "sole factor" causation. See Price Waterhouse, 490 U.S. at 265-66 (O'Connor, J., concurring) (focusing on "substantial factor" rather than "sole factor" causation).
170. Recall that, under Price Waterhouse, if there is less than "but for" causation (that is, if the defendant proves that it would have made the "same decision" absent the protected characteristic), the plaintiff loses. See supra note 119.
171. See Price Waterhouse, 490 U.S. at 261-70 (O'Connor, J., concurring).
172. This would raise the question of whether there is any other way of proving "but for" causation besides the strong version of McDonnell Douglas, short of an employer's admission that it used a protected characteristic as a "but for" factor in the challenged decision. The answer is that there is probably not another way to prove "but for" causation. See infra Part II.C.1. But even if this were true, the plaintiff in such a case would be required to use the strong version of McDonnell Douglas not because any mandate by the Court, but rather because the law required the plaintiff to prove "but for" causation and the strong version of McDonnell Douglas happened to be the only way to prove this standard.
173. See infra note 178 and accompanying text.
174. See infra Part II.C.1 (discussing the doctrinal argument) and note 182 (discussing the normative arguments).
175. The text below discusses three debates that are resolved once we understand the proper role for McDonnell Douglas as a nonmandatory method of proof. In addition, the understanding of how McDonnell Douglas works that underlies its nonmandatory nature resolves three additional debates that have long plagued disparate treatment law. See supra notes 80 (resolving the "pretext-plus" debate and laying to rest a long-running academic debate over the correctness of the Court's decision in Hicks) and 67 (resolving the debate over the availability of a "good faith" defense under McDonnell Douglas).
176. See supra Part I.A.
177. Sue supra notes 32-37 and accompanying text.
178. See supra note 44.
179. See Costa v. Desert Palace, Inc., 299 F.3d 838, 852-53 (9th Cir. 2002) (describing circuit split), aff'd, 539 U.S. 90 (2003).
180. See supra Part I.A.
181. See supra Part II.B.1 (noting that Desert Palace eradicated "direct evidence" requirement in 1991 Act cases), II.B.2.b-c (arguing that courts are under no obligation to apply the "direct evidence" doctrine in non-1991 Act cases).
182. See supra Part II.B.2.a.
183. In addition to the doctrinal argument in the text, there are also four normative arguments about why a plaintiff should never be required to prove "but for" causation: First, as suggested in the text immediately below, proving "but for" causation is extremely difficult. See infra notes 183-85 and accompanying text. Second, it is much more difficult for the plaintiff to prove "but for" causation (which requires proof of a negative-that is, proof of the lack of any independently sufficient factor in the decision) than it is for the defendant to prove a lack of "but for" causation (which requires only proof of a positive-that is, a single independently sufficient factor in the decision). Third, most of the relevant evidence on the issue of causation is within the defendant's control. In fact, most of it is in the defendant's head. See Katz, supra note 30, at 515-17. And finally, once the plaintiff proves "motivating factor" causation, the defendant has been established as a wrongdoer and the party responsible for the need to sort out "but for" causation. See RUTHERGLEN, supra note 3, at 48. These facts provide a strong normative argument against placing the burden on plaintiffs to prove "but for"-and an argument for burden-shifting mechanisms, such as those in the 1991 Act or Price Waterhouse.
184. See supra notes 109-10 and accompanying text.
185. See supra Part I.C.4.
186. See supra Part I.C.4.
187. See Price Waterhouse v. Hopkins, 490 U.S. 228, 271 (1988) (O'Connor, J., concurring), superseded by statute, Civil Rights Act of 1991, Pub. L. No. 102-166, 105 Stat. 1071; see also Costa v. Desert Palace, 299 F.3d 838, 852-53 (9th Cir. 2002), aff'd, 539 U.S. 90 (2003) (explaining various definitions of "direct evidence").
188. Theoretically, the defendant might admit that he not only used a protected factor (such as race or sex) in the challenged decision, but also that he would not have reached the same decision absent this factor. This type of admission would, of course, prove "but for" causation. However, such admissions are sufficiently unlikely that it would seem quite safe to discount this possibility.
189. Recall that what prevents a protected factor (such as race or sex) from being a "but for" cause is the existence of a second nonprotected factor (such as excessive tardiness) that would be sufficient to bring about the challenged employment decision. See. supra note 54 and accompanying text.
