BROWN v. NUCOR CORP. - Thomson Reutersstatic.legalsolutions.thomsonreuters.com/pdf/785_F3d_895.pdf · BROWN v. NUCOR CORP. 895 ... NUCOR CORPORATION; Nucor Steel Berkeley, Defendants

895BROWN v. NUCOR CORP.Cite as 785 F.3d 895 (4th Cir. 2015)

IV.

For the foregoing reasons, the judgmentof the district court is reversed and thecase is remanded for further proceedingsconsistent with this opinion.

REVERSED AND REMANDED

,

Quinton BROWN; Jason Guy; AlvinSimmons; Sheldon Singletary; GeraldWhite; Ramon Roane; Jacob Rave-nell, individually and on behalf of theclass they seek to represent, Plain-tiffs–Appellants,

v.

NUCOR CORPORATION; NucorSteel–Berkeley, Defendants–

Appellees.

No. 13–1779.

United States Court of Appeals,Fourth Circuit.

Argued: Sept. 17, 2014.

Decided: May 11, 2015.

Background: Black employees broughtclass action against employer under § 1981and Title VII alleging discriminatory jobpromotion practices and racially hostilework environment. The United States Dis-trict Court for the District of South Car-olina, C. Weston Houck, Senior DistrictJudge, decertified promotions class, andemployees appealed.

Holdings: The Court of Appeals, Gregory,Circuit Judge, held that:

(1) district court’s decision to reconsidercertification of class did not itself vio-late Court of Appeals’ mandate;

(2) employees satisfied commonality re-quirement for class certification;

(3) it was appropriate to treat plant atwhich employees worked as single enti-ty;

(4) employees did not waive argument thatdistrict court improperly determinedthat class certification was not war-ranted because individualized inquiriespredominated over common questions;and

(5) mandate rule precluded district courtfrom revisiting question of predomi-nance.

Vacated in part and remanded.

Agee, Circuit Judge, dissented and filedopinion.

1. Federal Courts O3585(3)Court of Appeals typically reviews dis-

trict court’s class certification order forabuse of discretion.

2. Federal Courts O3751Court of Appeals reviews de novo

whether district court contravenes priorexpress or implicit mandate issued byCourt of Appeals.

3. Federal Courts O3796Extraordinary exception to mandate

rule exists when there is showing thatcontrolling legal authority has changeddramatically.

4. Federal Courts O3796District court’s decision to reconsider

certification of class in employees’ actionalleging discriminatory job promotion

low nor properly briefed before us. Thus, wedecline to resolve it in the first instance. Thatsaid, if the court concludes that references to

the Order are indeed inadmissible, we believeredaction may be a prudent use of the districtcourt’s discretion.

896 785 FEDERAL REPORTER, 3d SERIES

practices did not itself violate Court ofAppeals’ mandate in prior appeal to certifyclass, in light of United States SupremeCourt’s subsequent decision in Wal–MartStores, Inc. v. Dukes, which provided suffi-ciently significant change in governing le-gal standard to permit limited reexamina-tion of whether class satisfied commonalityrequirement. Fed.Rules Civ.Proc.Rule23(a)(2), 28 U.S.C.A.

5. Federal Civil Procedure O174

Courts have no license to engage infree-ranging merits inquiries at class certi-fication stage; instead, claim’s merits maybe considered only when relevant to deter-mining whether prerequisites for class cer-tification are satisfied. Fed.Rules Civ.Proc.Rule 23, 28 U.S.C.A.

6. Federal Civil Procedure O184.10

Black employees satisfied commonali-ty requirement for class certification intheir Title VII action against employeralleging discriminatory job promotionpractices, where litigation concerned ap-proximately 100 class members in singlesteel plant, class members shared commonspaces, were in regular physical contactwith other departments, could apply forpromotions in other departments, andwere subject to same plant-wide policiesand practices, employees’ experts devel-oped reasonably reliable alternative bench-mark to extrapolate promotions data forperiod during which actual bidding infor-mation was unavailable, which showed sta-tistical disparity in promotions was statisti-cally significant at 2.54 standard deviationsfrom what would have been expected ifrace were neutral factor, and employeesprovided substantial evidence of unadulter-ated, consciously articulated, odious racismthroughout plant, including affirmative ac-tions by supervisors and widespread atti-tude of permissiveness of racial hostility.Civil Rights Act of 1964, § 703(k), 42

U.S.C.A. § 2000e–2(k); Fed.Rules Civ.Proc.Rule 23(a)(2), 28 U.S.C.A.


Plaintiffs must present common con-tention capable of being proven or dispro-ven in one stroke to satisfy commonalityrequirement for class certification. Fed.Rules Civ.Proc.Rule 23(a)(2), 28 U.S.C.A.


Commonality requirement for classcertification requires plaintiff to demon-strate that class members have sufferedsame injury. Fed.Rules Civ.Proc.Rule23(a)(2), 28 U.S.C.A.

9. Federal Civil Procedure O184.10

In determining whether black employ-ees satisfied commonality requirement forclass certification in their Title VII actionagainst employer alleging discriminatoryjob promotion practices, it was appropriateto treat plant at which they worked assingle entity, even though employees’ evi-dence of discrimination disproportionatelyconcerned single department, where de-clarants described frequent instances ofalleged promotions discrimination in otherdepartments, numerous complaints of dis-crimination were made to plant’s generalmanager, bidding for promotions wasplant-wide, and originating and destinationdepartment heads were both required toapprove promotions. Civil Rights Act of1964, § 703(k), 42 U.S.C.A. § 2000e–2(k);Fed.Rules Civ.Proc.Rule 23(a)(2), 28U.S.C.A.

10. Civil Rights O1140

Disparate impact liability under TitleVII requires identification of specific em-ployment practice that caused racially dis-parate results. Civil Rights Act of 1964,§ 703(k), 42 U.S.C.A. § 2000e–2(k).


11. Civil Rights O1140Unlike disparate treatment, disparate

impact theory does not require proof ofimproper intent to sustain Title VII viola-tion; instead, liability is premised on facial-ly neutral policies. Civil Rights Act of1964, § 703(k), 42 U.S.C.A. § 2000e–2(k).

12. Federal Civil Procedure O184.10To satisfy commonality requirement

for class certification in Title VII disparateimpact case, mere showing that policy ofdiscretion has produced an overall dispari-ty does not suffice; instead, plaintiffs whoallege such policy of discretion must dem-onstrate that common mode of exercisingdiscretion actually existed throughout com-pany. Fed.Rules Civ.Proc.Rule 23(a)(2),28 U.S.C.A.; Civil Rights Act of 1964,§ 703(k), 42 U.S.C.A. § 2000e–2(k).

13. Federal Courts O3733Party’s failure in its opening brief to

challenge alternate ground for districtcourt’s ruling waives that challenge.

14. Federal Courts O3733Plaintiffs did not waive argument that

district court improperly determined thatclass certification was not warranted be-cause individualized inquiries predominat-ed over common questions, even thoughtheir opening brief did not explicitly ad-dress predominance issue, where argu-ments in opening brief extended to districtcourt’s discussion of both predominanceand commonality. Fed.Rules Civ.Proc.Rules 23(a)(2), 23(b)(3), 28 U.S.C.A.;F.R.A.P.Rule 28(a)(8)(A), 28 U.S.C.A.

15. Federal Courts O3796Mandate rule precluded district court

from revisiting question of predominancein employees’ Title VII class action againstemployer alleging discriminatory job pro-motion practices, despite United StatesSupreme Court’s subsequent decision inWal–Mart Stores, Inc. v. Dukes, which

provided sufficiently significant change ingoverning legal standard to permit limitedreexamination of whether class satisfiedcommonality requirement, where Wal–Mart did not change, nor purport tochange, predominance analysis, Court ofAppeals had previously ruled that ‘‘puta-tive class satisfied both the predominanceand superiority requirements’’ for classcertification, and district court cited nonew facts or legal precedent to justifyrevisiting that determination once underly-ing question of commonality was resolved.Civil Rights Act of 1964, § 703(k), 42U.S.C.A. § 2000e–2(k); Fed.Rules Civ.Proc.Rules 23(a)(2), (b)(3), 28 U.S.C.A.

ARGUED: Robert L. Wiggins, Jr.,Wiggins, Childs, Quinn & Pantazis LLC,Birmingham, Alabama, for Appellants.Lisa Schiavo Blatt, Arnold & Porter LLP,Washington, D.C., for Appellees. ONBRIEF: Armand Derfner, D. Peters Wil-born, Jr., Derfner, Altman & Wilborn,Charleston, South Carolina; Ann K. Wig-gins, Wiggins, Childs, Quinn & PantazisLLC, Birmingham, Alabama, for Appel-lants. Cary A. Farris, John K. Linker, J.Shannon Gatlin, Alaniz Schraeder LinkerFarris Mayes, LLP, Houston, Texas; DirkC. Phillips, Sarah M. Harris, Arnold &Porter LLP, Washington, D.C.; J. TracyWalker, IV, Robert L. Hodges, MatthewA. Fitzgerald, McGuirewoods, LLP, Rich-mond, Virginia, for Appellees.

Before GREGORY, AGEE, andKEENAN, Circuit Judges.

Vacated in part, and remanded withinstructions by published opinion. JudgeGREGORY wrote the opinion, in whichJudge KEENAN joined. Judge AGEEwrote the dissenting opinion.


GREGORY, Circuit Judge:

This case concerns the certification of aclass of black steel workers who allegeendemic racial discrimination at a SouthCarolina plant owned by Nucor Corpora-tion and Nucor Steel Berkeley (collective-ly, ‘‘Nucor’’). Plaintiffs-appellants (‘‘theworkers’’) accuse Nucor of both discrimi-natory job promotion practices and a ra-cially hostile work environment under TitleVII of the Civil Rights Act of 1964 and 42U.S.C. § 1981. The district court original-ly denied class certification for both claims,and this Court reversed. See Brown v.Nucor Corp., 576 F.3d 149 (4th Cir.2009)(‘‘Brown I ’’).

The district court has revisited certifica-tion and decertified the promotions class inlight of the Supreme Court’s opinion inWal–Mart Stores, Inc. v. Dukes, ––– U.S.––––, 131 S.Ct. 2541, 180 L.Ed.2d 374(2011).1 We thus again confront the ques-tion of whether the workers’ have present-ed a common question of employment dis-crimination through evidence of racism inthe workplace. Despite Wal–Mart’s re-shaping of the class action landscape, wehold that the district court has for a sec-ond time erred in refusing to certify theworkers’ class, where (1) statistics indicatethat promotions at Nucor depended in parton whether an individual was black orwhite; (2) substantial anecdotal evidencesuggests discrimination in specific pro-motions decisions in multiple plant depart-ments; and (3) there is also significantevidence that those promotions decisionswere made in the context of a raciallyhostile work environment.

Against that backdrop, the district courtfundamentally misapprehended the reachof Wal–Mart and its application to theworkers’ promotions class. We thus va-cate the district court’s decision in partand remand for re-certification of theclass.

I.

The Nucor plant encompasses six pro-duction departments that work together tomelt, form, finish, and ship steel productsto customers. See Brown I, 576 F.3d at151. At the start of this litigation, 611employees worked at the plant. Seventy-one (11.62%) were black.2 There was,however, at most one black supervisor inthe production departments until after theEqual Employment Opportunity Commis-sion (‘‘EEOC’’) initiated charges that pre-ceded the putative class action.

The workers’ promotions claim rests onalternative theories of liability under TitleVII, which prohibits employment discrimi-nation because of an individual’s ‘‘race, col-or, religion, sex, or national origin.’’ 42U.S.C. § 2000e–2. The promotions claimfirst alleges a pattern or practice of racial-ly disparate treatment in promotions deci-sions. See Teamsters v. United States,431 U.S. 324, 336, 97 S.Ct. 1843, 52L.Ed.2d 396 (1977). Second, it chargesthat Nucor’s facially neutral promotionspolicies and procedures had a racially dis-parate impact. See Griggs v. Duke PowerCo., 401 U.S. 424, 431, 91 S.Ct. 849, 28L.Ed.2d 158 (1971); Wal–Mart, 131 S.Ct.at 2554.

Both theories are grounded in a statisti-cal analysis of racial disparities in job pro-

1. The district court refused to decertify theworkers’ hostile work environment claim.We have previously denied as untimely Nu-cor’s petition for interlocutory review of thatdecision. Nucor Corp. v. Brown, 760 F.3d341, 342 (4th Cir.2014).

2. By comparison, more than 38% of the avail-able local labor market is black, according toCensus data provided by the workers’ experts.


motions at the plant combined with anec-dotal evidence of discrimination. Theworkers’ statistical evidence spans thefour-year period preceding the litigation,between December 1999 and December2003. Because Nucor destroyed and/ordiscarded the actual bidding data for theperiod before 2001, the workers’ expertsestablished an alternative benchmark us-ing ‘change-of-status’ forms filed by thecompany whenever a promotion took placeat the plant. The experts extrapolatedcomparative statistics for that period usingan assumption that the racial compositionof the bidding pool for those jobs was thesame as for the post–2001 jobs analyzed(when Nucor retained actual bidding rec-ords).

The workers also presented abundantdirect and circumstantial anecdotal evi-dence of discrimination in promotions, in-cluding:

* Anecdotal evidence provided by theseven named plaintiffs and nine otherputative class members, claiming dis-crimination in specific promotions de-cisions in the Nucor production de-partments;

* A description of complaints, containedin affidavits and depositions, made toplant General Manager Ladd Hall,who the workers allege failed to mean-ingfully respond;

* Descriptions of retaliation againstthose who complained to manage-ment;

* A written copy of Nucor’s promotionspolicy and testimony that the policywas largely ignored in favor of givingunbridled discretion to supervisors;and

* Testimony by a white supervisor thathis department manager told him that‘‘I don’t think we’ll ever have a blacksupervisor while I’m here.’’

The facts undergirding the workers’ sep-arate hostile work environment claim, notdirectly at issue in this appeal, also bearon the promotions analysis. Those factsare disquieting in their volume, specificity,and consistency. Supervisors allegedlyroutinely referred to black workers as‘‘nigger’’ and ‘‘DAN (dumb ass nigger),’’with one supervisor reportedly stating‘‘niggers aren’t smart enough’’ to breakproduction records, while others toleratedthe routine use of epithets like ‘‘bolognalips,’’ ‘‘yard ape,’’ and ‘‘porch monkey.’’These epithets and others were broadcastover the plant-wide radio system—com-prising a network of walkie-talkies used tocommunicate—along with monkey noisesand the songs ‘‘Dixie’’ and ‘‘High Cotton.’’The workers’ declarations and depositionsfurther suggest that departmental supervi-sors and the plant’s general manager con-sistently ignored racial harassment carriedout by white workers, including the circu-lation of racist emails, the prominent dis-play of a hangman’s noose, the common-place showing of the Confederate flag, andan episode when a white employee drapeda white sheet over his head with eyes cutout in the form of a KKK hood.

In 2007, the South Carolina districtcourt denied the workers’ motion for classcertification for both the promotions andhostile work environment claims. In 2009,a divided panel of this Court reversed,concluding that the workers satisfied thethreshold requirements of Federal Rule ofCivil Procedure 23. We remanded thecase ‘‘with instructions to certify the appel-lants’ class action.’’ Brown I, 576 F.3d at160.

On February 17, 2011, the district courtfollowed our instructions to certify theclass, concluding that the workers satisfiedRule 23(b)(3)’s requirements that commonquestions predominate and that the classaction was superior to other litigation de-


vices to resolve the dispute. The districtcourt later declined to stay the case pend-ing a ruling in Wal–Mart, and it declinedto reconsider its order certifying the class.

The Supreme Court decided Wal–Martin June 2011, decertifying an unprecedent-ed nationwide class of approximately 1.5million female employees spread over3,400 stores. Wal–Mart held that theplaintiffs had failed to present a ‘‘commoncontention’’ of employment discriminationcapable of ‘‘classwide resolution,’’ as re-quired by Rule 23(a)(2). Wal–Mart, 131S.Ct. at 2551. Given the diffuse class andnumber of employment decisions at issue,the Supreme Court observed that ‘‘[w]ith-out some glue holding the alleged reasonsfor all those decisions together, it will beimpossible to say that examination of allclass members’ claims for relief will pro-duce a common answer to the crucial ques-tion why was I disfavored.’’ Id. at 2552(emphasis in original). The plaintiffs,Wal–Mart concluded, failed to meet thatstandard when they premised liability on acompany policy of decentralized subjectivedecision-making by local managers, com-bined with statistics showing gender-basedemployment disparities, limited anecdotalevidence, and expert testimony about acorporate culture that allowed for thetransmission of bias. See id. at 2551,2554–55.

On September 11, 2012, the districtcourt relied on Wal–Mart to decertify theworkers’ promotions class, invoking thecourt’s authority under Rule 23(c)(1)(C) toamend a certification order at any timebefore final judgment. Wal–Mart, thecourt observed, clarified and heightenedthe commonality requirement of Rule23(a)(2), requiring the workers to present‘‘significant proof’’ that Nucor ‘‘operatedunder a general policy of discrimination’’and that they suffered a common injury.

J.A. 10934 (quoting Wal–Mart, 131 S.Ct. at2553).

Under that standard, the district courtconcluded that decertification of the pro-motions class was required because: (1)this Court’s examination of the workers’statistical analysis in Brown I was notsufficiently ‘‘rigorous’’ to assess whether itraised questions common to the class un-der Rule 23(a)(2); (2) the workers’ statisti-cal and anecdotal evidence failed to estab-lish such commonality because it did notprovide ‘‘significant proof’’ that there exist-ed both a ‘‘general policy of discrimina-tion’’ and a ‘‘common injury’’; (3) the dele-gation of subjective decision-making toNucor supervisors was not, without more,a sufficiently uniform policy to present‘‘ ‘common’ issues appropriate for resolu-tion on a class-wide basis’’; and (4) even ifthe workers had identified a common ques-tion of law or fact satisfying Rule 23(a)(2),they failed to independently satisfy Rule23(b)(3)’s requirements that common is-sues predominate and that the class actionis a superior litigation device.

Although the court decertified the classfor the promotions claim, it refused to doso for the hostile work environment claim.The district court reaffirmed that theworkers had demonstrated that the ‘‘land-scape of the total work environment washostile towards the class.’’ J.A. 10964(quoting Newsome v. Up–To–Date Laun-dry, Inc., 219 F.R.D. 356, 362 (D.Md.2004)). Unlike the promotions claim, thecourt determined that the hostile environ-ment allegations required no showing of acompany-wide adherence to a common pol-icy of discrimination. Still, the court foundthat ‘‘there is significant evidence thatmanagement ignored a wide range ofharassment’’ and that the workers ‘‘mettheir burden to present significant proof ofa general policy of discrimination.’’ J.A.10968.


On September 30, 2013, the workers ap-pealed the district court’s decertification ofthe promotions class.

II.

[1, 2] We typically review a districtcourt’s certification order for abuse of dis-cretion. Doe v. Chao, 306 F.3d 170, 183(4th Cir.2002), aff’d on other grounds, 540U.S. 614, 124 S.Ct. 1204, 157 L.Ed.2d 1122(2004). We review de novo, however,whether a district court contravenes a pri-or express or implicit mandate issued bythis Court. United States v. Bell, 5 F.3d64, 66 (4th Cir.1993); S. Atl. Ltd. P’ship ofTenn. v. Riese, 356 F.3d 576, 583 (4thCir.2004) (‘‘We review de novo TTT wheth-er a post-mandate judgment of a districtcourt contravenes the mandate rule, orwhether the mandate has been ‘scrupu-lously and fully carried out.’ ’’ (quoting 2AFed. Proc., L.Ed. § 3:1016)).

Determining the appropriate standard ofreview thus requires a two step approach.First, we examine de novo whether thedistrict court’s decertification order violat-ed our mandate in Brown I to certify theworkers’ class. Second, if no such viola-tion occurred, we must determine anewwhether the district court abused its dis-cretion in decertifying the promotionsclass.

[3] As to the first question, an ‘‘ex-traordinary’’ exception to the mandate ruleexists when there is ‘‘a show[ing] that con-trolling legal authority has changed dra-matically.’’ Bell, 5 F.3d at 67 (alteration inoriginal). Moreover, Rule 23(c)(1)(C) pro-vides a district court with broad discretionto alter or amend a prior class certificationdecision at any time before final judgment.

Against that backdrop, the parties dis-agree about whether Wal–Mart providedsufficient justification for the district courtto invoke its powers to revisit certification.

Nucor maintains that Wal–Mart repre-sents a ‘‘sea change’’ and that ‘‘class ac-tions may proceed only in the most excep-tional of cases.’’ Resp’ts’ Br. 15, 20. Theworkers suggest, however, that the Su-preme Court instead largely reaffirmed ex-isting precedent. Appellants’ Br. 34.

The truth has settled somewhere in be-tween. See Scott v. Family Dollar Stores,Inc., 733 F.3d 105, 113–14 (4th Cir.2013)(discussing limitations on the scope ofWal–Mart’s holding); McReynolds v. Mer-rill Lynch, Pierce, Fenner & Smith, Inc.,672 F.3d 482, 487–88 (7th Cir.2012), cert.denied, ––– U.S. ––––, 133 S.Ct. 338, 184L.Ed.2d 157 (2012) (finding that Wal–Martprovided the basis for a renewed classcertification motion); DL v. District of Co-lumbia, 713 F.3d 120, 126 (D.C.Cir.2013)(surveying how Wal–Mart has changed theclass action landscape); Elizabeth Tippett,Robbing A Barren Vault: The Implica-tions of Dukes v. Wal–Mart for CasesChallenging Subjective Employment Prac-tices, 29 Hofstra Lab. & Emp. L.J. 433(2012) (using an empirical analysis to pre-dict Wal–Mart’s likely impact on class cer-tifications in the future). At the veryleast, Wal–Mart recalibrated and sharp-ened the lens through which a court exam-ines class certification decisions underRule 23(a)(2), an impact plainly manifestedby the number of certifications overturnedin its wake. See, e.g., EQT Prod. Co. v.Adair, 764 F.3d 347 (4th Cir.2014); Rodri-guez v. Nat’l City Bank, 726 F.3d 372, 376(3d Cir.2013); M.D. ex rel. Stukenberg v.Perry, 675 F.3d 832, 839, 841–44 (5th Cir.2012); Ellis v. Costco Wholesale Corp., 657F.3d 970, 974 (9th Cir.2011).

[4] In that light, we find that the dis-trict court’s decision to reconsider the cer-tification of the workers’ class did not itselfviolate our mandate in Brown I. Per thisCourt’s original remand instructions, thedistrict court certified both the promotions


and hostile work environment classes. Al-though the court had no discretion to thenreconsider questions decided by this Courtunder then-existing facts and law, Wal–Mart provided a sufficiently significantchange in the governing legal standard topermit a limited reexamination of whetherthe class satisfied the commonality re-quirement of Rule 23(a)(2).3 There are,however, instances described below whenthe district court unnecessarily revisitedother discrete determinations made by thisCourt in Brown I, such as whether theNucor plant should be treated analyticallyas a single entity, and whether the classindependently met the requirements ofRule 23(b)(3). The reconsideration ofthose determinations was not compelled byWal–Mart and contravened our mandatein Brown I.

Because the district court could reexam-ine whether the workers met the require-ment of commonality, we review thosefindings under the abuse of discretionstandard that typically applies to certifica-tion orders. See Amchem Prods., Inc. v.Windsor, 521 U.S. 591, 630, 117 S.Ct. 2231,138 L.Ed.2d 689 (1997) (‘‘The law givesbroad leeway to district courts in makingclass certification decisions, and their judg-ments are to be reviewed by the court ofappeals only for abuse of discretion.’’);Brown I, 576 F.3d at 152; Thorn v. Jeffer-son–Pilot Life Ins. Co., 445 F.3d 311, 317(4th Cir.2006). A district court abuses its

discretion when it materially misappliesthe requirements of Rule 23. See Gun-nells v. Healthplan Servs., Inc., 348 F.3d417, 424 (4th Cir.2003); Thorn, 445 F.3d at317–18 (‘‘A district court per se abuses itsdiscretion when it makes an error of lawor clearly errs in its factual findings.’’).The decisive question here is whether thedistrict court materially misapplied Rule23(a)(2) to the facts at hand in light ofWal–Mart.4

III.

Rule 23(a)(2) establishes that a class ac-tion may be maintained only if ‘‘there arequestions of law or fact common to theclass.’’ The district court determined thatWal–Mart required decertification of theworkers’ promotions class insofar as theSupreme Court’s interpretation of the rule(1) emphasized the analytical rigor re-quired to evaluate a plaintiff’s statisticalevidence of commonality at the class certi-fication stage, (2) placed the burden onplaintiffs to provide ‘‘significant proof’’ of a‘‘general policy of discrimination’’ and‘‘common injury,’’ and (3) relatedly estab-lished that a company’s policy of discre-tionary decision-making cannot sustainclass certification without a showing thatsupervisors exercised their discretion in acommon way.

Each of these arguments is consideredin turn.

3. Furthermore, this Court’s original mandatedid not entirely divest the district court of itsongoing authority under Rule 23(c)(1)(C) tomonitor the class and make changes whenappropriate. See Prado–Steiman v. Bush, 221F.3d 1266, 1273 (11th Cir.2000) (‘‘Class certi-fication orders TTT are not final judgmentsimpervious to lower court review and revi-sion.’’); Gene & Gene, L.L.C. v. BioPay, L.L.C.,624 F.3d 698, 702–03 (5th Cir.2010).

4. The dissent is skeptical that an appellatecourt can articulate a deferential standard of

review while then finding reversible error inmany of the factual and legal determinationsmade by a district court. See post at 930.Deference, however, clearly does not excuseus from conducting a detailed review of therecord. Nor does it blind us from factualfindings that were not supported and legaldeterminations that represent a fundamentalmisunderstanding of Wal–Mart’s scope. In-deed, we recently applied similar scrutinywhen overturning a district court’s class certi-fication order. See EQT Production, 764 F.3dat 357–58.


A.

[5] Wal–Mart reaffirmed existingprecedent that courts must rigorously ex-amine whether plaintiffs have met the pre-requisites of Rule 23(a) at the certificationstage, an analysis that will often overlapwith the merits of a claim. Wal–Mart, 131S.Ct. at 2551 (citing Gen. Tel. Co. of Sw. v.Falcon, 457 U.S. 147, 160–61, 102 S.Ct.2364, 72 L.Ed.2d 740 (1982)). But as theCourt later clarified, ‘‘Rule 23 grantscourts no license to engage in free-rangingmerits inquiries at the certification stage.’’Amgen Inc. v. Conn. Ret. Plans & TrustFunds, ––– U.S. ––––, 133 S.Ct. 1184,1194–95, 185 L.Ed.2d 308 (2013). Instead,the merits of a claim may be consideredonly when ‘‘relevant to determining wheth-er the Rule 23 prerequisites for class certi-fication are satisfied.’’ Id. at 1195.5

[6] This Court’s precedent and its ap-proach in Brown I are consistent withWal–Mart and Falcon. See Gariety v.Grant Thornton, LLP, 368 F.3d 356, 366(4th Cir.2004) (observing that ‘‘while anevaluation of the merits to determine thestrength of the plaintiffs’ case is not partof a Rule 23 analysis, the factors spelledout in Rule 23 must be addressed throughfindings, even if they overlap with issueson the merits’’). In Brown I, this Courtexpressly invoked Falcon’s requirement ofa rigorous analysis to determine compli-ance with Rule 23. 576 F.3d at 152. Moreimportant, of course, we actually conduct-

ed such an analysis, providing a detailedevaluation of the workers’ anecdotal andstatistical evidence to ensure that it pre-sented a common question under Rule23(a)(2). Id. at 153–56.

Contrary to the dissent’s assertion, wedo not (and Brown I did not) suggest thatRule 23 is a mere pleading standard. Seepost at 933. Far from it. It is true thatBrown I cautioned that ‘‘an in-depth as-sessment of the merits of appellants’ claimat this stage would be improper.’’ Id. at156. Such a statement, however, is consis-tent with the Supreme Court’s dictate inAmgen that a court should engage themerits of a claim only to the extent neces-sary to verify that Rule 23 has been satis-fied. Amgen, 133 S.Ct. at 1194–95.Brown I did precisely that.

1.

Even evaluated in a still more painstak-ing manner, the workers’ statistical evi-dence is methodologically sound whileyielding results that satisfy Wal–Mart’sheightened requirement of commonalitydiscussed below. The parties’ central dis-pute concerns the data used to analyze theperiod from December 1999 to January2001, when Nucor failed to retain actualbidding records. For that period, theworkers’ expert developed an alternativebenchmark that uses 27 relevant ‘change-of-status’ forms—filled out when an em-

5. The Wal–Mart majority confronted a splitamong courts regarding the depth of reviewnecessary to sustain class certification underRule 23. See Dukes v. Wal–Mart Stores, Inc.,603 F.3d 571, 582–84 (9th Cir.2010),rev’d, ––– U.S. ––––, 131 S.Ct. 2541, 180L.Ed.2d 374 (2011) (describing the split be-tween circuits); Wal–Mart, 131 S.Ct. at 2551–52. On one end of the spectrum, a number ofcourts liberally construed the SupremeCourt’s language in Eisen v. Carlisle & Jac-quelin, 417 U.S. 156, 94 S.Ct. 2140, 40L.Ed.2d 732 (1974), stating that ‘‘nothing in

either the language or history of Rule 23 TTT

gives a court any authority to conduct a pre-liminary inquiry into the merits of a suit inorder to determine whether it may be main-tained as a class action.’’ 417 U.S. at 177, 94S.Ct. 2140. On the other end, many courts,including this Circuit, heeded the SupremeCourt’s later call for a ‘‘rigorous analysis,’’ asannounced in Falcon. See 457 U.S. at 160,102 S.Ct. 2364. As Falcon held, ‘‘sometimesit may be necessary for the court to probebehind the pleadings before coming to rest onthe certification question.’’ Id.


ployee changes positions at the plant—toextrapolate promotions data because actualbidding information was unavailable.

Of course, it belabors the obvious toobserve that the alternative benchmark isa less precise measure than actual biddingdata. It is also clear, however, that plain-tiffs may rely on other reliable datasources and estimates when a company hasdestroyed or discarded the primary evi-dence in a discrimination case. More thantwo decades of this Court’s precedent af-firm as much. See Lewis v. BloomsburgMills, Inc., 773 F.2d 561 (4th Cir.1985);United States v. County of Fairfax, 629F.2d 932, 940 (4th Cir.1980); see generallyRamona L. Paetzold & Steven L. Willborn,The Statistics of Discrimination: UsingStatistical Evidence in DiscriminationCases § 4.03 (2014) (describing the use ofproxy data when actual data is unavailableor unreliable). In Lewis v. BloomsburgMills, Inc., this Court approved the use ofCensus data to establish a hypotheticalavailable pool of black female job appli-cants after a company discarded employ-ment applications for the relevant period.773 F.2d at 568.6 Plaintiffs then comparedthe ‘‘observed’’ annual rate of hires ofblack women with the ‘‘expected’’ ratesbased upon the proportional availability ofblack females in the labor pool. Id. Weendorsed a similar use of proxy data inUnited States v. County of Fairfax, involv-ing a county government that had de-

stroyed three years of employment appli-cations. 629 F.2d at 940. To analyzehiring during that time, plaintiffs assumedthat the proportion of black and womenapplicants for those years was the same asin the first year for which the countyretained records. Id. This Court ap-proved, concluding the alternative bench-mark was ‘‘the most salient proof of theCounty’s labor market.’’ Id.7

2.

The critical question is thus not whetherthe data used is perfect but instead wheth-er it is reliable and probative of discrimi-nation. To that end, a court must examinewhether any statistical assumptions madein the analysis are reasonable. See Paet-zold & Willborn, supra, § 4.16. The districtcourt here identified two assumptionsmade by the workers’ experts as proble-matic.

The district court first questioned theassumption that the job changes describedon the 27 forms represent promotions.See J.A. 10942. As an example of clearfactual error committed by the court, itquoted at length from the dissent inBrown I to argue that the forms mayrepresent job changes unrelated to pro-motions. J.A. 10942 (quoting Brown I, 576F.3d at 167–68 (Agee, J., concurring inpart and dissenting in part)). The formscited in Judge Agee’s original dissent,however, are plainly not among the 27

6. In Lewis, the company had ‘‘improperly dis-posed’’ of the relevant employment applica-tions, unlike the present case where there isno direct evidence of any impropriety. 773F.2d at 768. That fact, however, does notaffect our analysis of the workers’ alternativebenchmark.

7. The dissent cites Allen v. Prince George’sCounty, 737 F.2d 1299, 1306 (4th Cir.1984),to support its argument that a court has widediscretion to reject alternative benchmarks.Post at 940. In Allen, however, the defen-

dants produced actual ‘‘applicant flow data’’that contradicted the conclusions of the plain-tiffs’ statistics that were based on more gener-al workforce/labor market comparisons. Al-len, 737 F.2d at 1306. Here, like in Lewis,such actual applicant data is unavailable. SeeLewis, 773 F.2d at 568 (noting that ‘‘applicantflow data’’ was not available). Furthermore,Nucor has not presented any alternative sta-tistical study, or shown that data exists thatmay be more reliable than the alternativebenchmark used by the workers.


relied upon by the workers’ experts inconstructing the alternative benchmark.Compare J.A. 10942 (the district court’sdecertification order quoting the dissent inBrown I ), with J.A. 11005–11032 (copies ofthe actual change-of-status forms used inthe expert analysis). Worse still, the dis-sent in Brown I reached the question ofwhether the 27 forms represented pro-motions without the issue having beenraised, much less analyzed, by the districtcourt in its original order denying certifi-cation, see J.A. 8979, or by Nucor itself inits briefing before this Court in Brown I.8

The dissent in Brown I thus both engagedin sua sponte fact-finding to divine whichforms were used, and then got the factswrong.9 Using the flawed data, the dis-sent concluded in Brown I that ‘‘[o]n thisrecord, it is difficult, if not impossible todiscern whether the 2000 data based onthe nebulous change-of-status formsproves those positions were promotion po-sitions available for employee bidding andthus relevant to the formulation of statisti-cal evidence for the appellants’ claims.’’Brown I, 576 F.3d at 168 (Agee, J., concur-ring in part and dissenting in part). Thedistrict court expressly embraced that con-clusion in decertifying the promotions classafter Wal–Mart. J.A. 10942.

Upon examining the correct change-of-status forms, discerning whether they rep-resent promotions is a relatively straight-forward enterprise. Nineteen of the 27

forms expressly state they are for a pro-motion, for a ‘‘successful bidder’’ on a‘‘higher position,’’ or for a new positionthat was ‘‘awarded’’ or ‘‘earned.’’ Two ofthe forms describe changes in job classifi-cation accompanied by an increase in pay.One form notes that an inspector was a‘‘successful bidder’’ on a mill adjusterjob—a move referred to on another changeform as a promotion. Two forms are for a‘‘successful bidder’’ on a new positionwhere no new pay grade is noted. Theremaining three forms appear to involvechanges in positions or training that in-volved a decrease in pay, but there is noindication, or argument by Nucor or thedistrict court, that the exclusion of thoseforms would substantially undermine theprobativeness of the expert analysis.

The second assumption criticized by thedistrict court was that the bidding poolsfor the 27 positions filled between Decem-ber 1999 and January 2001 had the sameaverage racial composition as the pools forsimilar jobs analyzed from 2001 to Decem-ber 2003, when the company retained actu-al bidding data. Because of discovery lim-itations imposed by the district court, theinformation available regarding the 2001–2003 promotions was restricted to posi-tions similar to ones bid on by the namedplaintiffs, where there was at least oneblack bidder. However, because Nucorfailed to retain bidding records for 1999–2000, the data from that period could not

8. Nucor instead argued that the change ofstatus forms failed to capture whether blackemployees bid on the positions, and whetherthe positions were open for bidding in thefirst place. Given the lack of controversysurrounding whether the 27 forms describedpromotions, the forms themselves were notintroduced into the record until 2012, afterthe district embraced the fact-finding con-ducted by the dissent in Brown I and observedthat ‘‘the Court has never seen the 27 change-of-status formsTTTT’’ J.A. 10943. The workersthen appended all the forms to their motion to

‘‘alter and amend’’ the decertification order—a motion that was denied. J.A. 11005, 11083.Notably, it also appears that in 2006 theworkers’ expert provided Nucor with a list ofthe 27 employees used in the benchmarkanalysis. See J.A. 1409, 1438.

9. Given that history, we would be remiss notto acknowledge the irony inherent in the dis-sent’s insistence that we are now impermissi-bly making factual determinations withoutdue deference to the district court.


be limited to positions where there was aknown black bidder. Instead, the alterna-tive benchmark had to assume that therewas at least one black worker applying foreach promotion analyzed—an assumptionthat the district court concluded helpedrender the statistical analysis unreliable.But as we already determined in Brown I,the assumption does not fatally underminethe probativeness of the experts’ findings.The workers’ experts limited the recordsthey analyzed to the same positions identi-fied in the later period when bidding datawas available, positions for which therewas a black bidder. J.A. 1161–62. In itsoriginal order denying certification, thedistrict court observed that the assump-tions regarding bidding ‘‘may be reason-able and the statistics based thereon maybe relevant to prove discrimination at theplant,’’ but ‘‘the necessity of the assump-tions diminishes their probative value.’’ 10

J.A. 8987; see also Brown I, 576 F.3d at156. As we previously concluded, an in-cremental reduction in probative value—which is a natural consequence of the useof proxy data—does not itself render astatistical study unreliable in establishinga question of discrimination common to theclass. Brown I, 576 F.3d at 156. Indeed,to conclude otherwise would undermineour prior precedent in cases like Lewisand Fairfax, rendering plaintiffs unable tobring a statistics-based employment dis-crimination claim after a company has in-tentionally or inadvertently destroyed ac-tual applicant data.11 See Lewis, 773 F.2dat 568; Fairfax, 629 F.2d at 940.

3.

The dissent points to still more statisti-cal assumptions—assumptions not dis-cussed by either the district court or Nu-cor—to further question the reliability ofthe alternative benchmark. Specifically,the dissent suggests that the black work-ers may not have been qualified for higherpaying jobs and that they may have beendenied promotions because of disciplinaryrecords that were not themselves the re-sult of racial animus. See post at 940, 941–43. As to the qualifications of the work-ers, Nucor identifies nothing in the rec-ord—or in any factual findings by the dis-trict court—to suggest that black workersregularly applied for jobs for which theywere not qualified, such that the reliabilityof the study would be compromised. In-deed, the Nucor job postings explicitly list-ed the minimum qualifications required,and the workers’ experts reasonably as-sumed that individuals would normally ap-ply only if they believe they met suchqualifications. See J.A. 7763 (an exampleof a job posting); J.A. 1162. That is not tosay that patently unqualified workers didnot apply in isolated cases. But there isno reason to believe that such incidentswould have substantially reduced the relia-bility of the statistical conclusions. It alsobears repeating that it was Nucor thatfailed to retain or produce records thatwould have allowed the experts to takeother variables like qualifications moreprecisely into account. See J.A. 1165.

10. After we pointed to this language in BrownI, the district court did an about-face andchanged its conclusion to state that the statis-tics were ‘‘fundamentally unreliable.’’ J.A.10941.

11. The workers’ experts acknowledged thatthe incomplete data ‘‘undermined’’ their‘‘ability to use posting and bidding records toanalyze [those] promotions.’’ J.A. 1161. In

context, however, the experts were lamentingthe failure of Nucor to ‘‘produce all suchrecords.’’ J.A. 1161. As the experts conclud-ed, they were able to ‘‘calculate reliable statis-tics’’ for the limited universe of positions theyanalyzed, even though greater discoverywould have allowed them to make a more‘‘powerful’’ study of plantwide disparities.J.A. 1253–54; see also J.A. 1340–41.


The dissent, however, goes a step fur-ther in speculating that black workers mayhave been denied promotions because oftheir disciplinary records. See post at 940.Again, Nucor itself does not make thisargument. Instead, the argument the dis-sent constructs is based on the company’sself-serving responses to the workers’ in-terrogatories and requests for produc-tion—where Nucor asserts that some ofthe black workers were not chosen forpromotions due to disciplinary issues. Therecord, however, does not include disciplin-ary records for the named plaintiffs orputative class members. More fundamen-tal, the workers allege that any dispropor-tionate disciplinary action levied againstthem was itself a product of racial discrim-ination, with the disciplinary records thenused as a pretext in hiring decisions. Asworker Ramon Roane has stated:

Discipline, attendance, and safety allega-tions are similar factors that are notequally applied and that have been usedas an excuse to deny promotions to meand other persons of my race. Theattitudes I have experienced with whitesupervisors lead me to believe that myrace and that of other black employeesmakes a difference in how we are treat-ed and viewed for discipline[,] pro-motions[,] and training.

J.A. 1000; see also J.A. 1024 (AlvinSimmons’s statement that a white em-ployee was promoted over him despitethe fact that the white employee ‘‘hadbeen disciplined less than a year earlierfor ‘not paying attention’ when operatingequipment’’); J.A. 1111 (Earl Ravenell’s

statement that black workers were dis-proportionately singled out for disciplin-ary action); J.A. 6783 (Michael Rhode’sdescription of discrimination in disciplin-ary action). See generally J.A. 10960–10972 (the district court’s factual find-ings regarding the existence of a raciallyhostile work environment); Desert Pal-ace, Inc. v. Costa, 539 U.S. 90, 101–02,123 S.Ct. 2148, 156 L.Ed.2d 84 (2003)(allowing the use of circumstantial evi-dence to show that race was a motivat-ing factor in a ‘‘mixed-motive’’ case in-volving both legitimate and illegitimatereasons for an employment decision);Rowland v. Am. Gen. Fin., Inc., 340F.3d 187, 193–94 (4th Cir.2003) (allowingthe use of circumstantial evidence toshow that gender was ‘‘a motivating fac-tor’’ in a failure to promote an employ-ee). Given that background, it is easyto see why the district court chose notto advance the arguments that the dis-sent makes today.

Finally, the dissent criticizes the as-sumption that the 27 positions identifiedwere actually open for bidding.12 Post at939–40. That assumption, however, de-rives directly from Nucor’s stated policythat every job vacancy is posted on plantbulletin boards and is open to biddingplant-wide—a policy cited by Nucor’s ownexpert and the district court. See J.A.5887 (the Report of Finis Welch, observingthat ‘‘[o]pen positions are posted on bulle-tin boards and through email,’’ and that‘‘[a]ll employees in the plant are eligible tobid on a posted job’’); see also Resp’ts’ Br.9 (‘‘Department managers set the processin motion by sending postings for available

12. At times, the dissent seems to suggest thatstatistical assumptions themselves are to beviewed with great suspicion. What matters,however, is not whether an analysis makesassumptions based on imperfect data, butwhether those assumptions are reasonable.Indeed, statistics are not certainties but aremerely ‘‘a body of methods for making wise

decisions in the face of uncertainty.’’ W. Al-len Wallis & Harry V. Roberts, The Nature ofStatistics 11 (4th ed.2014); see also M.J. Mo-roney, Facts from Figures 3 (1951) (‘‘A statisti-cal analysis, properly conducted, is a delicatedissection of uncertainties, a surgery of sup-positions.’’).


promotions to Personnel employees, whoperformed a purely clerical role and adver-tised postings plantwide.’’); J.A. 8979 (thedistrict court’s original order denying cer-tification, finding that ‘‘[w]hen a position ina department becomes available, the job isposted on the plant’s e-mail system, whichis accessible to all employees in theplant’’). The dissent nonetheless arguesthat the statistical assumption was unrea-sonable.13 We disagree.

4.

With the alternative benchmark evi-dence included, the statistical disparity inpromotions is statistically significant at2.54 standard deviations from what wouldbe expected if race were a neutral factor.See Hazelwood Sch. Dist. v. United States,433 U.S. 299, 308 n. 14, 97 S.Ct. 2736, 53L.Ed.2d 768 (1977) (indicating that any-thing greater than two or three standarddeviations in racial discrimination cases issuspicious, at least for large sample sizes);Brown I, 576 F.3d at 156 n. 9 (applying theHazelwood standard to the workers’ statis-tical evidence); Jones v. City of Boston,752 F.3d 38, 46–47 (1st Cir.2014) (observ-ing that two standard deviations has be-come the commonly accepted threshold forsocial scientists and federal courts ‘‘in ana-lyzing statistical showings of disparate im-pact’’). According to the experts’ analysis,black employees constitute 19.24% of those

who applied for relevant promotions. Yetsuch employees were only 7.94% percentof those promoted.

Of course, statistical significance is notalways synonymous with legal significance.EEOC v. Fed. Reserve Bank of Richmond,698 F.2d 633, 648 (4th Cir.1983) rev’d onother grounds sub nom. Cooper v. Fed.Reserve Bank of Richmond, 467 U.S. 867,104 S.Ct. 2794, 81 L.Ed.2d 718 (1984).Indeed, the usefulness of statistical evi-dence often ‘‘depends on all of the sur-rounding facts and circumstances.’’Teamsters, 431 U.S. at 340, 97 S.Ct. 1843.Here, the surrounding circumstances andanecdotal evidence of discrimination, as de-scribed in greater detail below, are pre-cisely what help animate the statisticalfindings.14 As we held in Brown I andreaffirm today, ‘‘because the appellants’direct evidence alone was sufficient todemonstrate common claims of disparatetreatment and disparate impact, their sta-tistical data did not need to meet a two-standard-deviation threshold.’’ Brown I,576 F.3d at 156–57. Thus it is plain thatwhen the statistical disparity actually ex-ceeded two standard deviations, the dis-trict court abused its discretion in decerti-fying the class.

B.

[7] The district court further conclud-ed that the workers’ statistical and anecdo-

13. The record does indicate that ‘‘supervisorypositions’’ are not typically posted for biddingunder the Nucor hiring policy. J.A. 257.Neither Nucor nor the district court, howev-er, has provided any reason to believe thatany of the 27 records at issue describe opensupervisory jobs, as Nucor defined the term,and were thus not posted. Furthermore, thedissent suggests that there may have beenisolated instances when Nucor did not followits posting policy for non-supervisory jobs.The fact that a company does not follow itspolicy to a tee, however, does not fatally un-dermine a statistical assumption based uponsuch a policy.