190. Note that there is no similar problem with asking defendants to prove a lack of "but for" causation. Such a requirement simply obliges a defendant to advance at least one independently sufficient factor.
191. See supra Part I.C.4.
192. One might resist this conclusion by arguing that the Supreme Court has required certain plaintiffs to bear the full burden of proving "but for" causation. However, the only opinion in which the Supreme Court has done so unequivocally is Price Waterhouse, 490 U.S. at 262-63 (concluding that plaintiffs without "direct evidence" must bear full burden of proving "but for" causation). As noted above, this opinion has been overruled, at least in the 1991 Act context. See supra Part I.A; see also Desert Palace, Inc., 539 U.S. at 101-02 (holding, in 1991 Act case, that "direct evidence of discrimination is not required in mixed-motive cases"). And, as discussed above, Price Waterhouse has no place in non-1991 Act cases. See supra Part II.B.2.a; see also Katz, supra note 154 (manuscript at 14-20) (arguing that Price Waterhouse, having been legislatively overruled in 1991 Act cases, should not be understood to apply to non-1991 Act cases). Some writers believe the Court imposed such a burden in Hazen Paper, where it said that a plaintiff can only succeed where the protected factor (age) "had a determinative influence on the outcome" (firing). Hazen Paper Co. v. Biggins, 507 U.S. 604, 610 (1993). "Determinative influence" almost certainly means "but for." See Katz, supra note 30, at 501-03. But Hazen Paper did not address burdens of proof. The plaintiff there was not asking for a burden-shift, and the sentence in question says nothing about who has the burden of proving "determinative influence." See Hazen Paper, 507 U.S. at 610 (noting the necessity of proving "determinative influence," but failing to address who bears the burden of proof). In any event, to the extent that a Supreme Court opinion on this issue could be seen as requiring plaintiffs to bear the full burden of proving "but for," it would conflict with Congress' instruction that "sole factor" can never be required-and would have to yield. See supra notes 109-10 and accompanying text.
193. There are two additional arguments against the "direct evidence" doctrine. First, Congress has not used this term in any disparate treatment statute. In Desert Palace, the Court held that the fact that this term appears nowhere in the 1991 Act or its legislative history suggests that Congress simply did not intend to require plaintiffs suing under that Act to provide "direct evidence." Desert Palace, 539 U.S. at 91. But the same can be said for virtually every other disparate treatment statute. None refer to "direct evidence." Thus, it would seem, Congress did not intend to require plaintiffs under those statutes to provide "direct evidence." Some courts have already adopted this view. See supra note 148 and accompanying text. A second argument against the "direct evidence" doctrine is that there is no good definition for this term. We have seen that McDonnell Douglas provides one way to prove causation-one type of evidence. See Part I.C. Under virtually any definition, this method is not "direct"; it clearly relies on circumstantial evidence. Thus, we can envision two possible definitions of "direct evidence." First, "direct evidence" could be defined as any type of evidence other than McDonnell Douglas. But given the purpose of "direct evidence" (determining which plaintiffs must use McDonnell Douglas), this definition would be circular. It would require plaintiffs who use McDonnell Douglas (those without "direct evidence") to use McDonnell Douglas. The second option would be to define "direct evidence" as some subset of non-McDonnell Douglas evidence. That is, we could divide non-McDonnell Douglas evidence into two categories: "direct" and "indirect." The problem with this definition is that it would effectively foreclose the use of one of those subcategories of evidence. Plaintiffs with "direct evidence" would be allowed to use that type of evidence. But plaintiffs with "indirect evidence" would be required to use McDonnell Douglas-and be precluded from using the subcategory of non-McDonnell Douglas evidence defined as "indirect." Thus, there would seem to be no good definition for "direct evidence." See GEORGE RUTHERGLEN, EMPLOYMENT DISCRIMINATION LAW 49-50 (1st ed. 2001) (noting that the "direct evidence" distinction "transforms a question of degree-how closely evidence is connected to a fact in dispute-into a question of kind-whether it is connected closely enough to be 'direct'").
194. Desert Palace, 539 U.S. at 101-02.
195. See supra notes 177-78.
196. See supra Parts I.C.3-5; see also supra Part II.C.1 (arguing that lower courts need not apply the "direct evidence" doctrine, either in 1991 Act or non-1991 Act cases).
197. Aficionados of Monty Python will realize that I must give credit here to the movie, MONTY PYTHON AND THE HOLY GRAIL (Michael White Productions 1975), as well as its more recent incarnation in the Broadway show, Spamalot.