14. Indeed, the workers’ statistical analysismay actually underestimate the impact ofrace on promotions at Nucor. As workerEric Conyers stated in his declaration: ‘‘If Ibelieved that a truly level playing field existedat the company I would have bid on numer-ous other positions such as Roll Guide Build-er in the Beam Mill.’’ J.A. 1079. But theexpert analysis at issue could not capture theimpact of discrimination on depressed bid-ding rates.


tal evidence was insufficient for class certi-fication insofar as the evidence did notdemonstrate a uniform class-wide injurythat spanned the entire Nucor plant. Asthe court observed, Wal–Mart instructsthat plaintiffs must present a common con-tention capable of being proven or dispro-ven in ‘‘one stroke’’ to satisfy Rule23(a)(2)’s commonality requirement. Wal–Mart, 131 S.Ct. at 2551. Thus, a class-wide proceeding must be able to generatecommon answers that drive the litigation.Id.; see also Jimenez v. Allstate Ins. Co.,765 F.3d 1161, 1165 (9th Cir.2014) (observ-ing that ‘‘a class meets Rule 23(a)(2)’s com-monality requirement when the commonquestions it has raised are apt to drive theresolution of the litigation, no matter theirnumber’’ (internal quotation marks omit-ted)). For a claim based on discriminationin employment decisions, ‘‘[w]ithout someglue holding the alleged reasons for allthose decisions together, it will be impossi-ble to say that examination of all the classmembers’ claims for relief will produce acommon answer to the crucial questionwhy was I disfavored.’’ Wal–Mart, 131S.Ct. at 2552 (emphasis omitted); see alsoScott v. Family Dollar Stores, Inc., 733F.3d 105, 113 (4th Cir.2013).

[8] The workers here most generallypresent two such common contentions ca-pable of class-wide answers under TitleVII. Under a disparate treatment theory,the common contention is that Nucor en-gaged in a pattern or practice of unlawfuldiscrimination against black workers inpromotions decisions. See Teamsters, 431U.S. at 336, 97 S.Ct. 1843. Under theworkers’ disparate impact theory, the com-mon contention is that a facially neutralpromotions policy resulted in a disparateracial impact. See Griggs, 401 U.S. at429–31, 91 S.Ct. 849. As Wal–Mart ob-served, however, semantic dexterity incrafting a common contention is not

enough. Commonality instead ‘‘requiresthe plaintiff to demonstrate that the classmembers ‘have suffered the same inju-ry[.]’ ’’ Wal–Mart, 131 S.Ct. at 2551 (quot-ing Falcon, 457 U.S. at 157, 102 S.Ct.2364). As such, a court must examinewhether differences between class mem-bers impede the discovery of common an-swers. Id. at 2551.

In the absence of a common job evalua-tion procedure, Wal–Mart held that statis-tical proof of employment discrimination atthe regional and national level, coupledwith limited anecdotal evidence from somestates, was insufficient to show that thecompany maintained a ‘‘general policy ofdiscrimination’’ present in each storewhere class members worked. See Wal–Mart, 131 S.Ct. at 2553. Similarly, thedistrict court here found that the workers’statistical and anecdotal evidence was in-sufficient to show a general policy in allNucor departments that caused the classinjury.

The district court, however, failed toadequately appreciate three significant dif-ferences from Wal–Mart that make thecase largely inapposite to the facts athand.

1.

First, Wal–Mart discounted the plain-tiffs’ statistical evidence in large part be-cause the statistics failed to show discrimi-nation on a store-by-store basis. See Wal–Mart, 131 S.Ct. at 2555. As such, theplaintiffs could not establish that a storegreeter in Northern California, for in-stance, was subject to the same discrimina-tion as a cashier in New Hampshire.These dissimilarities between class mem-bers were exacerbated by the sheer size ofthe Wal–Mart class—1.5 million membersworking at 3,400 stores under ‘‘a kaleido-scope of supervisors (male and female),subject to a variety of regional policies


that all differed.’’ Id. at 2557 (quotingDukes v. Wal–Mart Stores, Inc., 603 F.3d571, 652 (9th Cir.2010) (Kozinski, J., dis-senting)). The scale and scope of the pu-tative class, combined with the nature ofthe evidence offered, was thus essential toWal–Mart’s holding. Had the class beenlimited to a single Wal–Mart store span-ning multiple departments, or had theplaintiffs’ evidence captured discriminationat a store level, a very different Rule23(a)(2) analysis would have been required.

In contrast to Wal–Mart, this litigationconcerns approximately 100 class membersin a single steel plant in Huger, SouthCarolina. The class members shared com-mon spaces, were in regular physical con-tact with other departments, could applyfor promotions in other departments, andwere subject to hostile plant-wide policiesand practices. See Brown I, 576 F.3d at151. Such differences are not merely su-perficial. Instead, a more centralized, cir-cumscribed environment generally increas-es the uniformity of shared injuries, theconsistency with which managerial discre-tion is exercised, and the likelihood thatone manager’s promotions decisions willimpact employees in other departments.That is particularly the case where, asdiscussed further below, the entire Nucorplant was allegedly infected by expressracial bias and stereotypes—a culture thatmanagement took few affirmative steps tomeaningfully combat.

Nonetheless, the district court analo-gized to Wal–Mart in finding that theworkers’ evidence of discrimination wasinsufficient because it disproportionately

concerned a single department—the BeamMill—and because there was an insuffi-cient showing that all departments operat-ed under a common policy of discrimina-tion. J.A. 10949–54. As such, a class-wideproceeding would not generate ‘‘commonanswers’’ as Wal–Mart required, the dis-trict court found. See Wal–Mart, 131S.Ct. at 2551.

[9] The district court, however, inap-propriately discounted, and often ignored,evidence that establishes discrimination inother Nucor departments. Although 11 ofthe 16 employees submitting declarationson behalf of the plaintiffs worked in theBeam Mill, the declarants describe fre-quent instances of alleged promotions dis-crimination in other departments. SeeJ.A. 1021–24; 1032–35; 1049–51; 1055–56;1061–63; 1085–86; 1091–92; 1103; 1110–11; 1118–19. Even the additional affida-vits obtained by Nucor, discussed in fur-ther detail below, present numerous alle-gations of discrimination in non-Beam Milldepartments. See J.A. 5992–95 (discrimi-nation in the Hot Mill and Melt Shop);6143–45 (discrimination in the Hot Mill);6174 (general observations of promotionsdiscrimination); 6369–70 (discrimination inthe Melt Shop); 6505–07 (discrimination inthe Hot Mill); 7036 (discrimination in theMelt Shop). The record additionally indi-cates numerous complaints of discrimina-tion made to the plant’s general manager,who allegedly did little to nothing in re-sponse. Such alleged tolerance of discrim-ination from top management at the plantsupports the workers’ contention of aclass-wide injury that affected them all.15

15. As the district court found in the context ofthe workers’ hostile work environment claim:

These affidavits support the Court’s conclu-sion that although allegations of a hostilework environment were most prevalent andsevere in the Beam Mill, employees fromall of the production departments were

subjected to abusive behavior. Specifical-ly, employees from every department re-ported seeing the Confederate flag, employ-ees from every department reported seeingracist graffiti; and employees from everydepartment reported receiving racially of-fensive e-mails. Furthermore, in several


The district court made a still morefundamental error by choosing to treat theNucor departments as autonomous opera-tions in the first place instead of part of asingle facility, contravening both thisCourt’s instructions in Brown I and thedistrict court’s own prior findings. Thedistrict court’s original order to certify theclass recognized that a department-by-de-partment approach had been foreclosed,writing:

Since the Fourth Circuit rejected thisCourt’s characterization of the produc-tion departments as separate environ-ments, the Court must proceed underthe assumption that the production de-partments were permeable, if not uni-tary. This assumption is buttressed bythe fact that Nucor’s bidding is plant-wide, and this Court already has heldthat ‘‘potential applicants are eligible toprove they would have applied for apromotion but for the discriminatorypractice.’’

J.A. 9705. Wal–Mart provided no groundsfor the court to reconsider that findingbecause nothing in the Supreme Court’sopinion suggests that single, localized op-erations must be analytically dissected intocomponent departments.16 Here, all of theworkers’ evidence concerns a single con-nected facility.

Even if not required by our prior ruling,treating the plant as a single entity re-mains sound. In addition to the direct and

circumstantial evidence of discriminationin promotions decisions in multiple depart-ments, racial bias in one Nucor plant de-partment itself diminished the promotionalopportunities for black workers in all thedepartments—including those who wantedpromotions into the infected departmentand those who sought promotions to otherdepartments and needed their supervisors’recommendations. To that end, the work-ers cogently observe that requirements fordual approvals for promotions—by origi-nating and destination departmentheads—‘‘carr[ied] the effects of racial dis-crimination from one department and su-pervisor to another, either by systemictolerance, acquiescence or design.’’ Appel-lants’ Reply Br. 24 (citing Smith v. Bray,681 F.3d 888, 897 & n. 3 (7th Cir.2012)).

Such a conclusion is further strength-ened by the workers’ hostile work environ-ment claim. As the district court itselffound, ‘‘the plaintiffs have submitted sig-nificant proof that the landscape of thetotal work environment at the Berkeleyplant was hostile towards African–Ameri-cans and that the defendants failed to take‘remedial action reasonably calculated toend the harassment.’ ’’ J.A. 10966; see alsoBrown I, 576 F.3d at 157–58. That envi-ronment, the workers argue, supportstheir showing of an atmosphere of system-ic tolerance of racial hostility by managersand supervisors, forming part of the over-all pattern or practice that ‘‘infected black

instances, employees who worked in onedepartment indicated they were harassedby employees from other departments, andmany employees reported observing whatthey considered to be racist symbols andracist graffiti in common areas of the plant.

J.A. 10968.

16. The dissent insists that Brown I’s determi-nation that the Nucor plant should be treatedas a single facility only extended to the hostilework environment claim. Post at 945. Yet

the discussion of the issue in Brown I wasspecifically premised on the district court’sfindings regarding both the ‘‘pattern or prac-tice’’ and the work environment claims.Brown I, 576 F.3d at 157. A district courtmay not typically relitigate ‘‘issues expresslyor impliedly decided by the appellate court.’’Bell, 5 F.3d at 66. Here, even the districtcourt has recognized that Brown I preventeda finding that the plant was not a unitaryenvironment in the context of the promotionsclaim. J.A. 9705 (Certification Order).


employees’ promotion opportunities.’’ Weagree.

2.

Second, the Wal–Mart plaintiffs’ theoryof commonality relied, in part, on showingthat the company maintained a corporateculture that facilitated the uniform trans-mission of implicit, or subconscious, biasinto the hiring process. See Wal–Mart,131 S.Ct. at 2548. To that end, the plain-tiffs’ expert testified the company was‘‘vulnerable’’ to ‘‘gender bias.’’ Id. at 2553.The Court, however, concluded that theexpert could not with specificity determinehow the culture concretely influenced indi-vidual employment decisions. Id. at 2553–54. The testimony was therefore insuffi-cient to show a common policy that pro-duced a common injury.

Here, however, the workers have provid-ed substantial evidence of unadulterated,consciously articulated, odious racismthroughout the Nucor plant, including af-firmative actions by supervisors and awidespread attitude of permissiveness ofracial hostility. The examples in the rec-ord are ubiquitous: bigoted epithets andmonkey noises broadcast across the plantradio system, emails with highly offensiveimages sent to black workers, a hangman’snoose prominently displayed, a white su-pervisor stating that ‘‘niggers aren’t smartenough’’ to break production records, andabundant racist graffiti in locker roomsand shared spaces. Moreover, no morethan one black supervisor worked in theNucor production departments until afterthe EEOC charge that preceded this liti-gation. It strains the intellect to posit anequitable promotions system set against

that cultural backdrop, particularly in lightof the other evidence presented.

The dissent rejects the idea that evi-dence of a racially hostile work environ-ment may help establish a claim for dispa-rate treatment in promotions decisions.17

Post at 945–46. Indeed, the dissent goesso far as to observe that ‘‘locker rooms andradios bear no relationship to promotionsdecisions.’’ Id. at 946. Such a perspec-tive, however, is perplexingly divorcedfrom reality and the history of workplacediscrimination. It is difficult to fathomhow widespread racial animus of the typealleged here, an animus that consistentlyemphasized the inferiority of black work-ers, bears no relationship to decisionswhether or not to promote an employee ofthat race. Although the dissent assertsthat ‘‘nothing in the record supports’’ mak-ing a connection between the work envi-ronment and promotions practices, we arenot limited to the record in making suchelementary judgments. Justice is notblind to history, and we need not avert oureyes from the broader circumstances sur-rounding employment decisions, and theinferences that naturally follow.

3.

Third, and related, the anecdotal evi-dence of discrimination in this case is sub-stantially more probative than that inWal–Mart. The Wal–Mart plaintiffs pre-sented affidavits from about 120 femaleemployees, representing approximatelyone affidavit for every 12,500 class mem-bers. Wal–Mart, 131 S.Ct. at 2556. Theaffidavits captured only 235 of Wal–Mart’s3,400 stores, and there were no affidavits

17. We do not suggest, of course, that evidenceof a hostile work environment is sufficient byitself to support a disparate treatment or dis-parate impact claim. Rather, we merely ob-serve that the substantial showing of endemic

prejudice at the plant—a prejudice that wasallegedly tolerated and/or encouraged bymanagement—heightens the probativeness ofthe workers’ other evidence.


from workers in 14 states. Id. The evi-dence thus fell far short of the benchmarkfor a showing of company-wide discrimina-tion established by Teamsters, 431 U.S.324, 97 S.Ct. 1843. In Teamsters, theplaintiffs produced statistical evidence ofracial bias combined with approximately 40accounts of discrimination from particularindividuals. Id. at 338, 97 S.Ct. 1843.Given the class size of approximately 334persons, there was roughly one anecdotefor every eight members of the class. Seeid. at 331, 338, 97 S.Ct. 1843; Wal–Mart,131 S.Ct. at 2556. ‘‘[T]he anecdotes camefrom individuals spread throughout thecompany who for the most part worked atthe company’s operational centers that em-ployed the largest numbers of the classmembers.’’ See Wal–Mart, 131 S.Ct. at2556 (internal quotation marks omitted).Similarly, this litigation includes anecdotal

evidence from more than 16 individuals 18

in a class that numbered approximatelyone-hundred ‘‘past and present black em-ployees at the plant’’ at the time litigationcommenced—an approximate ratio of oneanecdote for every 6.25 class members.19

See Brown I, 576 F.3d at 151 (describingthe class size).

Balanced against such evidence, the dis-trict court gave ‘‘limited weight’’ to ap-proximately 80 affidavits from Nucor em-ployees largely disclaiming discriminationat the plant—affidavits taken by companylawyers after the EEOC charges had beenfiled. See J.A. 10950–51. Common senseand prudence, however, instruct that theaffidavits do little to rebut the evidence ofdiscrimination insofar as they were givenunder potentially coercive circumstances,where the company reserved its ability to

18. This number includes both the 16 declara-tions introduced by the workers and otheraccounts of discrimination included in affida-vits obtained by Nucor after the EEOC chargewas filed. See, e.g., J.A. 5992–95, 6143–45,6174, 6369–70, 6505–07, 7036. Of the 16worker-filed declarations, Byron Turner’sstatement fails to mention specific instancesof promotions discrimination, but instead af-firms that that he was ‘‘affected by the samepractices that Ramon Roane and the othernamed plaintiffs’’ have raised. J.A. 1124.The dissent argues that the declaration ofWalter Cook also fails to mention promotions.Post at 949. Cook’s declaration, however,states that he heard white employees talkingabout a black worker’s application for anOperator position. According to Cook, theemployees stated they would ‘‘do everythingthat they could to make sure that niggerdidn’t get the job.’’ J.A. 1075. Further, thedissent argues that the declaration from Ken-neth Hubbard includes a complaint that Nu-cor in fact promoted him. Post at 949. Hub-bard’s declaration, however, accuses Nucor ofplacing him ‘‘in the position to get [him] outof the mill and the line of progression thatlead to supervisory positions.’’ J.A. 1097.Hubbard also observes that his trajectory atthe company was dramatically different fromthat of a white co-worker who started at the

plant at the same time and later became asupervisor. Id. Indeed, the dissent’s ap-proach to the affidavits, consistent with itsapproach to the anecdotal evidence through-out, appears to be to cherry pick facts froman 11,000 page record, strip those facts ofcontext, and then argue that they underminethe substantial, credible evidence of discrimi-nation that the workers have produced.

19. There is some uncertainty about the pre-cise size of the class. At the time the litiga-tion began, seventy-one workers at the Nucorplant were black. Brown I, 576 F.3d at 151.As the district court found, there was a totalof ‘‘ninety-four black employees who workedat the plant from 2001 through 2004.’’ Id. at152. The workers’ experts estimated thatthere may have been about 150 black workersin total who ‘‘were potentially affected by theselection decisions regarding promotion atNucor–Berkeley.’’ J.A. 1154. Even assuminga class size of 150, there would be more thanone anecdotal account of racial discrimina-tion for every 9.38 class members, a ratio thatremains in line with the evidence in Team-sters. Furthermore, that number does nottake into account the descriptions of discrimi-nation in promotions decisions in the affida-vits that Nucor itself obtained, as previouslydescribed.


use them against other employees in anyfuture lawsuit (a fact that was omittedfrom the Statement of Participation givento affiants). See J.A. 6003 (the Statementof Participation), 9379 (Nucor’s statementthat it intended ‘‘to use the affidavits forevery purpose permitted under the Feder-al Rules of Evidence,’’ including the oppo-sition to class certification and the im-peachment of witnesses); see also Kleinerv. First Nat’l Bank of Atlanta, 751 F.2d1193, 1202 (11th Cir.1985) (observing thatafter a class action has been filed, ‘‘[a]unilateral communications scheme TTT isrife with potential for coercion’’); Quezadav. Schneider Logistics Transloading &Distrib., No. CV 12–2188 CAS, 2013 WL1296761, at *5 (C.D.Cal. Mar. 25, 2013)(finding in a class action context that‘‘[f]ailing to inform the employees of theevidence-gathering purpose of the inter-views rendered the communications funda-mentally misleading and deceptive becausethe employees were unaware that the in-terview was taking place in an adversarialcontext, and that the employees’ state-ments could be used to limit their right torelief’’); Longcrier v. HL–A Co., 595F.Supp.2d 1218, 1228 (S.D.Ala.2008); Me-vorah v. Wells Fargo Home Mort., Inc.,No. C 05–1175 MHP, 2005 WL 4813532, at*4 (N.D.Cal. Nov. 17, 2005). Of course,companies may investigate allegations ofdiscrimination and take statements fromemployees. But when it comes to assess-ing the probative value of those state-ments—when weighed against the numer-ous declarations of employees who tookthe often grave risk of accusing an employ-er of a workplace violation—courts shouldproceed with eyes open to the imbalance ofpower and competing interests.20 More-over, as previously observed, the company-

obtained affidavits still contain numerousallegations of discrimination in promotionsdecisions—allegations that carry signifi-cant weight given the circumstances inwhich they were made. See J.A. 5992–95,6143–46, 6174, 6370, 6506, 7036.

Of course, a plaintiff need not ‘‘offerevidence that each person for whom it willultimately seek relief was a victim of theemployer’s discriminatory policy.’’ Team-sters, 431 U.S. at 360, 97 S.Ct. 1843; seealso EEOC v. Korn Indus., Inc., 662 F.2d256, 260 (4th Cir.1981). Instead, a bifur-cated class action proceeding allows for a‘‘liability’’ stage to first determine whetheran employer engaged in a pattern or prac-tice of discriminatory conduct. Teamsters,431 U.S. at 360, 97 S.Ct. 1843; Korn, 662F.2d at 260. Upon a finding of liability, asecond damages stage allows for the con-sideration of which individuals were specif-ically harmed by the policy. Teamsters,431 U.S. at 361, 97 S.Ct. 1843; Korn, 662F.2d at 260.

4.

Here, for a liability determination in adisparate treatment claim, the workers’statistical and anecdotal evidence, especial-ly when combined, thus provide preciselythe ‘glue’ of commonality that Wal–Martdemands. See Brown I, 576 F.3d at 156.Such a claim requires proof of a ‘‘system-wide pattern or practice’’ of discriminationsuch that the discrimination is ‘‘the regularrather than the unusual practice.’’ Team-sters, 431 U.S. at 336, 97 S.Ct. 1843; Coo-per, 467 U.S. at 875–76, 104 S.Ct. 2794; seealso Wal–Mart, 131 S.Ct. at 2552 n. 7. Therequired discriminatory intent may be in-ferred upon such a showing. See Team-sters, 431 U.S. at 339–40, 97 S.Ct. 1843;

20. The dissent is thus mistaken when it as-serts that we are articulating a new rule thatcourts categorically may not consider the affi-davits obtained by companies as part of an

investigation into allegations of discrimina-tion. See post at 951–52. Instead, our analy-sis concerns the weight that should be givento such affidavits in these circumstances.


Hazelwood, 433 U.S. at 308–09, 97 S.Ct.2736 (observing that ‘‘[w]here gross statis-tical disparities can be shown, they alonemay in a proper case constitute prima facieproof of a pattern or practice of discrimi-nation’’).

Whereas there may have been manyanswers in Wal–Mart to the question ofwhy any individual employee was disfa-vored, the workers here have sufficientlyalleged that there is only one answer tothe question of why Nucor’s black workerswere consistently disfavored.21 Unlike adisparate impact claim, a showing of dispa-rate treatment does not require the identi-fication of a specific employment policyresponsible for the discrimination. SeeTeamsters, 431 U.S. at 336 n. 16, 97 S.Ct.1843 (discussing the legislative history ofTitle VII and concluding that the words‘‘pattern or practice’’ should be interpretedaccording to their plain meaning). A pat-tern of discrimination, revealed throughstatistics and anecdotal evidence, can alonesupport a disparate treatment claim, evenwhere the pattern is the result of discre-tionary decision-making.

To hold otherwise would dramaticallyundermine Title VII’s prophylactic powers.As the Supreme Court observed in Griggs,a central purpose of Title VII is ‘‘toachieve equality of employment opportuni-ties and remove barriers that have operat-ed in the past to favor an identifiablegroup of white employees over other em-ployees.’’ 401 U.S. at 429–30, 91 S.Ct. 849;see also Albemarle Paper Co. v. Moody,422 U.S. 405, 417–18, 95 S.Ct. 2362, 45L.Ed.2d 280 (1975) (stressing Title VII’sprophylactic goals in addition to its pur-pose ‘‘to make persons whole for injuriessuffered on account of unlawful employ-ment discrimination’’). Here, where sub-

stantial evidence suggests a pattern of en-grained discriminatory decision-makingthat consistently disadvantaged blackworkers at Nucor, to deny class certifica-tion would significantly weaken Title VIIas a bulwark against discrimination.

C.

[10, 11] Statistics and anecdotes sug-gesting a pattern of discrimination, howev-er, are not enough alone to sustain a dis-parate impact claim. See Wal–Mart, 131S.Ct. at 2555; Watson v. Fort Worth Bank& Trust, 487 U.S. 977, 994, 108 S.Ct. 2777,101 L.Ed.2d 827 (1988). Disparate impactliability requires the identification of a spe-cific employment practice that caused ra-cially disparate results. See 42 U.S.C.§ 2000e–2(k); Watson, 487 U.S. at 986–87,108 S.Ct. 2777; Griggs, 401 U.S. at 431, 91S.Ct. 849. Unlike disparate treatment, thedisparate impact theory does not requireproof of improper intent to sustain a TitleVII violation. Teamsters, 431 U.S. at 349,97 S.Ct. 1843; Griggs, 401 U.S. at 429–31,91 S.Ct. 849 (finding the use of standard-ized tests resulted in a disparate impact).Instead, liability is premised on faciallyneutral policies. Griggs, 401 U.S. at 431,91 S.Ct. 849.

[12] Under Wal–Mart, a mere showingthat a ‘‘policy of discretion has producedan overall TTT disparity does not suffice.’’Wal–Mart, 131 S.Ct. at 2556. Instead,plaintiffs who allege such a policy of dis-cretion must demonstrate that a ‘‘commonmode of exercising discretion’’ actually ex-isted throughout a company. Id. at 2554;see also Tabor v. Hilti, Inc., 703 F.3d 1206,1229 (10th Cir.2013) (observing that ‘‘afterWal–Mart, federal courts TTT have gener-ally denied certification when allegedly dis-

21. Contrary to the dissent’s assertion, we donot find ‘‘in the first instance’’ that the work-er’s allegation is correct. Instead, we con-

clude that the district court clearly erred infinding that the allegation was not sufficientlysupported by the record.


criminatory policies are highly discretion-ary and the plaintiffs do not point to acommon mode of exercising discretion thatpervades the entire company’’ (internalquotation marks omitted)). Given thatstandard, the district court here found thatthe workers ‘‘failed to identify any factorthat unites the manner in which the vari-ous decision makers throughout the Berke-ley plant exercised their discretion.’’ J.A.10955.

Wal–Mart recognizes that in certaincases, ‘‘giving discretion to lower-level su-pervisors can be the basis of Title VIIliability under a disparate-impact theory,’’131 S.Ct. at 2554, because ‘‘an employer’sundisciplined system of subjective deci-sionmaking [can have] precisely the sameeffects as a system pervaded by impermis-sible intentional discrimination.’’ Id. (al-teration in original) (quoting Watson, 487U.S. at 990, 108 S.Ct. 2777). For a nation-wide class, Wal–Mart found that proving aconsistent exercise of discretion will bedifficult, if not impossible in some circum-stances. Id.; see also Davis v. CintasCorp., 717 F.3d 476, 488 (6th Cir.2013)(noting the difficulties Wal–Mart presentsfor parties seeking to certify a nationwideclass).

But for a localized, circumscribed classof workers at a single facility, a policy ofsubjective, discretionary decision-makingcan more easily form the basis of Title VIIliability, particularly when paired with aclear showing of pervasive racial hostility.In such cases, the underlying animus mayhelp establish a consistently discriminatoryexercise of discretion.

This Court’s recent opinion in Scott v.Family Dollar Stores, Inc. specifically pro-vides several ways that such a disparateimpact claim may satisfy Rule 23 afterWal–Mart, including: (1) when the exer-cise of discretion is ‘‘tied to a specificemployment practice’’ that ‘‘affected the

class in a uniform manner’’; (2) whenthere is ‘‘also an allegation of a company-wide policy of discrimination’’ that affectedemployment decisions; and (3) ‘‘whenhigh-level personnel exercise’’ the discre-tion at issue. Scott, 733 F.3d at 113–14.

The first and second of Scott’s alterna-tives are most relevant to this case. Aspecific employment practice or policy cancomprise affirmative acts or inaction. Cf.Ellison v. Brady, 924 F.2d 872, 881 (9thCir.1991) (explaining an employer’s re-sponsibility to act to rectify a hostile oroffensive work environment under TitleVII). Regarding affirmative acts, the dis-trict court has established that Nucor’spromotions practice provides that ‘‘[e]m-ployees in each of the production depart-ments may bid on positions available inother departments,’’ and that in order topromote one of the bidders, ‘‘the supervi-sor, the department manager, and the gen-eral manager must approve a writtenchange of status and then submit thechange of status form to the personneloffice.’’ J.A. 477–78.

For purposes of class certification, theworkers have provided sufficient evidencethat such a policy, paired with the exerciseof discretion by supervisors acting withinit, created or exacerbated racially dispa-rate results. The promotions system, re-quiring approvals from different levels ofmanagement, created an environment inwhich the discriminatory exercise of dis-cretion by one department head harmedthe promotions opportunities for all blackworkers at the plant by foreclosing onopportunities in that department and gen-erally impeding upward mobility. More-over, the disproportionate promotions ofwhite workers had to be ratified by thegeneral manager, Ladd Hall, who was thuson notice, or should have been on notice,that there were pronounced racial dispari-ties in department-level promotion prac-


tices, as indicated by the statistical andanecdotal evidence presented.

The workers have also presented suffi-cient evidence of a practice of inaction bythe general manager who ignored the evi-dence of, and complaints regarding, dis-crimination in promotions at the plant.See, e.g., J.A. 996–97, 1016, 1056, 1087,1104. Such managerial inaction occurreddespite Nucor’s status as an ‘‘Equal Op-portunity Employer’’ and its claim to havea ‘‘plantwide policy barring racial discrimi-nation.’’ Resp’ts’ Br. 6. One black worker,Ray Roane, has testified that he com-plained directly to Hall about discrimina-tion in promotions. J.A. 996–97. Hallthreatened his job. J.A. 997. Consistentwith that evidence, the workers observe inthe context of their hostile work environ-ment claim that despite a policy of investi-gating complaints of racial harassment,‘‘[n]ot even one of the five departmentmanagers has been shown to have lifted afinger to redress the racially hostile workenvironment found to exist both plant-wideand in each department.’’ Appellants’ Br.25. The workers have sufficiently allegedthat such a uniform policy of managerialinaction also contributed to racial dispari-ties in promotions decisions.

Consistent with Scott, the workers havefurther demonstrated that the exercise ofdiscretion at Nucor was joined by ‘‘a com-pany-wide policy of discrimination’’ thatwas encouraged, or at least tolerated, bysupervisors and managers. See Scott, 733F.3d at 114. In addition to the evidence ofa hostile work environment previously de-scribed in detail, one white supervisor hasexpressly stated in a deposition that heheard the head of the Beam Mill declare,‘‘I don’t think we’ll ever have a black su-pervisor while I’m here.’’ J.A. 1885–86.Such facts provide a critical nexus betweenthe racial animus at the plant and pro-motions decisions that impacted all black

workers by foreclosing opportunities forthem. Or, using Wal–Mart’s language,the evidence of pervasive racial hostility inthe working environment provides a ‘‘com-mon mode of exercising discretion thatpervade[d] the entire company.’’ Wal–Mart, 131 S.Ct. at 2554–55.

In the end, Wal–Mart simply ‘‘found itunlikely’’ that thousands of managersacross different regions ‘‘would exercisetheir discretion in a common way withoutsome common direction.’’ Tabor, 703 F.3dat 1222. Here, however, the workers haveprovided ample evidence supporting theirallegation of a common, racially-biased ex-ercise of discretion throughout the plant—demonstrated through alleged incidents ofspecific discrimination in promotions deci-sions, statistical disparities, and facts sug-gesting pervasive plant-wide racism. Thedistrict court abused its discretion in find-ing that such evidence was insufficient tomeet the burden that Wal–Mart imposes.

IV.

Nucor further argues that the workershave failed to contest the district court’sindependent finding that the putative classfailed to satisfy Rule 23(b)(3). As thecompany observes, the district court spe-cifically held that the class failed to meetthe rule’s requirements for a class actionseeking individualized money damages,namely, that common questions predomi-nate over individualized inquiries and thatthe class action is ‘‘superior to other avail-able methods for fairly and efficiently ad-judicating the controversy.’’ Fed.R.Civ.P.23(b)(3). The court remarked that ‘‘even ifthe Fourth Circuit subsequently concludesthat the plaintiffs have identified a com-mon issue that satisfies Rule 23(a)(2), thisCourt nonetheless finds that ‘common is-sues,’ as that term is defined by Wal–Mart, do not predominate over individual


issues with regard to the plaintiffs’ pro-motions claims.’’ 22 J.A. 10956.

[13] Nucor contends that nowhere inthe workers’ opening brief is the Rule23(b)(3) ruling addressed, and that anychallenge to that decision has thus beenwaived. The doctrine of waiver derivesfrom the Federal Rules of Appellate Pro-cedure, which require that the argumentsection of an appellant’s opening brief con-tain the ‘‘appellant’s contentions and thereasons for them, with citations to theauthorities and parts of the record onwhich the appellant relies.’’ Fed. R.App.P. 28(a)(8)(A); see also Mayfield v. Nat’lAss’n for Stock Car Auto Racing, Inc., 674F.3d 369, 376–77 (4th Cir.2012). ‘‘Failureof a party in its opening brief to challengean alternate ground for a district court’sruling TTT waives that challenge.’’ UnitedStates ex rel. Ubl v. IIF Data Solutions,650 F.3d 445, 456 (4th Cir.2011) (quotingRodriguez v. Hayes, 591 F.3d 1105, 1118 n.6 (9th Cir.2010)).

[14] The workers contend first, and weagree, that no waiver occurred becausetheir arguments in the opening brief ex-tended to the district court’s discussion ofboth predominance and commonality. Thesingle issue identified by the workers onappeal did not differentiate between thecourt’s findings on either question. Theissue, as presented, was this:

Was it error or an abuse of discretionfor the district court not to follow thisCircuit’s mandate holding that sufficientstatistical and non-statistical evidencehas been presented to certify a pattern-or-practice and disparate impact classcovering all six production departments

of the defendants’ manufacturing plantin Huger, South Carolina?

Consistent with that framing, the workers’opening brief describes the district court’sdecision in equally broad terms withoutdistinguishing between commonality andpredominance. See Appellants’ Br. 28–29(‘‘The district court erred as a matter oflaw by declining to follow this Court’smandate that held there is sufficient statis-tical and non-statistical evidence to certifya class covering all six production depart-ments.’’); Appellants’ Br. 3 (citing to theportion of the district court opinion wherepredominance is discussed).

Although more explicit separation of thepredominance and commonality inquirieswould no doubt have been wise, the work-ers’ arguments throughout their brief di-rectly respond to the issues the districtcourt raised in both contexts (issues that,as discussed below, were intertwined bythe court). The workers, for instance,specifically cite cases discussing predomi-nance when arguing about the extent towhich a court may look to merits in decid-ing certification. See Appellants’ Br. 34–35. Elsewhere, in discussing the sufficien-cy of the anecdotal evidence presented, theworkers argued in favor of our holding inBrown I that ‘‘[t]his evidence alone estab-lishes common claims of discriminationworthy of class certification.’’ Appellants’Br. 42 (citing Brown I, 576 F.3d at 153).Certification of the workers’ class requireda finding that Rule 23(b) was satisfied, inaddition to a finding of commonality underRule 23(a)(2). More generally, withoutlimiting its analysis to the question of com-monality, the workers’ opening brief ob-

22. This Court has previously observed that‘‘[i]n a class action brought under Rule23(b)(3), the ‘commonality’ requirement ofRule 23(a)(2) is ‘subsumed under, or super-seded by, the more stringent Rule 23(b)(3)requirement that questions common to theclass predominate over’ other questions.’’

Lienhart v. Dryvit Sys., Inc., 255 F.3d 138, 146n. 4 (4th Cir.2001) (quoting Amchem, 521U.S. at 609, 117 S.Ct. 2231). But as Wal–Mart made clear, the Rule 23(a) commonalityrequirement and the Rule 23(b)(3) predomi-nance requirement remain separate inquiries.Wal–Mart, 131 S.Ct. at 2556.

u0034828

Highlight


serves that ‘‘[t]he district court’s findingthat there is no pattern-or-practice evi-dence in the non-Beam Mill departmentsis directly contrary to the evidence and[the Fourth Circuit’s] mandate.’’ Appel-lants’ Br. 42–43.

It is true that the workers argumentsoften focus expressly on the question ofcommonality, as Wal–Mart focused itsanalysis. In that regard, however, theworkers have merely followed the districtcourt’s lead insofar as the court itselfraised the same arguments under Rule23(b)(3) as it did regarding commonalityunder Rule 23(a)(2).23 See J.A. 10958–59;see also United States v. Goforth, 465 F.3d730, 737 (6th Cir.2006) (observing that‘‘where an argument advanced in an appel-lant’s opening brief applies to and essen-tially subsumes an alternative basis foraffirmance not separately argued therein,the appellant does not waive that alterna-tive basis for affirmance’’). The districtcourt based its conclusion that commonissues did not predominate on the observa-tion that because the workers’ evidencedisproportionately concerns the Beam Mill,‘‘there is no ‘glue’ connecting the pro-motions decisions in the Beam Mill to thedecisions in the other departments.’’ J.A.10959. That is exactly the same argumentraised, and responded to by the workers,in the context of Rule 23(a)(2) commonali-ty. See J.A. 10950–54; Appellants’ Br. 42–47. Elsewhere in its Rule 23(b)(3) discus-sion, the court observes that ‘‘[a]lthoughthere are, to varying degrees, a few allega-tions of discrimination in promotions indepartments other than the Beam Mill,there is nothing to link these allegations tothe pattern of behavior alleged in theBeam Mill.’’ J.A. 10959. Again, this argu-ment is also made in the Rule 23(a)(2)

context and responded to in detail by theworkers there. Indeed, the district courtitself acknowledged that it ‘‘employ[ed] thelanguage of Wal–Mart ’’ regarding Rule23(a)(2) in discussing the requirements ofRule 23(b)(3). J.A. 10958–59. In respond-ing directly to the reasons given by thedistrict court for its predominance deter-mination, the workers have thus done farmore than take a mere ‘‘passing shot atthe issue.’’ See Belk, Inc. v. Meyer Corp.,679 F.3d 146, 152 n. 4 (4th Cir.2012) (find-ing that an issue was waived after a partymentioned the issue in a heading but failedto further develop the argument); see alsoWilliams v. Woodford, 384 F.3d 567, 587 n.5 (9th Cir.2004) (concluding that an appel-lant preserved a claim for review eventhough the argument consisted of ‘‘eightsentences in a footnote,’’ where the argu-ment identified the basis of disagreementwith the district court, the requested re-lief, and relevant citations to case law andthe record).

Nonetheless, the dissent argues that‘‘many different reasons underlay [the dis-trict court’s] predominance finding, includ-ing several individual questions that could‘overwhelm’ common ones.’’ Post at 924.But a plain reading of the district court’sopinion belies the idea that it made anypredominance arguments that were not re-sponded to by the workers. The onlyspecific argument cited by the dissent asunaddressed contends that because of theworkers’ reliance on anecdotal evidence, ajury ‘‘would have to delve into the meritsof each individual promotion decision.’’J.A. 10959; post at 924. Yet, as observedabove, the workers specifically argued thatthe anecdotal evidence establishes ‘‘com-mon claims of discrimination’’ that meritcertification, not merely a finding of com-

23. Even superficially, the district court in-cludes its predominance analysis under theheading of ‘‘Subjectivity as a Policy,’’ dove-

tailing a discussion of commonality, insteadof as a separate section of analysis. See J.A.10954, 10956.


monality. Appellants’ Br. 42 (quotingBrown I, 576 F.3d at 153). Indeed, suchan argument is consistent with the work-ers’ fundamental contention throughouttheir brief that plant-wide discriminationexisted.

As this Court has observed, the purposeof the waiver doctrine is to avoid unfair-ness to an appellee and minimize the ‘‘riskof an improvident or ill-advised opinionbeing issued on an unbriefed issue.’’ Unit-ed States v. Leeson, 453 F.3d 631, 638 n. 4(4th Cir.2006) (citing McBride v. MerrellDow & Pharm., Inc., 800 F.2d 1208, 1211(D.C.Cir.1986)). Given the briefing pre-sented, the fully developed record below,and the lack of any showing of unfairnessor prejudice, there is simply no reasonwhy we should exercise our discretion todiscard years of litigation on appeal be-cause of an inartful opening brief. See AHelping Hand, LLC v. Baltimore Cnty.,Md., 515 F.3d 356, 369 (4th Cir.2008) (ob-serving that even when an argument hasbeen waived, this Court may nonethelessconsider it if a ‘‘miscarriage of justicewould otherwise result’’ (internal quotationmarks omitted)); cf. In re Am. W. Air-lines, Inc., 217 F.3d 1161, 1165 (9th Cir.2000) (observing that a court may refuse tofind waiver and consider an argumentraised for the first time on appeal whenthe issue ‘‘is one of law and either does notdepend on the factual record, or the recordhas been fully developed’’).

[15] Independent of the adequacy ofthe workers’ opening brief, the districtcourt had no grounds to revisit the ques-tion of predominance in the first placegiven this Court’s remand instructions andmandate in Brown I. Unlike the require-ment of commonality under Rule 23(a)(2)discussed above, Wal–Mart did notchange, nor purport to change, the Rule23(b)(3) analysis. Indeed, any impact ofthe Supreme Court’s ruling on the ques-

tion of whether common questions pre-dominate is only incidental insofar as Wal–Mart recalibrated what constitutes a com-mon question in the first place. The ma-jority in Wal–Mart only invoked Rule23(b)(3) to argue that the rule’s well-estab-lished procedural protections should applyto the plaintiffs’ claims for backpay. SeeWal–Mart, 131 S.Ct. at 2559.

Following our instructions in Brown Ifor the district court to ‘‘certify the appel-lants’ class action,’’ the court found that‘‘the putative class satisfied both the pre-dominance and superiority requirements ofRule 23(b)(3).’’ J.A. 10930. The courtthen certified the class for those employedin all six Nucor operations departments.The district court cites no new facts orlegal precedent after Brown I to justifyrevisiting that determination once the un-derlying question of commonality has beenresolved.

Nonetheless, the dissent insists that ourdecision in Brown I ‘‘did not prevent thedistrict court in any way from consideringpredominance because our prior decisiondid not say anything about predominance.’’Post at 926 (emphasis added). Such aconclusion misconstrues both the plain lan-guage of our original mandate and ignoresthe district court’s equally plain under-standing of it. The pivotal question indetermining the scope of the mandate iswhether the district court was free onremand to find that the workers had notsatisfied the predominance requirement.If so, then our mandate did not reach theissue and the district court was free toreconsider it. But if the court did nothave such liberty, then we must askwhether ‘‘controlling legal authority haschanged dramatically’’ regarding Rule23(b)(3) such that the court could reconsid-er the question. See Bell, 5 F.3d at 67. Ifno such change has occurred, then thedistrict court could not revisit it.


As for the first question, the districtcourt had no discretion to find that theworkers’ class failed to satisfy Rule23(b)(3), after we expressly told it ‘‘to cer-tify the appellants’ class action and to en-gage in further proceedings consistentwith this opinion.’’ Brown I, 576 F.3d at160; see also Bell, 5 F.3d at 66 (requiringthat a district court ‘‘implement both theletter and spirit of the TTT mandate, takinginto account [our] opinion and the circum-stances it embraces’’ (internal quotationmarks and citation omitted)); UnitedStates v. Pileggi, 703 F.3d 675, 679 (4thCir.2013) (observing that the mandate rule‘‘forecloses relitigation of issues expresslyor impliedly decided by the appellatecourt’’ (quoting Bell, 5 F.3d at 66)); S. Atl.Ltd., 356 F.3d at 583 (observing that amandate must be ‘‘scrupulously and fullycarried out’’ (internal quotation marks andcitation omitted)).

Indeed, the district court itself recog-nized that we had ‘‘dictate[d] the generaloutcome to be reached (class certification)while leaving [the district court] to fill inthe details.’’ J.A. 9886 (Order Den. Mot.for Recons. 8 n.2). Of course, the courtcould have, and did, evaluate whether cer-tification was best under Rule 23(b)(2) or(b)(3). But it had no discretion to thenfind that the prerequisites of either rulewere not met. As the court observed,Nucor’s argument on remand that theworkers had failed to satisfy Rule 23(b)‘‘overlook[ed] the Fourth Circuit’s priorholding in this case.’’ J.A. 9704 (Certifica-tion Order).24 Thus, the dissent misstatesthe record when it maintains that our orig-inal decision did not ‘‘in any way’’ preventthe district court from considering predo-minance. Post at 926–27. Indeed, follow-

ing our instructions and findings in BrownI, the court proceeded to make the onlyfinding it could under Rule 23(b)(3), name-ly, that ‘‘common issues predominate andthat a class action is superior to any othermethod for adjudication of the claims inthis case.’’ The dissent is thus also misin-formed when it states we are now certify-ing ‘‘a Rule 23(b)(3) class action withoutany court ever finding that the Rule23(b)(3) requirements are satisfied.’’ Postat 927.

Given the fact that our prior ruling fore-closed the denial of certification on thebasis of Rule 23(b)(3), the district courtneeded some compelling reason to recon-sider the question. Bell, 5 F.3d at 67(describing the ‘‘extraordinary’’ exceptionto the mandate rule when there is ‘‘ashow[ing] that controlling legal authorityhas changed dramatically’’). But the courtcited no such reason and, unlike the ques-tion of commonality, Wal–Mart providednone. Indeed, as the district court itselfacknowledged, Wal–Mart only incidentallynarrowed an inquiry into whether commonquestions predominate by clarifying whatconstitutes a common question in the firstplace under Rule 23(a)(2). J.A. 10971–72.

V.

More than seven years have now elapsedsince the workers first filed their classcertification motion, and the district courttwice has refused to certify the class. Thenature of the allegations, the evidentiarysupport buttressing them, and the inher-ent cohesiveness of the class all demon-strate that the court’s failure to certify wasan error. Rule 23 provides wide discretionto district courts, in part, to promote the

24. The dissent also maintains that our man-date did not reach the question of predomi-nance because we amended our original opin-ion in Brown I to delete a specific reference toRule 23(b)(3). Post at 927. Such a deletion,

however, did not change either our mandateto certify—a mandate that required the courtto find the workers had met Rule 23(b)—orthe district court’s express understanding ofthat mandate.


systemic class action virtues of efficiencyand flexibility. The realization of suchbenefits, however, requires that a districtcourt exercise its judgment in a reasonedand expeditious manner.

The dissent rightly observes that themajority presses forward ‘‘[o]n the road toits desired result.’’ Post at 956. And thatresult is simple justice. At bottom, theworkers seek nothing more than thechance to speak with one voice about thepromotions discrimination they allegedlysuffered as one class on account of oneuniting feature: the color of their skin.The dissent would deny them that chancewhile leading this Court down a differentroad—a road that would further weakenthe class action as a tool to realize TitleVII’s core promise of equality.

We vacate the district court’s decertifi-cation of the workers’ promotions classand remand the case to the district courtwith instructions to certify the class.

VACATED IN PART, AND REMAND-ED WITH INSTRUCTIONS.

AGEE, Circuit Judge, dissenting:

We typically tread lightly when review-ing a class certification decision, affording‘‘substantial deference’’ to the districtcourt, especially when it provides ‘‘well-supported factual findings.’’ Ward v. Dix-ie Nat’l Life Ins. Co., 595 F.3d 164, 179(4th Cir.2010). Class certification proceed-ings often call for fact-intensive choicesrequiring intimate knowledge of the pecu-liarities of complex litigation. Id. We usu-ally trust that the district court has thebetter eye for these sorts of questions.

The majority today declines to followthat path. It instead takes issue withalmost every aspect of the district court’sdecision to decertify, reversing that court’sdetermination because of newfound factson appeal and different notions about the

nature of this case. In doing so, the ma-jority creates a split between this Courtand another, see Bennett v. Nucor Corp.,656 F.3d 802 (8th Cir.2011), overlooks aplain and decisive waiver from the appel-lants, and drains a critical Supreme Courtdecision of much of its meaning, see Wal–Mart Stores, Inc. v. Dukes, ––– U.S. ––––,131 S.Ct. 2541, 180 L.Ed.2d 374 (2011). Irespectfully dissent.

I. Predominance

A.

The district court decertified Plaintiffs’promotions classes for two distinct rea-sons. First, the court found that Plaintiffshad not identified a ‘‘question[ ] of law orfact common to the class,’’ as Rule 23(a)(2)of the Federal Rules of Civil Procedurerequires. Second, it held that any ques-tions common to the class members did not‘‘predominate over any questions affectingonly individual members,’’ so the classcould not be certified under Rule 23(b)(3).Each of these separate reasons—common-ality or predominance—provide an inde-pendent ground to decertify the class.See, e.g., Thorn v. Jefferson–Pilot Life Ins.Co., 445 F.3d 311, 319 (4th Cir.2006).