198. See supra note 21 (listing those who believe McDonnell Douglas is "dead"); see also Davis, supra note 6, at 862 (arguing that Desert Palace reduced McDonnell Douglas to "an empty formality" which should be abandoned).
199. Diamond v. Colonial Life & Accident Ins. Co., 416 F.3d 310, 318 & n.4 (4th Cir. 2005); Griffith v. City of Des Moines, 387 F.3d 733, 735-36 (8th Cir. 2004); McGinest v. GTE Serv. Corp., 360 F.3d 1103, 1122 (9th Cir. 2004); Hill v. Lockheed Martin Logistics Mgmt., Inc., 354 F.3d 277, 285 (4th Cir. 2004); Estades-Negroni v. Assocs. Corp. of N. Am., 345 F.3d 25, 30 (1st Cir. 2003), vacated, 362 F.3d 874 (1st Cir. 2004); Fye v. Okla. Corp. Comm'n, No. CIV-03-1477-C, 2006 WL 1716557, at *1 (W.D. Okla. June 21, 2006).
200. See supra note 48 and accompanying text.
201. See supra note 47 and accompanying text.
202. See Dunbar v. Pepsi-Cola Gen. Bottlers, Inc., 285 F. Supp. 2d 1180, 1196 (N.D. Iowa 2003) (remarking that Desert Palace ends the "false dichotomy" between McDonnell Douglas and the 1991 Act/Price Waterhouse). See generally T.L. Nagy, The Fall of the False Dichotomy: The Effect of Desert Palace v. Costa on Summary Judgment in Title VII Discrimination Cases, 46 S. TEX. L. REV. 137, 154-61 (2004) (arguing in favor of the Dunbar approach). Ironically, Dunbar was the case that began the use of a modified version of McDonnett Douglas. See Dunbar, 285 F. Supp. 2d at 1197. As noted below, this modified version continues to posit a false dichotomy.
203. See, e.g, Corbett, supra note 58, at 1576.
204. See id.; Van Delta, Le Roi Est Mori, supra note 15, at 72. Professor Zimmer takes a more nuanced view, arguing that McDonnett Douglas will only likely be used where the parties agree to use it and in those rare situations where a defendant fails to proffer a nondiscriminatory reason for its actions. See Zimmer, supra note 11, at 1932. However, like the other scholars, Zimmer assumes that someone must make a choice between McDonnell Douglas and the 1991 Act (either both of the parties by agreement, or the court). Id. at 1939.
205. See, e.g., Griffith, 387 F.3d at 735-36.
206. See, e.g., Rachid v. Jack in the Box, Inc., 376 F.3d 305, 312 (5th Cir. 2004).
207. See, e.g., Richardson v. Monitronics Int'I, Inc., 434 F.3d 327, 332-33 (5th Cir. 2005).
208. See supra Part II.A.
209. See supra notes 38-39 and accompanying text.
210. See, e.g., Bloomer v. United Parcel Serv., Inc., 94 F. App'x 820, 826 (10th Cir. 2004).
211. See, e.g., Rachid, 376 F.3d at 312.
212. See supra Part I.
213. As noted above, proponents of a "modified" McDonnell Douglas might argue that the modification to the framework at least ameliorates the problem of a mandatory McDonnell Douglas. The modification, after all, seems to give plaintiffs a choice as to whether to use the third-most difficult-stage of McDonnett Douglas. See supra note 137. However, as also discussed above, the modified version remains problematic, as it requires all litigants to use the first two stages of McDonnell Douglas and also exacts a high price for plaintiffs to opt out of the third stage (they must concede that the defendant had a legitimate reason). See supra note 137.
214. Dean Kaminshine argues, as I do, that McDonnell Douglas is not "dead." See Kaminshine, supra note 14, at 7. However, his argument is based on the claim that McDonnell Douglas and the 1991 Act are mutually exclusive; that McDonnell Douglas proves (and thus, he believes, requires) "but for" causation, while the 1991 Act requires "motivating factor" causation. See id. at 41. However, as we have seen above, McDonnell Douglas-or at least the basic version of it, which is all that can be possibly be required-does not prove "but for" causation. See supra Part I.C.3.
215. See supra Part II.A; see also infra Part III.E. (explaining why many plaintiffs should-and do, and will-choose McDonnell Douglas).
216. See supra Part I.C.5.
217. See supra Part I.A.
218. There is an additional flaw with the argument that McDonnell Douglas is "dead." This argument is based largely on the premise that McDonnell Douglas requires "but for" causation. As noted above, the "dead" camp believes that, given the choice, no plaintiff in her right mind would choose McDonnell Douglas over the 1991 Act framework. See supra note 203 and accompanying text. This argument is generally premised on the idea that the 1991 Act requires plaintiffs to prove only "motivating factor" causation, while McDonnell Douglas requires "but for" causation. See Corbett, supra note 58, at 1576; Van Delta, Le Roi Est Mort, supra note 15, at 117-19. However, as we have seen above, McDonnell Douglas does not represent a "but for" requirement. See supra Part I.C.3.
219. See, e.g., McCormick, supra note 15, at 183.
220. See, e.g., Malamud, supra note 6, at 2237-38.
221. See, e.g., McCormick, supra note 15, at 161 (noting that "courts. . . get so caught up in principles promoted by the test that they lose sight of the law," which has made it "nearly impossible to combat discrimination"); see also Wells v. Colo. Dep't of Transp., 325 F.3d 1205, 1221 (10th Cir. 2003) (Hartz J., concurring) ("The McDonnell Douglas framework only creates confusion and distracts courts from 'the ultimate question of discrimination vel non.' McDonnell Douglas has served its purpose and should be abandoned." (quoting U.S. Postal Serv. Bd. of Governors v. Aikens, 460 U.S. 711, 714 (1983))); Malamud, supra note 6, at 2279 (arguing that the McDonnell Douglas "proof structure actually does little to aid [courts'] analysis of the facts").
222. See supra note 74 (discussing propriety of instructing juries on pretext and burden-shifting aspects of McDonnell Douglas).
223. See generally Michael J. Zimmer, Slicing & Dicing of Individual Disparate Treatment Law, 61 LA. L. REV. 577, 584-85, 595-97 (2001) (explaining how courts "slice[ ] and dice[ ]. . . the evidence in the record in order to draw inferences in favor of the employer").
224. See, e.g., id. at 592-600.
225. Pretext evidence of causation should be viewed cumulatively with nonpretext evidence. And nonpretext evidence may be relevant to pretext. For example, racist statements by a decisionmaker might make it more likely that a proffered reason that was incorrect was also a lie, a cover-up, and a cover-up for discrimination. See Kaminshine, supra note 14, at 56 (noting that nonpretext evidence might be relevant to pretext and vice versa).
226. See Brodin, supra note 64, at 183, 215; Derum & Engle, supra note 8, at 1224-25; Rosenthal, supra note 15, at 335 ("Unfortunately for plaintiffs, the general rule that has developed [where employers offer] multiple justifications is that, to defeat employers' motions for summary judgment, plaintiffs must typically present evidence that each of these reasons is not the true reason for the action, and that discrimination was the real reason for the adverse action."); see also Wallace v. Methodist Hosp. Sys., 271 F.3d 212, 220 (5th Cir. 2001) ("The plaintiff must put forward evidence rebutting each of the nondiscriminatory reasons the employer articulates."); Clay v. Holy Cross Hosp., 253 F.3d 1000, 1007 (7th Cir. 2001) ("[The plaintiff] must present facts to rebut each and every legitimate, nondiscriminatory reason advanced by the [defendant] in order to survive summary judgment."); Herawi v. Ala. Dep't of Forensic Scis., 311 F. Supp. 2d 1335, 1346 (M.D. Ala. 2004) (requiring the plaintiff to disprove all of the reasons proffered by the defendant). But see Bryant v. Farmers Ins. Exch., 432 F.3d 1114, 1126 (10th Cir. 2005) (stating that "'when the plaintiff casts substantial doubt on many of the employer's multiple reasons, the jury could reasonably find the employer lacks credibility' " (quoting Tyler v. RE/MAX Mountain States, Inc., 232 F.3d 808, 814 (10th Cir. 2000))); Strate v. Midwest Bankcentre, Inc., 398 F.3d 1011, 1021 (8th Cir. 2005) (rejecting the defendant's argument that the plaintiff must rebut every reason offered by the defendant).