Because the district court provided twodifferent bases for its decision, Plaintiffswere required to contest both. They didnot. Plaintiffs’ opening brief nowherementions the topic of predominance. Nei-ther does it refer to Rule 23(b). And eventhough ‘‘the main concern in the predomi-nance inquiry’’ is ‘‘the balance betweenindividual and common issues,’’ Myers v.Hertz Corp., 624 F.3d 537, 549 (2d Cir.2010), a reader searches in vain for anymention of such a ‘‘balancing’’ in Plaintiffs’submissions. Instead, Plaintiffs’ openingbrief focuses solely on Rule 23(a) common-ality. The brief does not even contain asimple statement that the district courterred as to predominance for the same


reasons that it purportedly erred as tocommonality—not to say that such a state-ment would be sufficient, either. See Jim-enez v. Allstate Ins. Co., 765 F.3d 1161,1165 n. 4 (9th Cir.2014) (holding that ‘‘cur-sory statements that the district court’sorder also incorrectly applied Rule23(b)(3)’s [predominance] requirement’’ are‘‘not enough to preserve the issue for ap-peal’’).

An appellant must raise every issue thathe wishes to press in his opening brief. Ifthe appellant fails to address an issuethere, then we will deem the issue waivedor abandoned. We have repeated this ruleso often that it might rightfully be termedthe best-established rule in appellate pro-cedure. See, e.g., Metro. Reg’l Info. Sys.,Inc. v. Am. Home Realty Network, 722F.3d 591, 602 n. 13 (4th Cir.2013); Kens-ington Volunteer Fire Dep’t, Inc. v. Mont-gomery Cnty., 684 F.3d 462, 472 n. 4 (4thCir.2012); Mayfield v. Nat’l Ass’n forStock Car Auto Racing, Inc., 674 F.3d 369,376 (4th Cir.2012); A Helping Hand, LLCv. Balt. Cnty., 515 F.3d 356, 369 (4th Cir.2008); French v. Assurance Co. of Am.,448 F.3d 693, 699 n. 2 (4th Cir.2006). As arule that ‘‘all the federal courts of appealsemploy,’’ waiver ‘‘makes excellent sense.’’Joseph v. United States, ––– U.S. ––––, 135S.Ct. 705, 705, 190 L.Ed.2d 461 (2014)(Kagan, J., respecting denial of certiorari).

In past cases, we have endeavored toapply our waiver rule consistently, findingwaiver whenever a party fails to ‘‘develop[his] argument’’—even if his brief takes apassing shot at the issue. Belk, Inc. v.Meyer Corp., 679 F.3d 146, 152 n. 4 (4thCir.2012). We have further found argu-ments waived even though they mighthave had merit. See IGEN Int’l, Inc. v.Roche Diagnostics GmbH, 335 F.3d 303,308–09 (4th Cir.2003); Pleasurecraft Ma-rine Engine Co. v. Thermo Power Corp.,272 F.3d 654, 657 (4th Cir.2001). And we

have applied the doctrine despite its poten-tially significant impact. See, e.g., Carterv. Lee, 283 F.3d 240, 252 n. 11 (4th Cir.2002) (applying the doctrine in a deathpenalty case).

Given that Plaintiffs failed to challengethe district court’s ruling on predominance,the plain and consistent waiver rule de-feats their appeal. ‘‘[T]o obtain reversal ofa district court judgment based on multi-ple, independent grounds, an appellantmust convince us that every stated groundfor the judgment against him is incorrect.’’In re Under Seal, 749 F.3d 276, 289 (4thCir.2014); accord Maher v. City of Chi.,547 F.3d 817, 821 (7th Cir.2008); Jankovicv. Int’l Crisis Grp., 494 F.3d 1080, 1086(D.C.Cir.2007). Appellate courts have re-peatedly affirmed district court decisionsdenying class certification where plaintiffsfailed to contest a predominance finding.See, e.g., Little v. T–Mobile USA, Inc., 691F.3d 1302, 1306–08 (11th Cir.2012); Klayv. Humana, Inc., 382 F.3d 1241, 1268 (11thCir.2004), abrogated on other grounds byBridge v. Phoenix Bond & Indemnity Co.,553 U.S. 639, 128 S.Ct. 2131, 170 L.Ed.2d1012 (2008); Applewhite v. ReichholdChems., Inc., 67 F.3d 571, 573–74 (5thCir.1995). Nothing calls for a differentresult here.

B.

In view of their failure to raise thepredominance issue, Plaintiffs now suggestthat ‘‘[p]redominance and commonality TTT

are [both] part of Rule 23(b)(3),’’ such thata challenge concerning one should betreated as a challenge to both. Appellant’sReply Br. 2. They are mistaken.

Commonality, found in Rule 23(a)(2),asks whether the proposed class will ‘‘re-solve an issue that is central to the validityof each of one of the claims in one stroke.’’EQT Prod. Co. v. Adair, 764 F.3d 347, 360(4th Cir.2014). Predominance, found in


Rule 23(b)(3), presents a ‘‘far more de-manding’’ inquiry, Amchem Prods., Inc. v.Windsor, 521 U.S. 591, 624, 117 S.Ct. 2231,138 L.Ed.2d 689 (1997), namely whetherany common questions ‘‘pre-dominate overany questions affecting only individualmembers,’’ Fed.R.Civ.P. 23(b)(3). Thus,while a ‘‘common issue’’ will establish com-monality, that common issue only goes toone part of the predominance inquiry.Consequently, courts and parties must ad-dress these requirements separately, rath-er than muddle them together. See Vegav. T–Mobile USA, Inc., 564 F.3d 1256,1268–70 (11th Cir.2009); In re Ins. Bro-kerage Litig., 579 F.3d 241, 277 (3d Cir.2009); accord Ealy v. Pinkerton Gov’tServs., Inc., 514 Fed.Appx. 299, 305 (4thCir.2013) (‘‘[T]he Rule 23(a) commonalityrequirement[ ] and the Rule 23(b)(3) pre-dominance requirement remain separateinquiries and the inquiries should not be‘blended.’ ’’).

The majority excuses Plaintiffs’ waiverbecause it believes that Plaintiffs ‘‘followedthe district court’s lead’’ in combining thetwo issues. Maj. op. at 919. Thus, eventhough commonality and predominance arelegally distinct, the majority speculatesthat the district court did not treat themas such here. The majority’s analysis mis-characterizes the district court’s opinion.

The district court did not just repeatback its commonality findings in determin-ing that Plaintiffs’ class failed as to predo-minance. To the contrary, the court ex-pressly held that it could not find therequired predominance ‘‘even if the FourthCircuit subsequently conclude[d] thatplaintiffs have identified a common issuethat satisfies Rule 23(a)(2).’’ J.A. 10956.The court then explained—over severalpages—that many different reasons under-lay its predominance finding, includingseveral individual questions that could ‘‘ov-erwhelm’’ common ones. Amgen Inc. v.

Conn. Ret. Plans & Trust Funds, ––– U.S.––––, 133 S.Ct. 1184, 1196, 185 L.Ed.2d 308(2013). Because Plaintiffs heavily rely onanecdotal evidence, for instance, the dis-trict court correctly concluded that a jury‘‘would have to delve into the merits ofeach individual promotion decision’’ to de-termine whether each decision evidenceddiscrimination. J.A. 10959. Thus, a trialmeant to resolve class-wide issues wouldlikely devolve into a series of mini-trialsexamining each promotion decision madein the Nucor plant. The court furtheracknowledged that ‘‘individual damages de-terminations,’’ like those that would berequired here, can ‘‘cut against class certi-fication.’’ J.A. 10956. Although it con-cluded that such damages determinationsdid not, standing alone, compel decertifica-tion in this case, J.A. 10958, they did pro-vide the district court an additional basisfor caution in making its predominancefinding. See, e.g., Cooper v. So. Co., 390F.3d 695, 722–23 (11th Cir.2004), overruledon other grounds by Ash v. Tyson Foods,Inc., 546 U.S. 454, 126 S.Ct. 1195, 163L.Ed.2d 1053 (2006) (noting that individu-alized damage issues could swamp the ad-vantages coming from an initial, class-wideliability determination); accord Allison v.Citgo Petroleum Corp., 151 F.3d 402, 421–22 (5th Cir.1998), cited with approval inGunnells v. Healthplan Servs., Inc., 348F.3d 417, 445 n. 18 (4th Cir.2003); see alsoComcast Corp. v. Behrend, ––– U.S. ––––,133 S.Ct. 1426, 1433, 185 L.Ed.2d 515(2013) (explaining that individual damage-related questions might destroy predomi-nance); Windham v. Am. Brands, Inc.,565 F.2d 59, 71–72 (4th Cir.1977).

The district court appropriately resolvedpredominance separately from commonali-ty. Plaintiffs’ failure to address the predo-minance finding in any way ends theirappeal.


C.

The majority at least recognizes thatPlaintiffs should have been ‘‘more explicit’’in addressing predominance. Maj. op. at918; see also id. at 919 (acknowledgingthat Plaintiffs’ ‘‘express[ ]’’ argumentslargely concern commonality). Even so, itconcludes that certain oblique referencesin Plaintiffs’ briefs preserved a predomi-nance-related challenge on appeal. Theydo not.

Plaintiffs’ statement of the issue on ap-peal, for instance, does not help them. Seemaj. op. at 918. The statement asks onlywhether ‘‘it [was] error or an abuse ofdiscretion for the district court not to fol-low this Circuit’s mandate’’ when it decer-tified the class. See Appellant’s Br. 1.Here again, Plaintiffs never mention pre-dominance, and the statement does nototherwise indicate any specific complaintwith the district court’s predominanceholding. Even if it had, that referencewould not have been enough without somefurther argument on the matter—an argu-ment that Plaintiffs wholly failed to pro-vide. See Belk, Inc., 679 F.3d at 153 n. 6;11126 Balt. Blvd., Inc. v. Prince George’sCnty., Md., 58 F.3d 988, 993 n. 7 (4thCir.1995).

The majority also ignores Plaintiffs’waiver because their brief contains somebroadly stated attacks on the districtcourt’s decertification decision—attackspurportedly not ‘‘limit[ed] to the questionof commonality.’’ Maj. op. at 919. But inthe usual case, a generalized attack on thelower court’s decision does not preservethe specific arguments that might be sub-sumed within the broader one. Quite theopposite: a ‘‘generalized assertion of er-ror’’ will not suffice to preserve anything.MMG Fin. Corp. v. Midwest AmusementsPark, LLC, 630 F.3d 651, 659 (7th Cir.2011); see also, e.g., Garrett v. Selby Con-nor Maddux & Janer, 425 F.3d 836, 841

(10th Cir.2005); Norman v. United States,429 F.3d 1081, 1091 n. 5 (Fed.Cir.2005).Preservation would have little to recom-mend it if litigants could make nebulous,broadly worded arguments and trust ap-pellate courts to work out the details oncethe opposing party points out the default.

In much the same way, Plaintiffs did notpreserve their predominance challenge byciting a few cases that happen to touchupon the concept. See maj. op. at 918.The traditional rule provides that citationsto the ‘‘occasional case,’’ without any fullerdiscussion, do not preserve an argument.Pike v. Guarino, 492 F.3d 61, 78 n. 9 (1stCir.2007); see also Am. Wildlands v.Kempthorne, 530 F.3d 991, 1001 (D.C.Cir.2008) (‘‘A fleeting statement in the paren-thetical of a citation is no more sufficientto raise a claim than a cursory remark in afootnote[.]’’). Similarly, ‘‘[m]ere notationof the applicable law, without any argu-mentation as to how it applies to [this]case, does not raise the issue of its applica-tion on appeal.’’ Sou v. Gonzales, 450F.3d 1, 6 n. 11 (1st Cir.2006) (internalquotation marks and citations omitted hereand throughout); accord Johnson v. Unit-ed States, 734 F.3d 352, 360 (4th Cir.2013).

The majority’s analysis casts an inappro-priate role for an appellate court. Now, acourt must review each decision that anappellant cites and independently considerwhether any part of it might underminethe district court’s judgment for some rea-son that the appellant never raised. Thatconcept reconceives the appellate courts’role, as those ‘‘courts do not sit as self-directed boards of legal inquiry and re-search.’’ Nat’l Aeronautics & Space Ad-min. v. Nelson, 562 U.S. 134, 147 n. 10, 131S.Ct. 746, 178 L.Ed.2d 667 (2011); see alsoWalker v. Prince George’s Cnty., Md., 575F.3d 426, 429 n. * (4th Cir.2009) (‘‘Judgesare not like pigs, hunting for truffles bur-ied in briefs.’’). In addition, using the


majority’s new rule, appellants may nowlaunch late-in-the-day challenges to anypart of a district court’s certification deci-sion so long as they serendipitously cited acase canvassing Rule 23 in their openingbrief. This ‘‘preservation-by-citation’’ ap-proach renders the waiver rule a nullity.

D.

In the end, the majority declares itselfunwilling to exercise its ‘‘discretion’’ to‘‘discard years of litigation on appeal be-cause of an inartful brief.’’ Maj. op. at920. That approach seems to give pro selitigant treatment to a brief crafted byexperienced class counsel—counsel thathas appeared in our court before. Surelyit does not expect too much from veterancounsel to ask them to make their argu-ments straight up and square. All themore so when these counsel have beenspecifically cautioned about waiver on pre-vious occasions. See, e.g., Davis v. Coca–Cola Bottling Co. Consol., 516 F.3d 955,972–73 (11th Cir.2008) (holding that partyrepresented by same counsel had ‘‘aban-doned’’ claim by failing to raise it in hisopening brief); see also Angles v. DollarTree Stores, Inc., 494 Fed.Appx. 326, 330n. 6 (4th Cir.2012) (same); cf. Bennett,656 F.3d at 821 (holding that party repre-sented by same counsel had ‘‘essentiallyabandoned’’ argument by making only a‘‘conclusory challenge’’); Anderson v. Ca-gle’s, Inc., 488 F.3d 945, 959 (11th Cir.2007) (same).

The ‘‘purpose’’ of the preservation ruleis also not served by overlooking Plaintiffs’waiver. See maj. op. at 919–20. The rule‘‘ensures that the opposing party has anopportunity to reflect upon and respond inwriting to the arguments that his adver-sary is raising.’’ Hamilton v. SouthlandChristian Sch., Inc., 680 F.3d 1316, 1319(11th Cir.2012); see also United States v.Leeson, 453 F.3d 631, 638 n. 4 (4th Cir.

2006) (noting that late arguments are ‘‘un-fair to the appellee’’); Pignons S.A. deMecanique v. Polaroid Corp., 701 F.2d 1, 3(1st Cir.1983) (‘‘In preparing briefs andarguments, an appellee is entitled to relyon the content of an appellant’s brief forthe scope of the issues appealed[.]’’). Nu-cor never had a chance to address Plain-tiffs’ predominance arguments directly, asPlaintiffs waited until their reply brief tomake them. Plaintiffs argued in their re-ply brief, for example, that no ‘‘height-ened’’ predominance standard applies afterWal–Mart Stores, Inc. v. Dukes, ––– U.S.––––, 131 S.Ct. 2541, 2551, 180 L.Ed.2d 374(2011), and the majority agrees, see maj.op. at 921. There might very well bereason to believe otherwise, though Nucorhas never had a chance to make that argu-ment. See, e.g., Andrey Spektor, TheDeath Knell of Issue Certification andWhy That Matters After Wal–Mart v.Dukes, 26 St. Thomas L.Rev. 165, 172(2014) (suggesting that Wal–Mart ren-dered it harder for issues to predominate).It must be cold comfort to Nucor, then, tohear that it was not ‘‘prejudice[d]’’ bythese and other unanswerable arguments.Maj. op. at 920.

E.

The majority goes on to hold that themandate rule barred the district courtfrom examining Rule 23(b)(3) predomi-nance. See maj. op. at 920–22. That viewis factually and legally incorrect. The de-cision in the prior appeal in this case didnot prevent the district court in any wayfrom considering predominance becauseour prior decision did not say anythingabout predominance.

In its original class certification decisionin 2007, the district court held that Plain-tiffs did not satisfy three of Rule 23(a)’sfour requirements. It expressly declinedto consider ‘‘the remaining requirements of


Rule 23(b).’’ J.A. 8997. On appeal, theparties’ submissions focused solely on Rule23(a). A majority of the Court then re-viewed these ‘‘Rule 23(a) factors’’ andfound them ‘‘satisfied.’’ Brown v. NucorCorp., 576 F.3d 149, 160 (4th Cir.2009)(‘‘Brown I ’’). The Brown I majority ini-tially went on to hold, in a single sentenceat the end of the opinion, that ‘‘the re-quirements of [Rule] 23(b)(3) ha[d] alsobeen satisfied for these claims.’’ SeeBrown v. Nucor Corp., 576 F.3d 149, 160(4th Cir.2009). Nucor then petitioned forrehearing en banc, arguing, among otherthings, that neither the lower court nor theparties had previously analyzed the Rule23(b) issue. See Nucor Pet. for Reh’g at 9,Brown I, 576 F.3d 149 (No. 08–1247), ECFNo. 53. In response, the Brown I panelamended its opinion and excised any men-tion of Rule 23(b)(3). See Order, Brown v.Nucor Corp., No. 08–1247 (4th Cir. Oct. 8,2009). One can easily discern why the opin-ion was amended: Brown I could not de-cide a fact-intensive issue—that is, the pre-dominance issue under Rule 23(b)(3)—when the parties had not yet argued it andthe district court had not yet addressed it.See Transamerica Leasing, Inc. v. Instit.of London Underwriters, 430 F.3d 1326,1332 (11th Cir.2005) (explaining that themandate rule and the broader law of thecase doctrine ‘‘cannot apply when the issuein question was outside the scope of theprior appeal’’). In fact, up to that point,Plaintiffs had never even sought certifica-tion under Rule 23(b)(3); they sought tocertify only a Rule 23(b)(2) class or, in thealternative, a so-called ‘‘hybrid’’ action.

By removing any reference to Rule23(b), Brown I left it to the district courtto determine in the first instance whetherPlaintiffs’ class met that provision’s re-quirements. The district court compliedwith both the letter and the spirit ofBrown I, and it correctly took ‘‘into ac-count [the] opinion and the circumstances

it embrace[d].’’ United States v. Bell, 5F.3d 64, 66 (4th Cir.1993); see also, e.g.,Lindy Pen Co. v. Bic Pen Corp., 982 F.2d1400, 1404–05 (9th Cir.1993) (affirming dis-trict court’s decision not to order account-ing or damages, despite appellate court’sinstructions to ‘‘order an accounting and toaward damages,’’ where district court act-ed in line with the ‘‘spirit’’ of the mandate).An appellate mandate ‘‘does not reachquestions which might have been decidedbut were not.’’ United States v. Lentz, 524F.3d 501, 528 (4th Cir.2008). And ‘‘[w]hilea mandate is controlling as to matterswithin its compass, on the remand a lowercourt is free as to other issues.’’ Spraguev. Ticonic Nat’l Bank, 307 U.S. 161, 168,59 S.Ct. 777, 83 L.Ed. 1184 (1939). Simplyput, the Brown I mandate did not apply toRule 23(b)(3), nor could it.

On remand after Brown I, the districtcourt initially certified the two promotionsclasses under Rule 23(b)(3). The courtlater reconsidered, as it was entitled to dounder Rule 23, which provides that ‘‘[a]norder that grants or denies class certifica-tion may be altered or amended beforefinal judgment.’’ Fed.R.Civ.P. 23(c)(1)(C);see also Fed.R.Civ.P. 54(b). ‘‘[C]ertifica-tions are not frozen once made,’’ Amgen,Inc., 133 S.Ct. at 1202 n. 9, and a districtcourt has ‘‘considerable discretion to de-certify the class,’’ Cent. Wesleyan Coll. v.W.R. Grace & Co., 6 F.3d 177, 189 (4thCir.1993). See also Prado–Steiman ex rel.Prado v. Bush, 221 F.3d 1266, 1273 (11thCir.2000). The district court could revisitits interlocutory decision regardless ofwhether, as the majority puts it, ‘‘newfacts or legal precedent [arose] afterBrown I.’’ Maj. op. at 920.

In effect, the majority today certifies aRule 23(b)(3) class action without anycourt ever finding that the Rule 23(b)(3)requirements are satisfied. It cannot gen-uinely contend that Brown I did the work,


as ‘‘the Fourth Circuit has never allowedthe rigorous Rule 23 analysis to be accom-plished implicitly.’’ Partington v. Am.Int’l Specialty Lines Ins. Co., 443 F.3d334, 341 (4th Cir.2006). And the districtcourt ultimately did not make such a find-ing either. The majority’s decision to cer-tify in part on this illusory mandate, then,substantially damages Rule 23(b)(3)’s ‘‘vitalprescription.’’ Amchem, 521 U.S. at 623,117 S.Ct. 2231. The Supreme Court re-cently reminded us that ‘‘plaintiffs wishingto proceed through a class action mustactually prove—not simply plead—thattheir proposed class satisfies each require-ment of Rule 23, including TTT the predo-minance requirement of Rule 23(b)(3).’’Halliburton Co. v. Erica P. John Fund,Inc., ––– U.S. ––––, 134 S.Ct. 2398, 2412,189 L.Ed.2d 339 (2014). At least as topredominance, Plaintiffs have yet to proveanything.

* * *

Plaintiffs did not challenge the districtcourt’s predominance ruling and do notcredibly explain why they failed to do so.The district court’s decision should there-fore be affirmed on that basis alone.

II. Relevant Standards

Even ignoring Plaintiffs’ waiver of thepredominance issue, they have not estab-lished that the district court abused itsdiscretion in finding insufficient commonal-ity. To see why, it is first necessary torecognize the standard that appellatecourts use in reviewing a district court’sclass-certification decision. Then, thestandard that the district court used inevaluating the evidence at the certificationstage must be considered.

A.

1.

A district court’s ultimate class-certifica-tion decision—that is, how it applied the

Rule 23 factors—is reviewed for an abuseof discretion. See, e.g., EQT Prod. Co.,764 F.3d at 357; Ward, 595 F.3d at 179;Monroe v. City of Charlottesville, Va., 579F.3d 380, 384 (4th Cir.2009); Gregory v.Finova Capital Corp., 442 F.3d 188, 190(4th Cir.2006). But reciting the standardis not enough; there must be genuine re-spect and adherence paid to the limits thatit imposes.

The abuse-of-discretion standard doesestablish some substantial limits, repre-senting ‘‘one of the most deferential stan-dards of review.’’ Matthew Bender & Co.v. West Publ’g Co., 240 F.3d 116, 121 (2dCir.2001). Under it, the appellate courtmay reverse only when ‘‘the [trial] court’sexercise of discretion, considering the lawand the facts, was arbitrary and capri-cious.’’ United States v. Mason, 52 F.3d1286, 1289 (4th Cir.1995). We act onlywhen the decision could not ‘‘have beenreached by a reasonable jurist,’’ or whenwe may call it ‘‘fundamentally wrong,’’‘‘clearly unreasonable, arbitrary, or fanci-ful.’’ Bluestein v. Cent. Wis. Anesthesiolo-gy, S.C., 769 F.3d 944, 957 (7th Cir.2014);accord Am. Copper & Brass, Inc. v. LakeCity Indus. Prods., Inc., 757 F.3d 540, 543(6th Cir.2014) (characterizing review of aclass certification decision as ‘‘very limit-ed’’).