227. See, e.g., Hicks v. St. Mary's Honor Ctr., 90 F.3d 285, 290-91 (8th Cir. 1996) (affirming the lower court's holding that the plaintiff did not adequately demonstrate that racial discrimination rather than personal animosity motivated his termination); Brodin, supra note 64, at 215 ("What makes no sense is requiring the plaintiff to disprove theories not put into play by defendant."); Derum & Engle, supra note 8, at 1224-25 (noting that even when the plaintiffs rebut all of the defendant's nondiscriminatory reasons, courts may nevertheless presume that personal
228. See Derum & Engle, supra note 8, at 1224-28 (explaining the "personal animosity" presumption and presenting cases where courts have applied this presumption); cf. Brodin, supra note 64, at 215-29 (discussing how the "personality" excuse is the "ultimate pretext" which could potentially "eviscerate[ ] Title VII's protections").
229. See supra Part I.C.1.
230. See supra Part I.C.4.
231. As noted above, it is possible that there is a third alternative framework (in addition to Price Waterhouse and the 1991 Act), under which the plaintiff must bear the full burden of proving "but for" causation (rather than shifting that burden to the defendant upon a showing of "motivating factor," as in Price Waterhouse and the 1991 Act). See supra Part II.B.2, especially text accompanying notes 166 to 171. But as also noted above, such a plaintiff-must-prove-"but for" framework should not be seen as existing in current doctrine (and would be normatively flawed). See supra Part II.C.1, especially text accompanying notes 183 to 192.
232. See supra Part II.A.
233. See supra Part I.C.2-3.
234. See McDonnell Douglas Corp. v. Green, 411 U.S. 792, 805 (1973) (holding that an employee "must be given a full and fair opportunity to demonstrate by competent evidence that the presumptively valid reasons from his rejection were in fact a coverup").
235. Tex. Dep't of Cmty. Affairs v. Burdine, 450 U.S. 248, 258 (1981); see, e.g., Aka v. Wash. Hosp. Ctr., 156 F.3d 1284, 1298 (D.C. Cir. 1998) (stating that "courts traditionally treat explanations that rely heavily on subjective considerations with caution"); Rosenthal, supra note 15, at 370-72 (noting that a few courts have expressed suspicion when a defendant presents subjective reasons).
236. See Hart, supra note 8, at 767.
237. See supra Part I.C.1.
238. See supra note 228.
239. See supra Part I.C.1.
240. See Derum & Engle, supra note 8, at 1187-88, 1241; Krieger, supra note 15, at 1161-64; David N. Rosen & Jonathan M. Freiman, Remodeling McDonnell Douglas: Fisher v. Vassar College and the Structure of Employment Discrimination Law, 17 QUINNIPIAC L. REV. 725, 747-48 (1998); Selmi, supra note 80, at 325-26; Selmi, supra note 7, at 563; John Valery White, The Irrational Turn in Employment Discrimination Law: Slouching Toward a Unified Approach to Civil Rights Law, 53 MERCER L. REV. 709, 717 n.37 (2002); see also Michael E. Gold, Towards a Unified Theory of the Law of Employment Discrimination, 22 BERKELEY J. EMP. & LAB. L. 175, 188 (2001) (suggesting that McDonnell Douglas depends for its operation on the assumption that discrimination is prevalent); McCormick, supra note 15, at 176 (same); Sullivan, supra note 21, at 930 (same). Perhaps the most colorful version of the assumption-of-discrimination argument is made by Professor Galloway, who equates a failure to assume discrimination in such cases with Holocaust denial. See Deborah A. Galloway, St. Mary's Honor Center v. Hicks: Questioning the Basic Assumption, 26 CONN. L. REV. 997, 1036 (1994).
241. See supra Part I.C.2.
242. This argument is related to, but distinct from, the argument that McDonnell Douglas works by a process of elimination. The elimination argument concludes that, once the defendant's proffered reason has been eliminated, discrimination is the only possible, explanation. As discussed above, in Part I.C.2, this argument is erroneous. The prevalence-of-discrimination argument is that, once the proffered reason and the common reasons of the prima facie case have been eliminated, discrimination is-or should be considered-the most likely explanation. An excellent discussion of this reasoning, as well as its empirical flaws, can be found in Malamud, supra note 6, at 2254-62.