Of course, deference does not equalblind acceptance. If, for instance, thedistrict court entirely fails to undertakesome part of the requisite analysis, thenit may be appropriate to reverse. See,e.g., EQT Prod., 764 F.3d at 371 (vacatingand remanding a certification order wherethe district court failed to conduct an ap-propriately rigorous analysis of Rule 23’srequirements). But when our review ven-tures into intensely factual matters or ar-eas of practical concern, then our defer-ence must be at its greatest—indeed, we


must stand aside in those circumstancesunless the lower court was ‘‘clearlywrong.’’ Windham, 565 F.2d at 65; ac-cord CGC Holding Co., LLC v. Broad &Cassel, 773 F.3d 1076, 1086 (10th Cir.2014) (‘‘[A]s long as the district court ap-plies the proper Rule 23 standard, we willdefer to its class certification ruling pro-vided that decision falls within the boundsof rationally available choices given thefacts and law involved in the matter athand.’’).

We do not then reverse anytime wedisagree with the result that the districtcourt reaches. See First Penn–Pac. LifeIns. Co. v. Evans, 304 F.3d 345, 348 (4thCir.2002). Rather, ‘‘the [abuse-of-discre-tion] standard draws a line TTT betweenthe unsupportable and the merely mistak-en, between the legal error, disorder ofreason, severe lapse of judgment, and pro-cedural failure that a reviewing court mayalways correct, and the simple disagree-ment that, on this standard, it may not.’’Evans v. Eaton Corp. Long Term Disabil-ity Plan, 514 F.3d 315, 322 (4th Cir.2008);see also Cooter & Gell v. Hartmarx Corp.,496 U.S. 384, 405, 110 S.Ct. 2447, 110L.Ed.2d 359 (1990) (holding that the dis-trict court did not abuse its discretionwhere it ‘‘applied the correct legal stan-dard and offered substantial justificationfor its finding’’).

These principles might strike some astruisms, but they carry special force in theclass-certification context. ‘‘Granting ordenying class certification is a highly fact-intensive matter of practicality,’’ Monrealv. Potter, 367 F.3d 1224, 1238 (10th Cir.2004), so much so that ‘‘[h]ighly fact-based,complex, difficult matters’’ arise as a mat-ter of routine, Amchem, 521 U.S. at 630,117 S.Ct. 2231 (Breyer, J., concurring inpart and dissenting in part). Unsurpris-ingly, then, we give district courts ‘‘broaddiscretion in deciding whether to allow the

maintenance of a class action.’’ Roman v.ESB, Inc., 550 F.2d 1343, 1348 (4th Cir.1976); see also Lowery v. Circuit CityStores, Inc., 158 F.3d 742, 757–58 (4thCir.1998), vacated 527 U.S. 1031, 119 S.Ct.2388, 144 L.Ed.2d 790 (1999), reaff’d inrelevant part, 206 F.3d 431 (4th Cir.2000).As with any other decision that appellatecourts review for abuse of discretion, weshould affirm a certification decision evenif we are convinced that ‘‘reasons clearlyexisted for taking the other course.’’ Lew-is v. Bloomsburg Mills, Inc., 773 F.2d 561,564 (4th Cir.1985); accord Simmons v.Poe, 47 F.3d 1370, 1382 (4th Cir.1995).

2.

An appellate court must be even morecareful in reviewing any factual findingsunderlying the district court’s decision, aswe review those only for clear error.Thorn, 445 F.3d at 317–18; see also Fed.R.Civ.P. 52(a)(6). ‘‘The clear error stan-dard TTT protects district courts’ primacyas triers of fact.’’ Evans, 514 F.3d at 321.Our opinions have repeatedly emphasizedthat clear-error review is ‘‘narrow,’’ Walk-er v. Kelly, 593 F.3d 319, 323 (4th Cir.2010), ‘‘highly deferential,’’ Green v. John-son, 515 F.3d 290, 301 (4th Cir.2008), and‘‘particularly circumscribed,’’ Jiminez v.Mary Washington Coll., 57 F.3d 369, 378(4th Cir.1995). We may reverse findingsreviewed under this standard only when,having reviewed the entire record, we are‘‘left with the definite and firm convictionthat a mistake has been committed.’’United States v. Heyer, 740 F.3d 284, 292(4th Cir.2014). If the district court chosebetween ‘‘two permissible views of the evi-dence,’’ or if it otherwise offered a ‘‘plausi-ble’’ account of that evidence, Anderson v.City of Bessemer City, N.C., 470 U.S. 564,574, 105 S.Ct. 1504, 84 L.Ed.2d 518 (1985),then its factual findings are ‘‘conclusive,’’Walker, 593 F.3d at 323. And as with theabuse-of-discretion standard, we cannot re-


verse merely because we would have de-cided the matter differently. SeeAnderson, 470 U.S. at 573, 105 S.Ct. 1504.

3.

Despite these deferential standards ofreview, the majority identifies reversibleerror in virtually every legal and factualjudgment that the district court rendered.Yet in searching the majority’s opinion forany of the hallmarks of deference—expla-nations as to how the district court clearlyerred, or full analysis of how the districtcourt abused its discretion—we find verylittle.

In truth, the majority seems to applyjust about every standard of review but adeferential one. For the most part, themajority offers bare statements that thedistrict court erred, apparently becausethe district court decided things differentlythan the majority would have. For in-stance, it insists that Plaintiffs’ statisticalevidence is simply ‘‘less precise’’ and re-jects out-of-hand the district court’s viewthat the evidence was ‘‘fundamentally un-reliable.’’ Maj. op. at 904, 906. Likewise,it draws its own conclusions about theanecdotal evidence, reciting certain por-tions of certain affidavits and declaringthem enough. It makes credibility deter-minations, categorically rejecting Nucor’sevidence as ‘‘self-serving,’’ id. at 906, or‘‘coercive,’’ id. at 913, while embracing con-trary statements from Plaintiffs becausethe majority finds them ‘‘credible,’’ id. at913. And it offers its own notions aboutwhat is ‘‘plain,’’ id. at 908, ‘‘elementary,’’id. at 912, or ‘‘common sense,’’ id. at 913.The majority does so even while decryingthe dangers of ‘‘cherry pick[ing] facts froman 11,000 page record.’’ Id. at 913. Inshort, the majority opinion shows little re-spect for a district court that is far morefamiliar with each page of the record thanwe are.

Contravening our ‘‘axiomatic’’ ruleagainst factual findings on appeal, CoreCommc’ns, Inc. v. Verizon Md. LLC, 744F.3d 310, 324 (4th Cir.2014), the majorityeventually finds in the first instance that‘‘there is only one answer to the questionof why Nucor’s black workers were consis-tently disfavored,’’ maj. op. at 913. Thisadventuresome approach is rather jarringwhen placed against the more measuredmethods found in some of our other classcertification decisions. See, e.g., EQTProd., 764 F.3d at 371 (remanding forfurther consideration of class certificationafter determining that district court mis-applied the relevant standards); Gariety v.Grant Thornton, LLP, 368 F.3d 356, 366(4th Cir.2004) (same). Making mattersworse, the majority offers no good reasonfor it. Instead, it engages in a ratherextended discussion of the Brown I dissentand then declares any attack on the major-ity’s factfinding today ‘‘iron[ic].’’ Maj. op.at 905.

Too often, we fail to give standards ofreview the attention that they deserve.We see them recited in boilerplate andthen dispensed with when the perceivedexigencies of a case seem to call for it.But ‘‘[s]tandards of review are TTT an ele-mental expression of judicial restraint,which, in their deferential varieties, safe-guard the superior vantage points of thoseentrusted with primary decisional respon-sibility.’’ Evans, 514 F.3d at 320–21. Anappellate court should not be so quick toignore them.

B.

We must next consider the districtcourt’s role in deciding the certificationmotion in the first place. The majorityimplies that the district court too readilydismissed Plaintiffs’ efforts to certify. Butthe district court was not just permitted to


take a hard look at Plaintiffs’ submis-sions—it was required to.

1.

Although plaintiffs shoulder the burdenof demonstrating that a proposed classcomplies with Rule 23, the district courthas an ‘‘independent obligation to performa rigorous analysis to ensure that all of theprerequisites have been satisfied.’’ EQTProd., 764 F.3d at 358. Among otherthings, this ‘‘rigorous analysis’’ requiresthe district court ‘‘to resolve a genuinelegal or factual dispute relevant to deter-mining the requirements.’’ In re Hydro-gen Peroxide Antitrust Litig., 552 F.3d305, 320 (3d Cir.2008).

‘‘[C]areful attention to the requirementsof [Rule] 23 remains TTT indispensable’’even in cases ‘‘alleging racial or ethnicdiscrimination.’’ E. Tex. Motor FreightSys., Inc. v. Rodriguez, 431 U.S. 395, 405,97 S.Ct. 1891, 52 L.Ed.2d 453 (1977).Thus, ‘‘a Title VII class action, like anyother class action, may only be certified ifthe trial court is satisfied, after a rigorousanalysis, that the prerequisites of [theRule] have been satisfied.’’ Gen. Tel. Co.of Sw. v. Falcon, 457 U.S. 147, 161, 102S.Ct. 2364, 72 L.Ed.2d 740 (1982); see alsoDesert Palace, Inc. v. Costa, 539 U.S. 90,99, 123 S.Ct. 2148, 156 L.Ed.2d 84 (2003)(noting the ‘‘conventional rule[s] of civillitigation TTT generally appl[y] in Title VIIcases’’). And there is no ‘‘entitlement toclass proceedings for the vindication ofstatutory rights,’’ Am. Express Co. v. Ital-ian Colors Rest., ––– U.S. ––––, 133 S.Ct.2304, 2309, 186 L.Ed.2d 417 (2013), TitleVII included. Thus, the Court must becareful not to bend and twist the ‘‘rigorousanalysis’’ that Rule 23 compels merely forthe sake of abstract notions of Title VII’sobjectives and purposes. Cf. Touche Ross& Co. v. Redington, 442 U.S. 560, 578, 99S.Ct. 2479, 61 L.Ed.2d 82 (1979) (‘‘[G]ener-

alized references to the ‘remedial pur-poses’ of [a statute] will not justify readinga provision more broadly than its languageand the statutory scheme reasonably per-mit.’’). To do so would not only ignore theSupreme Court’s warnings; it might alsohave unforeseen effects in the many otherareas of law in which Rule 23 is implicated.

In basic terms, the rigorous-analysisstandard tests whether plaintiffs have pre-sented substantial evidence of compliancewith Rule 23. Plaintiffs may ‘‘not simplyplead’’ that the relevant requirements havebeen met, but must ‘‘actually prove’’ it.Halliburton, 134 S.Ct. at 2412; accordMonroe, 579 F.3d at 384. To meet thatstandard, plaintiffs must summon ‘‘eviden-tiary proof,’’ Comcast, 133 S.Ct. at 1432,and ‘‘affirmatively demonstrate [their]compliance with the Rule,’’ Wal–Mart, 131S.Ct. at 2551. ‘‘[S]ome evidence’’ is notenough. In re Initial Pub. Offerings[‘‘IPO’’] Sec. Litig., 471 F.3d 24, 33 (2dCir.2006).

Before certifying a class action, courtswill require a plaintiff to establish by apreponderance of the evidence that theaction complies with each part of Rule 23.See In re U.S. Foodservice Inc. PricingLitig., 729 F.3d 108, 117 (2d Cir.2013);Carrera v. Bayer Corp., 727 F.3d 300, 306(3d Cir.2013); Messner v. NorthshoreUniv. HealthSystem, 669 F.3d 802, 811(7th Cir.2012); Ala. Elec. Pension Fund v.Flowserve Corp., 572 F.3d 221, 228 (5thCir.2009), abrogated in other respects byHalliburton, 134 S.Ct. 2398; accord In reTitanium Dioxide Antitrust Litig., 284F.R.D. 328, 336 (D.Md.2012); In re MillsCorp. Sec. Litig., 257 F.R.D. 101, 104(E.D.Va.2009); In re Safety–Kleen Corp.Bondholders Litig., No. 3:00–1145–17, 2004WL 3115870, at *2 (D.S.C. Nov. 1, 2004);see also Anthony F. Fata, Doomsday De-layed: How the Court’s Party–NeutralClarification of Class Certification Stan-


dards in Wal–Mart v. Dukes ActuallyHelps Plaintiffs, 62 DePaul L.Rev. 675,681 (2013) (reading Wal–Mart to apply apreponderance-of-the-evidence standard).

2.

‘‘[T]he factors spelled out in Rule 23must be addressed through findings, evenif they overlap with issues on the merits.’’Gariety, 368 F.3d at 366; accord In reRail Freight Fuel Surcharge Antitrust Li-tig., 725 F.3d 244, 249 (D.C.Cir.2013)(recognizing that certification will some-times ‘‘resemble[ ] an appraisal on themerits’’). Obviously, ‘‘[a] court may notsay something like ‘let’s resolve the meritsfirst and worry about the class later’ TTT

or ‘I’m not going to certify a class unless Ithink that the plaintiffs will prevail.’ ’’ Sza-bo v. Bridgeport Machs., Inc., 249 F.3d672, 677 (7th Cir.2001), cited with approvalin Wal–Mart, 131 S.Ct. at 2552. But over-lap ‘‘cannot be helped,’’ as certification‘‘generally involves considerations that areenmeshed in the factual and legal issuescomprising the plaintiff’s cause of action.’’Wal–Mart, 131 S.Ct. at 2551–52. Com-pare Brown I, 576 F.3d at 156 (citingEisen v. Carlisle & Jacquelin, 417 U.S.156, 177, 94 S.Ct. 2140, 40 L.Ed.2d 732(1974), and refusing to inquire into Plain-tiffs’ statistics because it would be an im-permissibly ‘‘in-depth assessment of themerits’’), with Wal–Mart, 131 S.Ct. at 2552& n. 6 (admonishing courts not to ‘‘mistak-enly cite[ ]’’ Eisen for the incorrect ideathat merits inquiries are barred).

3.

Contrast these well-defined and rigorousstandards with the ambiguous and limit-less ones found in the majority opinion.The majority acknowledges the ‘‘rigorousanalysis’’ that lower courts must perform,but abandons that standard soon aftermentioning it. Instead, it treats the evi-

dentiary standard for certification as onedifferent from that required for a party toprevail on the merits, never acknowledgingthat this view breaks from the manycourts (including those in our Circuit) thatapply the preponderance standard. Nordoes it even tell us what a ‘‘rigorous analy-sis’’ might consist of. Instead, it merelyinvokes Amgen, a case that addresseswhat questions may be considered on classcertification, not what evidence will sufficeto answer them. 133 S.Ct. at 1194–95.Having rendered the rigorous analysis lessrigorous than other courts’ (though towhat degree, one does not know), the ma-jority then proceeds to apply its weakenedtest, repeatedly using mere allegations—or, sometimes, allegations ‘‘proven’’ by al-legations—to justify certification. See,e.g., maj. op. at 906, 910, 910, 912, 914, 915,917, 917, 921. The necessary implication isthat the majority’s ‘‘rigorous analysis’’ con-sists of very little.

One finds a further hint at the level ofproof that the majority means to applywhen it embraces Brown I ’s metric. Maj.op. at 903. Brown I held that ‘‘allega-tions’’ of disparate treatment were enoughto establish commonality, a conclusion atodds with Wal–Mart. Compare Brown I,576 F.3d at 153, with Wal–Mart, 131 S.Ct.at 2553 (distinguishing between an ‘‘other-wise unsupported allegation’’ and the ‘‘sig-nificant proof’’ required to establish a com-mon policy). The majority in Brown I alsosaid that anecdotes from three employeesconcentrated in a single departmentproved a common policy of discrimination.576 F.3d at 153. And it held that statisticalevidence of ‘‘relatively weak probative val-ue’’ was enough, even though problems inthat evidence—the statistical evidenceseen here—might ‘‘very well discredit’’ itat some later stage. Id. at 156 & n. 10. Inshort, Brown I required the plaintiffs tosummon an exceptionally low, almost non-


existent level of proof at the class-certifica-tion stage.

The majority’s decision to reanimateBrown I’s negligible evidentiary standardleaves this circuit alone on an island. TheBrown I majority suggested that its le-nient view of the necessary evidencealigned with the Second Circuit’s decisionin Caridad v. Metro–North CommuterRailroad, 191 F.3d 283, 293 (2d Cir.1999).See Brown I, 576 F.3d at 157 (citing Cari-dad, 191 F.3d at 293). But by the timeBrown I was issued, the Second Circuithad already repudiated any part of Cari-dad suggesting a lesser burden of proofthan a preponderance of the evidence. SeeIn re IPO, 471 F.3d at 42 (‘‘[O]ur conclu-sions necessarily preclude the use of a‘some showing’ standard, and to whateverextent Caridad might have implied such astandard for a Rule 23 requirement, thatimplication is disavowed.’’). Only one cir-cuit followed Brown I’s lead and acceptedsuch a low degree of proof: the NinthCircuit, in its now-reversed decision inDukes v. Wal–Mart Stores, Inc. See 603F.3d 571, 595–96 & n. 17 (9th Cir.2010).(citing Brown I, 576 F.3d at 156). In themeantime, another circuit rejected BrownI outright. See Bennett, 656 F.3d at 816 n.2 (declining to ‘‘follow’’ Brown I’s findingthat sufficient evidence established com-monality, as ‘‘Brown [I] was decided with-out the benefit of the Supreme Court’srecent opinion in Dukes ’’).

All in all, despite assurances otherwise,the majority treats Rule 23 as somethingakin to a pleading standard. It is not.See Wal–Mart, 131 S.Ct. at 2551. Werethe rule written as the majority envisionsit, district courts would get to ‘‘duck hardquestions.’’ West v. Prudential Sec., Inc.,282 F.3d 935, 938 (7th Cir.2002). Butframing class certification as a mere plead-ing standard ‘‘amounts to a delegation ofjudicial power to the plaintiffs.’’ Id. ‘‘[A]district court’s certification order often

bestows upon plaintiffs extraordinaryleverage, and its bite should dictate theprocess that precedes it.’’ Oscar PrivateEquity Invs. v. Allegiance Telecom, Inc.,487 F.3d 261, 267 (5th Cir.2007), abrogatedin other respects by Erica P. John Fund,Inc. v. Halliburton Co., ––– U.S. ––––, 131S.Ct. 2179, 180 L.Ed.2d 24 (2011).

III. Commonality

With the proper standards in mind, itbecomes evident that the district court didnot abuse its discretion in finding thatPlaintiffs failed to establish commonality.

‘‘In this case, proof of commonality nec-essarily overlaps with [Plaintiffs’] meritscontention that [Nucor] engages in a pat-tern or practice of discrimination.’’ Wal–Mart, 131 S.Ct. at 2552. Plaintiffs mustestablish a unifying policy of discrimina-tion at certification, or ‘‘it will be impossi-ble to say that examination of all the classmembers’ claims for relief will produce acommon answer to the crucial question [of]why was I disfavored.’’ Id. In otherwords, Plaintiffs cannot simply identify agroup of people who they allege have suf-fered some type of Title VII injury. Id.To certify the class, Plaintiffs must be ableto trace that injury to a single, commonsource. Id.; accord Ellis v. Costco Whole-sale Corp., 657 F.3d 970, 981 (9th Cir.2011); see also William B. Rubenstein,Newberg on Class Actions § 3:19 (5thed.2014) (citing Brown I as an example ofa case that approached commonality‘‘loosely’’ and explaining that Wal–Martarticulated ‘‘a more explicit definition ofcommonality’’). Plaintiffs here must iden-tify a common policy with common injuryto members of a class spanning more thana decade, covering Nucor’s entire SouthCarolina production facility, and touchingupon dozens of relevant decisionmakers.That task can be decidedly difficult, espe-cially given that Plaintiffs premise their

u0034828

Highlight


class in part on a disparate treatment the-ory. See Stastny v. S. Bell Tel. & Tel. Co.,628 F.2d 267, 274 n. 10 (4th Cir.1980); seealso Garcia v. Johanns, 444 F.3d 625, 633(D.C.Cir.2006) (‘‘Establishing commonalityfor a disparate treatment class is particu-larly difficult where, as here, multiple deci-sionmakers with significant local autonomyexist.’’).

A plaintiff who brings a class-widecharge of discrimination must traverse a‘‘wide gap’’ between his claim of individualmistreatment and a class-wide harm. Fal-con, 457 U.S. at 157, 102 S.Ct. 2364. Theplaintiff could do so in one of two ways.See Wal–Mart, 131 S.Ct. at 2553. First,he might identify a ‘‘biased testing proce-dure’’ that is used to evaluate applicantsand employees. Id. By all accounts, Plain-tiffs do not identify that sort of procedurehere. Second, a plaintiff might offer ‘‘sig-nificant proof’’ that an employer ‘‘operatedunder a general policy of discriminationTTT [that] manifested itself in hiring andpromotion practices in the same generalfashion.’’ Id. This second route forms thefocus of this case.

Plaintiffs offer two types of evidencethat they say bridge the gap between indi-vidual and class-wide claims: statistical ev-idence and anecdotal evidence. Whetherexamining these two categories of evidenceseparately or together, the district courtdid not abuse its discretion in deeming thePlaintiffs’ case insufficient.

A. Statistical Evidence

1.

Plaintiffs first present a statistical studycomparing a hypothesized, weightedbenchmark of black bidders for promotionsto the number of black employees thatthey assumed Nucor promoted during therelevant period. This evidence performs adouble duty, as it goes to Plaintiffs’ dispa-

rate impact claim and their disparatetreatment claim.

As to the disparate impact claim, thissort of statistical evidence should identifydisparities that are ‘‘sufficiently substan-tial’’ to raise ‘‘an inference of causation.’’Anderson v. Westinghouse Savannah Riv-er Co., 406 F.3d 248, 281 (4th Cir.2005).Without ‘‘substantial’’ disparities, we can-not be confident that a challenged policyproduced an injury common to the class.See Wal–Mart, 131 S.Ct. at 2551.

As to the disparate treatment claim,‘‘gross statistical disparities’’ ‘‘may in aproper case constitute prima facie proof ofa pattern or practice of discrimination.’’Hazelwood Sch. Dist. v. United States, 433U.S. 299, 307–08, 97 S.Ct. 2736, 53 L.Ed.2d768 (1977); accord Ardrey v. United Par-cel Serv., 798 F.2d 679, 683 (4th Cir.1986).But see Warren v. Halstead Indus., Inc.,802 F.2d 746, 753 (4th Cir.1986) (‘‘[S]tatis-tics cannot alone prove the existence of apattern or practice of discrimination[.]’’).But not every case will present the trulyegregious and unexplained disparities thatleave no room for any inference other thanintentional discrimination. Moreover,‘‘[i]nferring past discrimination from sta-tistics alone assumes the most dubious ofconclusions: that the true measure of ra-cial equality is always to be found in nu-meric proportionality.’’ Md. TroopersAss’n, Inc. v. Evans, 993 F.2d 1072, 1077(4th Cir.1993).

2.

The majority observes that Plaintiffs’evidence is ‘‘statistically significant at 2.54standard deviations from what would beexpected if race were a neutral factor.’’Maj. op. at 908. Statistical significance,however, is a necessary but not sufficientcondition to finding a discriminatory prac-tice or policy; statistical significance doesnot axiomatically equate with legal signifi-


cance. See EEOC v. Fed. Reserve Bankof Richmond, 698 F.2d 633, 648 (4th Cir.1983) (‘‘[S]tatistical significance as meas-ured by the standards of acceptable statis-tical principles will not necessarily be le-gally significant[.]’’), rev’d sub nom onother grounds, Cooper v. Fed. ReserveBank of Richmond, 467 U.S. 867, 104S.Ct. 2794, 81 L.Ed.2d 718 (1984). Highstatistical significance levels might lackpractical and legal significance, for in-stance, because ‘‘a high significance levelmay be a misleading artifact of the study’sdesign.’’ Kadas v. MCI SystemhouseCorp., 255 F.3d 359, 362 (7th Cir.2001).Thus, determining what is legally signifi-cant—as opposed to statistically signifi-cant—‘‘is a legal determination properlymade by the court and not by an expert.’’Fed. Reserve Bank of Richmond, 698 F.2dat 648; cf. United States v. Philip Mor-ris USA, Inc., 449 F.Supp.2d 1, 706 n. 29(D.D.C.2006) (criticizing one of Plaintiffs’experts for his undue reliance on statisti-cal significance).