243. See supra Part I.C.1.
244. See FED. R. CIV. P. 56(c). A need-for-assumption-of- discrimination critic might argue that, absent an assumption of widespread discrimination, the facts would be in equipoise, requiring summary judgment for the defendant. The argument is that, absent such an assumption, a factfinder would have no way to choose which inference to draw in the pretext chain: no way to determine (1) whether an incorrect proffered reason was a good faith error or a lie, (2) whether a lie was a benign lie or a cover-up, or (3) whether a cover-up was for a non discriminatory reason or a discriminatory reason. Absent such guidance, the argument goes, summary judgment would be appropriate. However, this argument confuses the standard for a plaintiffs verdict (a preponderance of evidence/more likely than not) with the standard for summary judgment. Where two competing fact scenarios are in equipoise, a factfinder should find for the defendant (under the preponderance standard). However, whether two competing fact scenarios are in equipoise is a fact question, which cannot be resolved on summary judgment. The whole point is that it is the factfinder's role to determine the likelihood that each of two competing scenarios occurred. While assumptions as to the widespread nature of discrimination might aid the factfinder in making such a determination, such assumptions play no role on summary judgment. See JACK H. FRIEDENTHAL ET AL., CIVIL PROCEDURE 475 & n.15 (4th ed. 2005) (noting that where competing inferences are equally plausible, summary judgment should be denied: as "the trial judge need only determine that the non-movant's [here, plaintiffs] evidence on an issue is 'facially plausible and capable of being accepted by a rational factfinder' in order to deny a summary-judgment motion" (quoting James Joseph Duane, The Four Greatest Myths About Summary Judgment, 52 WASH. & LEE L. REV. 1523, 1560 (1995))). Put differently, equipoise goes to whether the plaintiff met its burden of persuasion-a factual question. On summary judgment, the question is essentially whether the plaintiff met its burden of production-of producing evidence that would allow a reasonable factfinder to find the fact in question. See supra note 79 and accompanying text (noting that factfinders can infer guilt from the fact that the defendant ran from the scene of an accident-even if an equally likely explanation might exist).
245. See supra Part I.C.2.
246. The validity of such an argument is beyond the scope of this Article.
247. See supra notes 78-80 and accompanying text.
248. It is true that the prima facie case relies on the elimination method. See Tex. Dep't of Cmty. Affairs v. Burdine, 450 U.S. 248, 253-54 (1981) ("The prima facie case serves an important function in the litigation: it eliminates the most common nondiscriminatory reasons for the plaintiffs rejection."). However, the analysis in the text highlights what a weak inference this is (at least without an assumption of widespread discrimination). To the extent that we rely on the prima facie case solely as a way to trigger the defendant's obligation to proffer a nondiscriminatory reason, the weakness of this inference does not seem problematic. See supra note 91. But it is more problematic when we speak of the "continuing relevance" of the prima facie case at the pretext stage. See Reeves v. Sanderson Plumbing Prods., Inc., 530 U.S. 133, 143 (2000) ("[A]lthough the presumption of discrimination 'drops out of the picture' once the defendant meets its burden of production. . . the trier of fact may still consider the evidence establishing the plaintiffs prima facie case and 'inferences properly drawn therefrom... on the issue of whether the defendant's explanation is pretextual."' (quoting St. Mary's Honor Ctr. v. Hicks, 509 U.S. 502, 511 (1993); Burdine, 450 U.S. at 256'n. 10)).
249. See Dare v. Wal-Mart Stores, Inc., 267 F. Stipp, 2d 987, 991 (D. Minn. 2003) (criticizing-and rejecting-McDonnell Douglas as a false dichotomy); Hart, supra note 8, at 758 (arguing that "courts applying the McDonnell-Douglas framework mistakenly assume that employment decisions are motivated by a single factor-either honest business judgment or dishonest discriminatory motivation" and explaining that "employment decisions are not either-or events, but events with multiple motivations"); Krieger, supra note 15, at 1179 ("Within the pretext paradigm, it is simply not possible for an employment decision to be both motivated by the employer's articulated reasons and tainted by intergroup bias; the trier of fact must decide between the two."); Stonefield, supra note 64, at 113 (stating the assumption that "a dichotomous result [lies] at the end of the McDonnell Douglas fact-finding road"); see also Price Water-house v. Hopkins, 490 U.S. 228, 247 (1989) (plurality opinion) ("Where a decision was the product of a mixture of legitimate and illegitimate motives... it simply makes no sense to ask whether the legitimate reason was the 'true reason'. . . for the decision-which is the question asked by [McDonnell Douglas]." (internal quotation marks and citation omitted)), superseded by statute, Civil Rights Act of 1991, Pub. L. No. 102-166, 105 Stat. 1071; Thomas v. Eastman Kodak Co., 183 F.3d 38, 58 (1st Cir. 1999) (discussing "truth versus lies" claim); Miller v. CIGNA Corp., 47 F.3d 586, 597 (3d Cir. 1995) (en banc) (same); id. at 600 (Greenburg, J., concurring) (same); Waltman v. Int'l Paper Co., 875 F.2d 468, 481 (5th Cir. 1989) (same); Green, supra note 8, at 91-92 (arguing that McDonnell Douglas presents an either-or paradigm); Kaminshine, supra note 14, at 18 & n.87 (same); Stonefield, supra note 64, at 113 (same).
250. See 110 CONG. REC. 13,837, 13,837 (1964) (statement of Sen. Case) ("If anyone ever had an action that was motivated by a single cause, he is a different kind of animal from any I know of."); RUTHERGLEN, supra note 3, at 47 ("[F]ew cases are likely to involve employers who rejected the plaintiff for discriminatory reasons alone."); Krieger, supra note 15, at 1215 ("It will be the rare employer indeed who can accurately identify the reasons why he hired or promoted one employee over another, fired another, or set salary increases as he did.... [I]n the real world one simply cannot control the multiplicity of causal antecedents so as to determine the causal efficacy of race, gender, national origin, or age.").
251. Hart, supra note 8, at 758.
252. E.g., Bryant v. Farmers Ins. Exch., 432 F.3d 1114, 1126 (10th Cir. 2005); Clay v. Holy Cross Hosp., 253 F.3d 1000, 1007-08 (7th Cir. 2001); Wallace v. Methodist Hosp. Sys., 271 F.3d 212, 220 (5th Cir. 2001); see also Rosenthal, supra note 15, at 335 ("[E]mployers. . . often articulate numerous justifications....").
253. See supra Parts I.C.2-3, III.B.
254. Again, I make no claim as to the likelihood that a factfinder will find only some of a defendant's multiple proffered reasons to be pretextual. See supra note 89. It is certainly possible that factfinders might tend to assess the credibility of the employer on a global level, finding either all or none of the employer's proffered reasons to be truthful. The point is that there is nothing about the framework that mandates such a binary view.
255. See supra note 248 and accompanying text.
256. In fact, even this choice is not binary. As noted in Part I.C.1, a factfinder can conclude that a proffered reason is not correct without concluding it was a cover-up for discrimination.
257. See supra notes 67-74 and accompanying text.
258. Cf. supra Part I.C.2-3 (noting that McDonnell Douglas cannot prove "but for" causation because it cannot eliminate all nondiscriminatory reasons for the defendant's actions).
259. For example, the plaintiff might be able to prove that the tardiness policy had never been applied before despite tardiness by other employees, or that it was not applied uniformly among employees of different races, sexes, or ages. Such comparative evidence is not pretext evidence. See supra note 76.
260. See Krieger, supra note 15, at 1179. Professor Krieger focuses on unconscious bias-that is, stereotypes outside of the decisionmaker's awareness. But the problem-this limitation on the utility of McDonnell Douglas-is the same whether the bias that infects the proffered reason is conscious or unconscious.
261. Moreover, as I will argue below, it works in many cases where other methods will not. See infra Part III.E.
262. See Krieger, supra note 15, at 1241 (arguing that McDonnell Douglas is incapable of addressing subtle or unconscious discrimination); see also Hart, supra note 8, at 756-57, 765 (noting that many courts applying McDonnell Douglas require plaintiffs to show pretext by showing the employer lied, an approach that only leads to liability for conscious, not unconscious, discrimination); Wax, supra note 15, at 1147-49 (stating that "[t]he McDonnell Douglas formulation is clearly geared to a narrow view of discriminatory intent: its operation depends on a defendant's possessing a conscious or deliberate state of mind" and arguing that McDonnell Douglas does not adequately address unconscious bias).
263. See, e.g., Hart, supra note 8, at 757-58.
264. It is open to debate whether it makes sense to hold people accountable for mental processes that are outside of their consciousness, both from a moral perspective (whether it makes sense to assign blame) and a practical perspective (whether there is any prospect of deterrence). However, such questions are beyond the scope of this Article. This Article will simply address the criticism that McDonnell Douglas is incapable of detecting unconscious bias-taking as a starting point the assumption of these critics that doing so would be a good thing.