Nevertheless, the majority seems to de-fer to Plaintiffs’ experts and assume legalsignificance because the statistical evi-dence crosses the two-standard-deviationthreshold, the threshold for statistical sig-nificance at a 95% confidence level. Yet‘‘courts of law should be extremely cau-tious in drawing any conclusions fromstandard deviations in the range of one tothree.’’ EEOC v. Am. Nat’l Bank, 652F.2d 1176, 1192 (4th Cir.1981); see alsoKingsley R. Browne, Statistical Proof ofDiscrimination: Beyond ‘‘Damned Lies’’,68 Wash. L.Rev. 477, 503 (1993) (‘‘Randomdisparities of this magnitude are pervasivein the workplace and are not suggestive ofa nonrandom cause, let alone an illegalone.’’). In specific cases, even higher num-bers may not be enough. EEOC v. West-ern Electric Co., Inc., 713 F.2d 1011 (4thCir.1983), provides one example. There,we held that a district court clearly erred

in finding a policy or practice of discrimi-nation, even though statistics showed over-all disparities of 4.7955 and 5.883 standarddeviations. Id. at 1018–19.

Similarly, other courts have rejected sta-tistical evidence even though the evidencemet the two-standard-deviation threshold.See, e.g., Carpenter v. Boeing Co., 456 F.3d1183, 1201 (10th Cir.2006) (7.95 and 38.03standard deviations); Lopez v. LaborersInt’l Union Local No. 18, 987 F.2d 1210,1213–14 (5th Cir.1993) (3.26 and 3.01 stan-dard deviations); Waisome v. Port Auth.of N.Y. & N.J., 948 F.2d 1370, 1376 (2dCir.1991) (2.68 standard deviations);EEOC v. Chi. Miniature Lamp Works, 947F.2d 292, 300 (7th Cir.1991) (20.1 standarddeviations); Gay v. Waiters’ & DairyLunchmen’s Union, Local No. 30, 694F.2d 531, 551 (9th Cir.1982) (2.45 standarddeviations). In short, ‘‘there is nothingmagical about two or three standard devia-tions.’’ Ramona L. Paetzold & Steve L.Willborn, The Statistics of Discrimination§ 4:13 (2014).

3.

Instead of assuming ‘‘that any particularnumber of ‘standard deviations’ ’’ estab-lishes a discriminatory policy, courts mustevaluate statistical evidence on a ‘‘case-by-case basis.’’ Watson v. Fort Worth Bank& Trust, 487 U.S. 977, 995 n. 3, 108 S.Ct.2777, 101 L.Ed.2d 827 (1988) (pluralityopinion); see also Int’l Bhd. of Teamstersv. United States, 431 U.S. 324, 340, 97S.Ct. 1843, 52 L.Ed.2d 396 (1977). Nei-ther ‘‘courts [n]or defendants [are] obligedto assume that plaintiffs’ statistical evi-dence is reliable.’’ Watson, 487 U.S. at996, 108 S.Ct. 2777. And we must alwayskeep in mind that we are looking for reli-able indications of ‘‘gross’’ or ‘‘substantial’’disparities that amount to ‘‘significantproof.’’ Wal–Mart, 131 S.Ct. at 2551,


2553; Hazelwood, 433 U.S. at 307–08, 97S.Ct. 2736.

The duty to test the relevant statisticalevidence attaches at the class certificationstage, Comcast, 133 S.Ct. at 1433, as ‘‘reli-ance on unverifiable evidence is hardlybetter than relying on bare allegations,’’Unger v. Amedisys, Inc., 401 F.3d 316, 324(5th Cir.2005). District courts must probethe validity of statistical evidence, as ‘‘anymethod of measurement’’ would otherwisebecome ‘‘acceptable so long as it c[ould] beapplied classwide, no matter how arbitrarythe measurements may be.’’ Comcast, 133S.Ct. at 1433; accord Rail Freight FuelSurcharge Antitrust Litig., 725 F.3d at254; Am. Honda Motor Co., Inc. v. Allen,600 F.3d 813, 815 (7th Cir.2010).

In this case, the district court evaluatedPlaintiffs’ statistical evidence, reasonablyfound it wanting, and explained in detailwhy that was so. It should not then besaid that the district court clearly erred byrefusing to give weight to unconvincingevidence. And when one takes a closerlook, Plaintiffs’ statistical evidence truly isfundamentally unconvincing, not just—asthe majority calls it—‘‘less precise.’’ Maj.op. at 904.

4.

‘‘[T]rial judges may evaluate the dataoffered to support an expert’s bottom-lineopinions to determine if that data providesadequate support to mark the expert’s tes-timony as reliable.’’ Milward v. AcuitySpecialty Prods. Grp., Inc., 639 F.3d 11, 15(1st Cir.2011). And in any case involvingexpert testimony, ‘‘a court may concludethat there is simply too great an analyticalgap between the data and the opinion of-fered.’’ Gen. Elec. Co. v. Joiner, 522 U.S.136, 146, 118 S.Ct. 512, 139 L.Ed.2d 508(1997).

Plaintiffs’ own experts conceded thatthey used problematic data. In support of

a motion to compel, one of Plaintiffs’ ex-perts affirmed under oath that the infor-mation he had received thus far was ‘‘in-complete in a number of important waysthat ma[d]e it impossible to calculate reli-able statistics.’’ J.A. 399. Because of this‘‘inadequate’’ data, the expert opined thathe could not calculate ‘‘proper statistics’’ orperform ‘‘any of th[e] three standard formsof statistical analysis.’’ J.A. 403, 409.Without additional data, it was concededly‘‘impossible to calculate TTT statistical pat-terns that might show whether or not acommon issue of fact exists in this case.’’J.A. 403–04. Ultimately, the expert didnot receive any of the additional data thathe professed to need for a scientificallyvalid analysis. But, despite his swornstatements that the task was ‘‘impossible,’’he and another expert nevertheless pro-duced statistical analyses based on the ‘‘in-complete’’ and ‘‘inadequate’’ data.

Plaintiffs’ experts’ report confirms thatthey used incomplete data to support andreach their conclusions. For instance,even though the experts drew conclusionsabout positions throughout the Nucorplant, they did not employ any data fromeither the shipping or maintenance depart-ments. J.A. 1154. They used only a ‘‘lim-ited amount of data’’ for the remainingdepartments. J.A. 1153. And althoughPlaintiffs’ experts chose to use biddingdata to determine an expected number ofblack promotions, they conceded that in-complete data ‘‘undermined’’ their ‘‘abilityto use posting and bidding records to ana-lyze [those] promotions.’’ J.A. 1161. Nu-cor’s expert identified other basic issues inPlaintiffs’ experts’ data that the majorityopinion ignores. See J.A. 5892. For in-stance, Plaintiffs’ experts included a pro-motion won by an external candidate intheir pool—even though this case only con-cerns internally filled promotions. Theyfurther overlooked seven selections of


black employees for promotions. See J.A.5891.

The district court did not clearly err indiscrediting this incomplete work anddeeming it unworthy of evidentiary weight.

5.

a.

To further understand why Plaintiffs’statistical evidence is problematic, it helpsto consider how it came about. In discov-ery, Nucor produced bidding packets andother promotion-related applicant datacovering certain promotions from January2001 to February 2006. Plaintiffs’ analysisof the 2001–2006 data indicated that theblack selection rate fell only 0.84 standarddeviations from the mean—a statisticallyinsignificant result. See J.A. 5872. Fortu-nately for Plaintiffs, the district court lim-ited the use of the actual data to theJanuary 2001 to December 2003 period.But an analysis of that period’s data didnot produce a statistically significant dis-parity, either. At best, analysis of the2001–2003 data produced disparities fallingonly 1.53 standard deviations from themean. See J.A. 1449.

Left with no results from actual recordsthat suggested discrimination, Plaintiffs’experts set about creating extrapolated‘‘benchmark’’ figures for promotions bid-ding between December 1999 and January2001. They began by using so-called‘‘change-of-status’’ forms plucked frompersonnel records to identify 27 purportedpromotions during the period. The ex-perts then constructed a hypothetical bid-ding pool by essentially guessing that bid-ders in early years were racially identicalto bidders in later ones. See J.A. 1162.With their theoretical promotion and bidfigures established, Plaintiffs’ experts thencalculated an expected black promotionrate and compared it to the ‘‘actual’’ blackpromotion rate for the same period. Tied

with the actual promotions figures from2001 through 2003, Plaintiffs’ extrapolatedfigures produced the number on which themajority now relies—2.54 standard devia-tions.

b.

Plaintiffs’ experts, however, based theirextrapolations on several erroneous as-sumptions that render their model unrelia-ble.

It begins with the change-of-statusforms, which Nucor used to record anychange of employee status. Because theforms also recorded demotions, pay in-creases, reassignments, and transfers, onecannot and should not assume that everyform reflects a posted promotion. But upto the time that the district court decerti-fied the promotions classes, Plaintiffs hadnever provided the 27 relevant change-of-status forms to the district court. Quiteunderstandably, the district court wantedmore concrete assurance that Plaintiffs’selected forms showed actual promotions.The district court never got that assur-ance, and it was ‘‘not inclined’’ to ‘‘take[Plaintiffs’] word for it.’’ J.A. 10943.Plaintiffs did eventually submit the 27 rel-evant change-of-status forms—but only af-ter the district court decertified the pro-motions classes. As it turns out, thoseforms do little to dispel the concern thatPlaintiffs misidentified promotions. Forexample, two forms seem to show trans-fers, not promotions, J.A. 11006 (Reyn-olds), 11028 (Forsell), while another justreflects training, J.A. 11029 (Green). Oth-ers do not involve pay raises, suggestingno promotion occurred. See J.A. 11006(Haselden), 11030 (Cooper). Certain otherforms are ambiguous, failing to indicatewhether pay rates changed or what thenature of the position change was. See,e.g., J.A. 11022 (Anderson), 11024 (Pros-kine), 11025 (Pope). Most of the forms fail


to indicate whether Nucor posted the rele-vant opening for bidding. See, e.g., J.A.11006–15, 11019–21, 11023, 11026–32. So,the district court was reasonably con-cerned that the 27 purported promotions—representing nearly half of the promotionsin Plaintiffs’ statistical analysis—were sus-pect and statistically useless.

The problems with Plaintiffs’ experts’model continue to mount when the hypoth-esized bidding pools for the purported pro-motions are examined. Plaintiffs’ expertshypothesized that at least one black em-ployee bid on each of the 27 assumedpromotion opportunities. But that ap-proach rejects the prospect of an all-whitebidding pool during the projected period,something likely to randomly happen fromtime to time given Nucor’s 11% blackworkforce. Consequently, Nucor’s expertconcluded that Plaintiffs’ experts’ model‘‘overstat[ed] the expected number of Afri-can American selections’’ between Decem-ber 1999 and January 2001, as the modelvery likely inflated the number of blackbidders. J.A. 5912. And indeed, Plain-tiffs’ experts calculated that black workersapplied to jobs at a substantially higherpace than their actual percentage of theworkforce, further suggesting some degreeof inflation. Compare J.A. 1157 (notingthat workforce was ‘‘11.3% African–Ameri-can’’), with J.A. 1162 (‘‘The racial composi-tion of the bidders TTT was 19.24% Afri-can–American.’’).

An ‘‘inflated pool’’ like the one thatPlaintiffs used ‘‘can undermine the validityof a statistical study to determine imba-lances.’’ Smith v. Va. CommonwealthUniv., 84 F.3d 672, 677 (4th Cir.1996).When a statistical model overestimates thenumber of black bidders, for instance, thenblack bidding rates artificially rise andblack selection rates artificially fall.These effects might explain, for instance,why the black bidder selection rate for

January 2001 to December 2003—when ac-tual data was available—was three timeshigher than the calculated selection ratefor December 1999 to January 2001—whenprojected data was used. If, during theprojected period, the hypothesized numberof black bidders in the pool (artificially)rose while the number of black bidderselections stayed the same, then the hy-pothesized black selection rate would be(artificially) driven down during the pro-jected period.

c.

The majority nevertheless dubs the ex-trapolated data ‘‘sound.’’ Maj. op. at 903.That conclusion, however, reflects an un-willingness to confront genuine concernsover statistical validity.

For instance, although admitting thatthe change-of-status forms are ambiguous,the majority blames Nucor for not explain-ing how these ambiguities would affectPlaintiffs’ statistical accuracy. Maj. op. at905. That burden was not Nucor’s. Cf.Cooper v. Smith & Nephew, Inc., 259 F.3d194, 199 (4th Cir.2001) (noting that the‘‘proponent of the testimony’’ bears theburden of proving that it is reliable). Re-cently, for example, the Court affirmed adistrict court’s refusal to consider statisti-cal evidence offered to show disparate im-pact because the evidence contained anumber of ‘‘mistakes and omissions’’ in itsanalysis. EEOC v. Freeman, 778 F.3d463, 467 (4th Cir.2015). The Court did soeven though the plaintiff there raised thevery same argument that the majority nowembraces: that the employer never‘‘show[ed] that correcting the errors wouldnegate the disparate impact.’’ Brief forAppellant at 26, Freeman, 778 F.3d 463(No. 13–2365), 2014 WL 320746. TheCourt appropriately rejected that argu-ment then; it should have done the samenow.


Rather than focusing on the reliability ofthe extrapolated statistics, the majorityprefers to revisit the Brown I dissent. Seemaj. op. at 904–05. That dissent notedsome of the concerns mentioned here: notall change-of-status forms used to extrapo-late openings reflect promotions, manyforms are unclear, and few forms indicatewhether positions were posted. SeeBrown I, 576 F.3d at 168 (Agee, J., concur-ring in part and dissenting in part). Toillustrate these concerns, the dissent exam-ined ‘‘the change-of-status forms found inthe record for 2000.’’ Id. Bear in mindthat, at least up to that point, Plaintiffshad never produced the particular change-of-status forms that they relied upon toguesstimate their statistics. Nor had theyinformed the Court that the forms in therecord were not those upon which theybased their statistical evidence. So, theBrown I dissent used the only change-of-status forms that were available to assesswhether they could credibly support Plain-tiffs’ alleged statistical disparities. Id. Al-though the majority labels this exercise‘‘sua sponte fact-finding,’’ maj. op. at 905,the discussion in the Brown I dissent con-sisted of nothing more than explication byexample.

The majority then attempts to tie thedistrict court’s decertification decision tothe ‘‘error’’ that the majority mistakenlyidentifies in the Brown I dissent. Accord-ing to the majority, the district court com-mitted ‘‘clear factual error’’ by assumingthat the change-of-status forms discussedin the Brown I dissent were those thatPlaintiffs relied upon to build their statisti-cal model. But here’s the rub: the districtcourt expressly disclaimed that very as-sumption. The district court noted that, atthe time of decertification, Plaintiffs stillhad not produced the relevant forms. So,it had ‘‘never seen the 27 change-of-statusforms upon which [Plaintiffs’] experts ap-parently relied.’’ J.A. 10943. Thus, the

district court cited the Brown I dissentonly to emphasize the potential problemsinherent in using the forms and why itneeded to see them. See J.A. 10942–43.The majority’s protracted discussion of theBrown I dissent therefore does nothing torehabilitate Plaintiffs’ evidence, resting asit does on a twofold misreading of theBrown I dissent and the district court’sdecertification decision.

Nor does the majority explain why in-flated black bidding rates can be excused.Rather than address that obstacle, the ma-jority assures the reader that the problemcauses only ‘‘an incremental reduction inprobative value’’ that does not ‘‘fatally un-dermine the probativeness of the experts’findings.’’ Maj. op. at 906. But it is hardto minimize these defects so quickly whenPlaintiffs’ experts offered few explanationsfor their assumptions or any assessment ofthe expected impact of those assumptions.The experts did not say, for instance,whether black bidding rates varied duringthe years for which data was available. Ifthey had shown that the rates remainedsteady, then one might assume that thosesame rates applied to the extrapolatedyears. But if the rates varied, then Plain-tiffs’ experts’ assumptions are not sustain-able. Oddly, the majority again blamesNucor for not summoning any evidencegoing to variation, but that tack once morereverses the burden of proof. ‘‘It is theplaintiffs’ burden to demonstrate compli-ance with Rule 23,’’ not Nucor’s. EQTProd., 764 F.3d at 358. The majority fur-ther finds that Plaintiffs’ experts reason-ably assumed that ‘‘every’’ position wasposted for bidding. But Plaintiffs them-selves submitted testimony identifying sev-eral unposted positions. See, e.g., J.A.1010, 1051, 1091, 1110. Nucor’s stated pol-icies also indicated that, at least for a time,‘‘[v]acant supervisory positions [were] not[to] be posted for bidding.’’ J.A. 257.


The majority stresses that, as a generalmatter, plaintiffs may employ extrapolateddata to prove discrimination. Maj. op. at904. That can be true in some cases, butextrapolated data must still be statisticallyvalid. And the majority ignores a signifi-cant and telling distinction between thiscase and past ones: Plaintiffs’ experts ex-trapolated two data points—the composi-tion of the applicant pool and the successrates—whereas experts in our prior casesonly extrapolated one data point. SeeLewis, 773 F.2d at 568; United States v.Cnty. of Fairfax, Va., 629 F.2d 932, 940(4th Cir.1980).

The majority’s cited cases also involveddefendants who wrongfully destroyed rele-vant evidence. See Lewis, 773 F.2d at 568(noting that the defendant ‘‘improperly dis-posed’’ of applicant records); Cnty. ofFairfax, 629 F.2d at 936 n. 4 (noting thatthe defendant destroyed applicant data‘‘[i]n violation of the record keeping regu-lations of [two statutes]’’). In a situationinvolving spoliation of evidence, the Courtcommonly draws adverse inferencesagainst the spoliators. But this recordcontains no evidence of spoliation.

Regardless, no authority requires thedistrict court to find extrapolated data con-vincing in every case. Our precedentholds just the opposite. In Allen v. PrinceGeorge’s County, 737 F.2d 1299, 1306 (4thCir.1984), for example, the district courtrelied solely upon actual applicant flowdata ‘‘to the exclusion of all [other] statisti-cal evidence,’’ including evidence craftedfrom alternative benchmarks. We af-firmed, emphasizing that we could not‘‘second-guess’’ a fact-bound decision con-cerning ‘‘the relative weights to be accord-ed to the parties’ respective evidence.’’ Id.The district court here did essentially thesame thing as the district court in Allen,giving weight for good reason to the actualdata available to the exclusion of the spec-

ulative extrapolation evidence. As in Al-len, we should not say that the districtcourt clearly erred in doing so.

6.

a.

Plaintiffs’ statistical evidence also doesnot apply controls for non-discriminatoryfactors that could very well have causedany observed disparities. See Lowery, 158F.3d at 764. Seniority, for instance, influ-ences promotions decisions at Nucor. See,e.g., J.A. 257. Disciplinary issues also ledNucor to reject certain applicants for pro-motion—including frequent bidder JasonGuy, who is black. See J.A. 659–67; seealso Coates v. Johnson & Johnson, 756F.2d 524, 544 n. 20 (‘‘[A]n employee’s priordiscipline record seems likely to be a ma-jor, if not the most important, factor in [anemployment] decision.’’). But Plaintiffs’experts admitted that they did not controlfor these or any other ‘‘additional factorsbeyond the control for each job posting.’’J.A. 1164. The majority would wish theseconsiderations away, reasoning that Nucornever raised them. But Nucor’s expertnoted the need to ‘‘control for characteris-tics that would seem to affect the chance ofselection,’’ which would include matterslike seniority and discipline. See J.A.5893. Anyway, we could have affirmed thedistrict court’s decision here on ‘‘any basissupported by the record.’’ Defenders ofWildlife v. N.C. Dep’t of Transp., 762 F.3d374, 392 (4th Cir.2014).

The majority also tries to summon itsown justifications for these omissions, im-plying that records were not available tocontrol for matters like discipline. Maj.op. at 906. Even Plaintiffs’ experts con-ceded that they were. See J.A. 1165 (ac-knowledging that Nucor had maintainedand produced ‘‘bidders’ training, discipline,and bidding records’’); see also J.A. 5893(Nucor’s expert observing that ‘‘separate


discipline and training files [were] provid-ed to Drs. Bradley and Fox and [him]’’).And, based on allegations and personalassessments from Plaintiffs themselves,the majority assumes that potential ex-planatory variables are themselves raciallybiased. See maj. op. at 906–07. Yet hereagain, Plaintiffs’ experts do not assume so,perhaps because there is no concrete evi-dence of such taint in the record. SeeOttaviani v. State Univ. of N.Y. at NewPaltz, 875 F.2d 365, 375 (2d Cir.1989)(holding that district court correctly re-quired the plaintiffs to account for poten-tial explanatory variable where the plain-tiffs alleged but did not prove that thevariable was biased). And even if onewere to indulge the majority’s assumptionthat discipline at Nucor was itself biased,that outcome would not justify excludingthe variable from the statistical modelcompletely. ‘‘[T]ainted variables shouldnot be routinely excluded from the regres-sion equation. Instead, the effects of theinclusion of a tainted variable must beassessed and minimized.’’ Paetzold &Willborn, supra, § 6:13. The majority’sreasons, then, do not fill the gaps in Plain-tiffs’ experts’ work.

The failure to control for non-race-relat-ed explanatory variables ‘‘is sufficientlyserious so as to weaken the statisticalstudy’s probativeness.’’ Lowery, 158 F.3dat 764; see also Smith, 84 F.3d at 676;accord Rodriguez v. Nat’l City Bank, 726F.3d 372, 384–85 (3d Cir.2013); Morgan v.United Parcel Serv. of Am., Inc., 380 F.3d459, 468 (8th Cir.2004); Munoz v. Orr, 200F.3d 291, 301 (5th Cir.2000); Sheehan v.Daily Racing Form, Inc., 104 F.3d 940,942 (7th Cir.1997); Penk v. Or. St. Bd. ofHigher Educ., 816 F.2d 458, 465 (9th Cir.1987). A trier of fact must determinewhether racial discrimination—rather thanchance or some other ‘‘confounding fac-tor[ ]’’—caused an alleged disparity. In reNavy Chaplaincy, 738 F.3d 425, 429

(D.C.Cir.2013). Only a controlled modelcan provide that answer, and Plaintiffs’experts’ evidentiary model did not meetthat definition.

b.

In most every employment case, a validstatistical model must account for one par-ticularly important explanatory variable:the applicant pool’s qualifications. ‘‘[T]herelevant comparison is between the per-centage of minority employees and thepercentage of potential minority applicantsin the qualified labor pool.’’ Carter v.Ball, 33 F.3d 450, 456 (4th Cir.1994); seealso City of Richmond v. J.A. Croson Co.,488 U.S. 469, 501–02, 109 S.Ct. 706, 102L.Ed.2d 854 (1989); McNairn v. Sullivan,929 F.2d 974, 979 (4th Cir.1991). If courtswere to accept statistical models contain-ing unqualified applicants, then employerscould be punished merely because of a‘‘dearth of qualified nonwhite applicants(for reasons that are not [the employers’]fault).’’ Wards Cove Packing Co. v. Ato-nio, 490 U.S. 642, 651, 109 S.Ct. 2115, 104L.Ed.2d 733 (1989). Thus, ‘‘statisticsbased on an applicant pool containing indi-viduals lacking minimal qualifications forthe job [are] of little probative value.’’Watson, 487 U.S. at 997, 108 S.Ct. 2777;see also Paetzold & Willborn, supra, § 4:3(‘‘[W]hen considering potential discrimina-tion in promotions within an organization,only employees qualified for promotionshould be considered in the proxy pool.’’).Furthermore, ‘‘[n]o rational enterprisethat has several qualified candidates for aposition selects among them by lot; itpicks the best qualified.’’ Mason v. Cont’lIll. Nat’l Bank, 704 F.2d 361, 364 (7thCir.1983). So, a truly effective statisticalmodel will not just account for minimumqualifications, but should control for thevariations in skills even among minimallyqualified applicants.