265. See, e.g., Hart, supra note 8, at 754-55 & n.75 (cataloging examples). Professor Hart argues that the law does not in fact require proof of dishonesty. See id. at 755-66. However, she does not argue the same for McDonnell Douglas. Rather, she (correctly) asserts that McDonnell Douglas does require dishonesty, but argues that, in light of the 1991 Act and Desert Palace, courts should not require plaintiffs to use McDonnell Douglas. See id. at 758.
266. See, e.g., id. at 757-58.
267. See Martin Katz, No Intent, No foul?, LEGAL TIMES, May 21, 2007, at 35.
268. See supra notes 78-79 and accompanying text.
269. See, e.g., Forrester v. Rauland-Borg Corp., 453 F.3d 416, 419 (7th Cir. 2006).
270. See supra Part I.C.1.
271. The most common form of evidence offered in such cases is comparative evidence (e.g., that other nonminority employees stole money but were not fired). Such comparative evidence is not properly thought of as evidence of pretext in the McDonnell Douglas sense. See supra note 76.
272. At least some critics of McDonnell Douglas do not advocate scrapping it. Rather, they propose supplementing it. See, e.g., Green, supra note 8, at 144-56.
273. See supra Part I.C.3-5.
274. See. e.g., Krieger, supra note 15, at 1213.
275. See, e.g., Sun v. Bd. of Trs. of the Univ. of 111., 473 F.3d 799, 813 (7th Cir. 2007) (holding that "stray" remarks that are insufficiently related to a decision will not defeat a motion for summary judgment); Shorter v. ICG Holdings, Inc., 188 F.3d 1204, 1209-10 (10th Cir. 1999) (same); EEOC v. Clay Printing Co., 955 F.2d 936, 941-42 (4th Cir. 1992) (same). While it is beyond the scope of this Article, the weight that should be given to such remarks would seem to be a factual question for a jurynot a question of relevance and therefore admissibility. Any such remark would seem to make it more likely that a decision by the speaker was biased.
276. There has been a fair amount of discourse about using experts for another purpose: to testify about the prevalence or likelihood of subtle forms of discrimination in workplace culture. See Lee, supra note 80, at 487 & n.46 (summarizing commentators who have called for increased use of experts on structural discrimination). However, it seems unclear how an expert might prove discrimination in a particular decision absent some evidence specific to the decision, or at least specific to the decisionmaker. Such evidence would likely take one of the forms discussed in the text. And in any event, putting on expert testimony of this sort is expensive.
277. See, e.g., Sublett v. John Wiley & Sons, Inc., 463 F.3d 731, 740-41 (7th Cir. 2006) (holding that summary judgment was appropriate where comparators were not sufficiently similar to the plaintiff).
278. See Trans World Airlines, Inc. v. Thurston, 469 U.S. 111, 121-22 (1985) (stating that McDonnell Douglas is "designed to assure that the 'plaintiff [has] his day in court despite the unavailability of direct evidence'" (quoting Loeb v. Textron, Inc., 600 F.2d 1003, 1014 (1st Cir. 1979))).
279. Notably, despite the large number of critics of McDonnell Douglas, none has proposed a true alternative to this method of proof. Most critics have proposed adoption of one of the two alternative frameworks (Price Waterhouse or the 1991 Act). See, e.g., Hart, supra note 8, at 791. As this Article has shown, these frameworks are not mutually exclusive with McDonnell Douglas; they do not provide alternative methods of proof. See supra Part H.A. While some critics have proposed a more radical solution-a more aggressive burden-shifting mechanism that would require any employer who takes an adverse action against a female or minority worker to prove a lack of discrimination-that proposal is also not a method of proof. See RUTHERGLEN, supra note 3, at 42. So even this more radical solution is not an alternative to McDonnell Douglas. (Moreover, this radical solution has additional problems. While it might lead to more judgments against employers, it is contrary to the general philosophy of our legal system: innocent until proven guilty. Moreover, such burden-shifting schemes are generally justified based on the likelihood that the presumed fact (discrimination) is the actual state of affairs. See CHRISTOPHER B. MUELLER & LAIRD C. KIRKPATRICK, EVIDENCE 105 (3d ed. 2003) (noting that burdens of proof are often allocated according to the likelihood that a party will have access to particular facts). As noted above in Part III.C, the likelihood that any particular adverse employment action was the product of discrimination is a subject of considerable debate.)