By this point, Plaintiffs and their ex-perts should have known better than toignore other explanatory factors. In arelated case challenging promotions prac-tices at a different Nucor facility, theEighth Circuit found that similarly sub-standard work from the same expert didnot create a triable question of fact onsummary judgment. See Bennett, 656 F.3dat 812. In so holding, the Eighth Circuitemphasized that the expert’s statistics had‘‘little force’’ because they ‘‘assumed thatall applicants were qualified for promotionto each available position.’’ Id. at 818.The Eighth Circuit is not alone. Othercourts have criticized Plaintiffs’ principalexpert for employing his ‘‘warm body hy-pothesis,’’ which ‘‘assumes that every per-son is just as qualified and skilled andexperienced as everyone else.’’ Davis v.Ala. Dep’t of Educ. Dep’t of DisabilityDetermination Serv., 768 F.Supp. 1471,1477 (N.D.Ala.1991); accord Adams v.Austal, U.S.A., L.L.C., No. 08–00155–KD–N, 2011 WL 1558790, at *8 (S.D.Ala. Apr.25, 2011); Rollins v. Ala. Cmty. Coll. Sys.,No. 2:09cv636–WHA, 2010 WL 4269133, at*8–9 (M.D.Ala. Oct. 25, 2010); Bennett v.Nucor Corp., No. 3:04CV00291 SWW, 2007WL 2333193, at *3 (E.D.Ark. Aug. 13,2007); Yapp v. Union Pac. R.R. Co., 229F.R.D. 608, 619 (E.D.Mo. Aug.5, 2005);Rhodes v. Cracker Barrel Old CountryStore, Inc., No. Civ.A. 4:99–CV–217–H,2002 WL 32058462, at *65 (N.D.Ga. Dec.31, 2002). We even affirmed a districtcourt’s choice to exclude work from thesame expert precisely because he did notincorporate adequate controls. SeeAnderson, 406 F.3d at 262–63 (agreeingwith the district court’s view that the ex-pert had ignored ‘‘actual job performanceor job requirements’’ even though he ‘‘con-ceded’’ that he could have ‘‘use[d] a controlfactor that would control for the actual jobtitle or the job duties’’).

Plaintiffs’ experts assumed that all per-sons in each bidding pool were equallyqualified because ‘‘only persons who decid-ed to bid based on the posted qualificationswere included.’’ J.A. 1162. This opaquelanguage obscures another faulty assump-tion built into the model: the experts as-sumed that only qualified persons appliedfor each promotion opportunity. It takesno expertise to comprehend that some peo-ple ‘‘might be discouraged from applyingbecause of a self-recognized inability tomeet the [opening’s] standards.’’ Dothardv. Rawlinson, 433 U.S. 321, 330, 97 S.Ct.2720, 53 L.Ed.2d 786 (1977). But onecould hardly assume that every job appli-cant is so discerning, and even the majori-ty seems unwilling to make that assump-tion. See maj. op. at 906. The majorityprefers to guess that the number of un-qualified applicants will be so triviallysmall as to be statistically irrelevant, and itmakes that guess simply because the jobannouncement includes job requirements.In practical effect, the majority has readthe ‘‘qualified applicants’’ limitation foundin our prior cases out of the law, as mostevery job opening provides some minimaldescription of what skills are required.

‘‘A statistical study that fails to correctfor explanatory variables, or even to makethe most elementary comparisons, has novalue as causal explanation[.]’’ PeopleWho Care v. Rockford Bd. of Educ., 111F.3d 528, 537 (7th Cir.1997). Plaintiffspresented just such a study here, and thedistrict court did not clearly err in reject-ing it.

7.

Lastly, Plaintiffs’ statistical evidence im-properly aggregates data in a way thatdistorts the results.

a.

The objective in a class action—even ina proceeding that alleges disparate treat-


ment—is to identify a common, uniformpolicy. ‘‘While in a case alleging intentionaldiscrimination, such as this one, a plaintiffneed not isolate the particular practice andprove that such practice caused the dis-crimination, plaintiffs must make a signifi-cant showing to permit the court to inferthat members of the class suffered from acommon policy of discrimination that per-vaded all of the employer’s challenged em-ployment decisions.’’ Love v. Johanns,439 F.3d 723, 728 (D.C.Cir.2006).

Thus, if the class challenges a policyimplemented at the nationwide level, thenplaintiffs might use applicable statisticsshowing nationwide disparities to establishthe policy’s effects. Conversely, if theclass challenges policies implemented on aplant-by-plant or department-by-depart-ment basis, then the class must summonstatistics showing disparities at that level.Otherwise, non-uniform decisions made byone discriminatory decisionmaker mightcreate disparities that, when aggregatedwith other, neutral decisions, misleadinglyindicate discrimination across the wholegroup of decisionmakers.

Wal–Mart demonstrates these conceptswell. There, the plaintiffs offered statis-tics purporting to show regional and na-tional disparities in employment decisionsat Wal–Mart. Wal–Mart, 131 S.Ct. at2555. Those decisions, however, weremade at the store level. Id. at 2547. Be-cause of that disconnect, the SupremeCourt held that plaintiffs’ statistics did notestablish a common policy. Once again,the broader disparities might have been‘‘attributable only to a small set of Wal–Mart stores’’ and did not ‘‘establish theuniform, store-by-store disparity uponwhich plaintiffs’ theory of commonality de-pend[ed].’’ Id. at 2555. In essence, Wal–Mart agreed with our own, earlier casesindicating that statistics should not be ag-gregated together to create disparities

that are not actually representative of theclass as a whole. Compare Stastny, 628F.2d at 279–80 (requiring the plaintiffs’statistics to focus on the ‘‘locus of autono-my’’), with Elizabeth Tippett, Robbing aBarren Vault: The Implications of Dukesv. Wal–Mart for Cases Challenging Sub-jective Employment Practices, 29 HofstraLab. & Emp. L.J. 433, 447 (2012), citedwith approval by Scott v. Family DollarStores, Inc., 733 F.3d 105, 113 (4th Cir.2013) (explaining that Wal–Mart requiresthat plaintiffs’ statistics focus on ‘‘the locusof the subjective decision-making’’).

In requiring the plaintiffs’ statistics tobe centered at the level of relevant deci-sionmaking, Wal–Mart did not distinguishbetween nationwide and other class ac-tions. Rather, Wal–Mart asked whetherthe plaintiffs there were too dissimilar tobring their claims together, regardless ofhow many claims there might be. Thus,courts have applied principles from Wal–Mart in cases involving classes of roughlythe same size as the class at issue here.See, e.g., Wang v. Chinese Daily News,Inc., 737 F.3d 538, 544 (9th Cir.2013) (200class members); Ealy, 514 Fed.Appx. at304–08 (150 class members). Even statis-ticians agree that Wal–Mart reachesclasses big and small. See, e.g., Dr. MaryDunn Baker, Class Certification Statisti-cal Analysis Post–Dukes, 27 ABA J. Lab.& Empl. L. 471, 479 (2012) (‘‘[T]he size ofthe putative class or the number of estab-lishments the defendant operates will havelittle to do with whether the Dukes com-monality approach is applicable.’’). So,even though Plaintiffs here challenge prac-tices in one plant, they still must offerstatistics showing disparities among all therelevant decisionmakers, regardless of thatone-plant focus. See Rubenstein, supra,§ 24:40 (‘‘Courts have certified [only] limit-ed classes when the facts show that nouniform personnel policies are applied


among the various plants, departments, orlevels of employees.’’).

b.

Here, as the Brown I majority agreed,the evidence indicates ‘‘that each depart-ment manager’’ in each of Nucor’s sixproduction departments ‘‘has unbridleddiscretion to make promotions within hisdepartment utilizing whatever objectiveor subjective factors he wishes.’’ BrownI, 576 F.3d at 151. Department manag-ers took full advantage of that discretion,developing processes that they recurrent-ly characterized as unique and indepen-dent. See J.A. 7887, 7894–95, 7900,7906–07. Indeed, these processes wereso varied that one supervisor declaredthat he had ‘‘no idea what other depart-ments d[id].’’ J.A. 8109. Even the deci-sionmakers varied. In some depart-ments, such as the hot mill and shippingdepartments, supervisors and the depart-ment managers made promotion deci-sions. In other departments, such asmaintenance and the cold mill, pro-motions decisions were a more collabora-tive effort involving even lower-level leadmen. These different decisionmakersthen applied different standards. In thebeam mill, for example, the process cen-tered upon interviews alone. In contrast,the melt shop looked to applicants’ workhistory, safety record, psychological inter-view, job skills, training, attendance, andscores on a job-specific aptitude test.Nucor’s general manager quite reason-ably described the promotions processeswhen he said that ‘‘each departmentha[d] their own way of doing [pro-motions].’’ J.A. 1723.

Plaintiffs’ own expert found that eachdepartment had its own procedures, and atleast eight different criteria—not including‘‘numerous other idiosyncratic factors’’—might or might not be considered in mak-

ing any employment decision. J.A. 1518–19. ‘‘Different supervisors,’’ he explained,‘‘utilized different criteria weightingschemes with little consistency among theselection officials and among the differenthiring/promotion/transfer opportunities.’’J.A. 1525. Taking all this dissimilaritytogether, the expert concluded that Nu-cor’s selection process was only ‘‘consistentin its inconsistency.’’ J.A. 1519.

Yet Plaintiffs’ statistical evidence incor-rectly assumed the exact opposite: perfect,plant-wide consistency as to promotions.Given that promotions decisions weremade at the department or supervisor lev-el using different and independent criteria,we cannot rightfully assume that a plant-wide disparity resulted from a uniformproblem arising in the same way in eachNucor department. See Wal–Mart, 131S.Ct. at 2555. Put differently, the districtcourt reasonably found that the ‘‘locus ofautonomy’’ rested at the departmental lev-el, not a plant-wide one. We cannot thenassume that department decisions weremade in lockstep, such that plant-wide dis-parities necessarily reflect common, de-partmental ones. See Bolden v. WalshConstr. Co., 688 F.3d 893, 896 (7th Cir.2012) (rejecting aggregate data because itdid not necessarily imply that ‘‘all 25 su-perintendents behaved similarly, so itwould not demonstrate commonality’’).

We have already seen these conceptsplay out in another employment discrimi-nation action involving a similar Nucorfacility. Applying Wal–Mart, the EighthCircuit rejected statistics—from the sameexpert—that reflected plant-wide dispari-ties in promotions at an Arkansas Nucorplant. Bennett, 656 F.3d at 815–16. Justas in this case, the statistical evidencethere indicated that different departmentsin the plant applied different criteria forpromotions decisions. Id. at 815. Theplant-wide evidence therefore ‘‘ha[d] little


value in the commonality analysis’’ becauseit ‘‘did not differentiate between the hiringand promotion decisions made in each de-partment.’’ Id. The Eighth Circuit foundthat, in those sorts of circumstances, ‘‘abottom-line analysis [wa]s insufficient todemonstrate that any disparate treatmentor disparate impact present in one depart-ment was also common to all others.’’ Id.at 815–16.

As in Bennett, Nucor here provided itsown analysis that demonstrated how thestatistical disparities varied among the dif-ferent departments in the plant. Nucor’sexpert measured how selection rates var-ied between white and black applicants ona department-by-department basis overthe period for which bidding informationwas available. With proper controls ap-plied, the expert found that race differ-ences between departments could vary byas much as 2.44 standard deviations. J.A.5894. In other words, some departmentsexperienced decidedly smaller disparitiesin selection rates, undermining any infer-ence of uniformity and commonality amongall departments.

Given the wide variance in promotionspractices at the Nucor facility, the districtcourt did not clearly err in rejecting astatistical study that failed to account forthat variance.

c.

The majority finds, however, that Nu-cor’s entire plant should be treated ‘‘as asingle entity’’ when it comes to promotionsdecisions. Maj. op. at 911 (alluding toBrown I, 576 F.3d at 158). Although themajority suggests otherwise, Brown I didnot decide this issue. Brown I held thatthe district court should treat Nucor’s var-ious production departments as a singlefacility only for purposes of Plaintiffs’ hos-tile work environment claim. 576 F.3d at158 (‘‘[T]he affidavits of employees in one

department are admissible to prove aplant-wide hostile environment that affect-ed employees in other departments, andthe plaintiffs have satisfied the commonali-ty requirement for their hostile work envi-ronment claim.’’ (emphasis added)); seealso id. at 157 (discussing how a ‘‘hostileenvironment determination’’ must be madein the context of discussing Plaintiffs’ ‘‘sin-gle entity’’ argument). It said nothingabout the uniformity of promotions deci-sions across the plant. Id. The Brown Imajority did so because Plaintiffs likewisefocused their ‘‘single entity’’ argument ononly the hostile work environment claim.See Brief for Appellant at 25–35, Brown I,576 F.3d 149 (No. 08–1247), 2008 WL2307453. Thus, as with predominance, thedistrict court was not constrained in decid-ing the ‘‘single facility’’ issue, as no BrownI mandate existed as to that issue.

Nonetheless, the majority concludes thatfacts establishing a single hostile work en-vironment claim also establish a commonpromotions policy. Maj. op. at 911–12.Yet ‘‘[d]isparate treatment TTT is inherent-ly different from hostile work environ-ment. The federal courts treat the twotypes of cases differently for good reason.’’See Pollard v. E.I. DuPont de NemoursCo., 213 F.3d 933, 943 (6th Cir.2000), rev’don other grounds, 532 U.S. 843, 121 S.Ct.1946, 150 L.Ed.2d 62 (2001). And no courthas held that a common hostile work envi-ronment establishes that a facility must betreated as a single entity for purposes ofevery other kind of employment discrimi-nation claim.

In finding a common environment,Brown I focused on shared locker roomsand spaces, plant-wide email, and plant-wide radio systems. 576 F.3d at 158.When it comes to a hostile work environ-ment claim, those facts may matter: racialslurs and ‘‘monkey noises’’ uttered in acommon space or transmitted via plant-


wide radio can affect whoever hears them.See Burlington Indus., Inc. v. Ellerth, 524U.S. 742, 762, 118 S.Ct. 2257, 141 L.Ed.2d633 (1998) (‘‘[A]nyone who has regular con-tact with an employee can inflict psycho-logical injuries by his or her offensive con-duct.’’). But locker rooms and radios bearno relationship to promotions decisions;certainly nothing in the record supportssuch a concept. Only supervisors can in-flict the ‘‘pain’’ of a denied promotion, andthey can do so only when empowered bycompany structure, not common spaces.We should not assume that dozens of su-pervisors acted in concert merely becausetheir employees might have changedclothes in the same room. Nor should weassume—in the face of expressly differentcriteria applied to different groups of em-ployees—that applicants in each depart-ment nevertheless suffer the same injurymerely because of their physical proximityto one another at some point during aworkday. Though the majority insiststhat ‘‘centralized, circumscribed environ-ments’’ will ‘‘generally’’ increase ‘‘consis-tency’’ in managerial decisionmaking, maj.op. at 910, Plaintiffs’ own expert madeclear that this hypothesized general rulecannot apply here, see J.A. 1519 (‘‘Thebest sentiment I can muster in favor of the[Nucor] selection procedure is that it isconsistent in its inconsistency.’’). See also,e.g., Tabor v. Hilti, Inc., 703 F.3d 1206,1229 (10th Cir.2013) (affirming denial ofclass certification where ‘‘Plaintiffs chal-lenge[d] a highly discretionary policy forgranting promotions’’).

The majority also notes that the generalmanager formally approved promotions inthe plant. Maj. op. at 917. Without say-ing so explicitly, the majority seems topropose that the general manager provid-ed some common, plant-wide direction thatdrove common, plant-wide disparities. Yeteven the Brown I majority recognized thatthe general manager played no genuine

role in the promotions decisionmaking pro-cess. 576 F.3d at 152 (‘‘Although, by poli-cy, the plant’s general manager approvesall promotions and handles discriminationand harassment investigations, the recordsuggests that each department managerhas unbridled discretion to make pro-motions within his department utilizingwhatever objective or subjective factors hewishes.’’). The evidence confirms thatproposition. Promotions, the general man-ager explained, were ‘‘not [his] area ofresponsibility,’’ as he had ‘‘departmentmanagers that ma[d]e those decisions.’’J.A. 8163. Nucor instead trained its de-partment managers to make promotionsdecisions and implement the anti-discrimi-nation policy.

The majority nevertheless says the gen-eral manager engaged in ‘‘inaction.’’ Maj.op. at 912, 917. The majority’s theory—premised on an assumed culture of ‘‘odiousracism’’ and passive enabling—resembles atheory that Wal–Mart out-and-out reject-ed. See 131 S.Ct. at 2553–54 (refusing tocredit evidence asserting that a ‘‘strongcorporate culture,’’ enabled by policies ofdiscretion, permitted bias in pay deci-sions); accord Davis v. Cintas Corp., 717F.3d 476, 489 (6th Cir.2013).

Even if one assumes that such a theorywere viable and relevant here, it would notprove commonality. ‘‘Inaction’’—lettingsupervisors do as they wish—is just dis-cretion by another name. ‘‘[I]t is a policyagainst having uniform employment prac-tices.’’ Wal–Mart, 131 S.Ct. at 2554.‘‘Wal–Mart tells us that local discretioncannot support a company-wide class nomatter how cleverly lawyers’’ (or judges)‘‘may try to repackage local variability asuniformity.’’ Bolden, 688 F.3d at 898; ac-cord In re Navy Chaplaincy, No. 1:07–mc–269 (GK), 306 F.R.D. 33, 52, 2014 WL4378781, at *15 (D.D.C. Sept. 4, 2014).Were it otherwise, one could find a com-


mon policy in most every case, as mostevery company has a management head atthe top that could be accused of not doingenough. Beyond that, Plaintiffs’ expertsnever traced their identified disparities tothe general manager, and their reportsnever even mention him. For good rea-son. Individual acts of discretion, not thegeneral manager’s purported acquiescence,would have caused any disparities and theinjuries that they reflect. Thus, the not-very-common common policy does notpresent a common injury.

Nucor also used a plant-wide ‘‘dual-ap-proval’’ scheme, under which promotionsrequired approval from both ‘‘originating’’and ‘‘destination’’ department heads. Themajority sees this as a case of potential‘‘cat’s paw’’ liability, wherein a non-deci-sionmaker influences the ultimate decision-maker’s choice in a discriminatory way.Maj. op. at 911–12 (citing Smith v. Bray,681 F.3d 888, 897 & n. 3 (7th Cir.2012)).But nothing other than speculation indi-cates that dual approval was used to effectdiscrimination in any common way, andany cat’s paw must be the ‘‘proximatecause’’ of the discriminatory harm to beactionable. Staub v. Proctor Hosp., 562U.S. 411, 131 S.Ct. 1186, 1192, 179 L.Ed.2d144 (2011). Not even Plaintiffs’ statisticalexperts attempt to tie their disparities to adual-approval policy.

The majority surmises that a discrimina-tory supervisor in one department couldhave theoretically used dual approval toinflict his animus upon employees outsidehis own department. But if a racist de-partment head had tried to use the dual-approval scheme to disadvantage blackworkers, he would not have been able toreach all or even most of the promotionsdecisions in the plant, dual approval not-withstanding. A discriminatory depart-ment head in the beam mill, for instance,would have had no say when it came to a

cold mill employee seeking a higher posi-tion within the cold mill, hot mill, meltshop, maintenance department, or shippingdepartment. Perhaps, then, the majority’sconcept—if properly supported with evi-dence—might justify a class of personsapplying in and out of a particularly pro-blematic department. In fact, the districtcourt proposed certifying just such a classas to the beam mill. See J.A. 10953–54 &n.16. But it would not justify the plant-wide class action that Plaintiffs now meanto bring. Cf. Ellis, 657 F.3d at 983 (‘‘Adisparity in only 25% of the regions, how-ever, would not show that discriminationmanifested in promotions practices in thesame general fashion.’’).

* * *

In sum, the district court did not clearlyerr in choosing not to rely on Plaintiffs’statistical evidence. Faced with evidencebased on questionable data, uncontrolledexplanatory variables, and poorly struc-tured methodologies, the district court didnot act irrationally in determining thatsuch evidence was of negligible credence.The ‘‘troubling effects of statistical infer-ences require thoughtful consideration ineach case,’’ Mister v. Ill. Cent. Gulf R.R.Co., 832 F.2d 1427, 1437 (7th Cir.1987), andthat consideration is sorely lacking fromthe work of Plaintiffs’ experts. Thus, Plain-tiffs’ evidence, with its many deficiencies,does not establish the common policy nec-essary for class certification. The districtcourt did not abuse its discretion in mak-ing that finding.

B. Anecdotal Evidence

Plaintiffs also present affidavits fromsixteen employees in support of certifyingthe promotions classes. The district courtdid not abuse its discretion in refusing tocertify Plaintiffs’ proposed class based onthis limited evidence.


1.

In their original class certification mo-tion, Plaintiffs never argued that anecdotalevidence, standing alone, could establish acommon policy of discrimination. Rather,Plaintiffs presented the anecdotal evidenceonly to supplement their statistical evi-dence. See Brown I, 576 F.3d at 164(Agee, J., dissenting). The Brown I ma-jority constructed its own theory of thecase, finding that Plaintiffs could in factadvance their case on anecdotal evidence‘‘alone.’’ Id. at 153. Plaintiffs now takeup the Brown I majority’s theory in thisappeal.

Plaintiffs made the better choice in theirinitial offering, as anecdotes only help tellthe story. They are meant to bring ‘‘thecold numbers convincingly to life,’’ Team-sters, 431 U.S. at 339, 97 S.Ct. 1843, pro-viding ‘‘texture’’ for statistical evidence.Robinson v. Metro–North Commuter R.R.Co., 267 F.3d 147, 168 (2d Cir.2001), abro-gated on other grounds by Wal–Mart, 131S.Ct. at 2560–62. But standing alone, ‘‘an-ecdotal evidence TTT [will] rarely, if ever,TTT show a systemic pattern of discrimina-tion.’’ O’Donnell Constr. Co. v. Dist. ofColumbia, 963 F.2d 420, 427 (D.C.Cir.1992); accord Briggs v. Anderson, 796F.2d 1009, 1019 (8th Cir.1986) (observingthat plaintiffs ‘‘punished themselves’’ bychoosing to rely on anecdotal evidence);EEOC v. Bloomberg L.P., 778 F.Supp.2d458, 470–71 & n. 8 (S.D.N.Y.2011) (collect-ing cases); see also Michael Selmi, Theo-rizing Systemic Disparate TreatmentLaw: After Wal–Mart v. Dukes, 32 Berke-ley J. Emp. & Lab. L. 477, 501 (2011)(‘‘[A]necdotal evidence is always of margin-al significance in a pattern or practiceclaim.’’).

In discrimination cases, courts move an-ecdotal evidence to the background be-cause such evidence does not prove much.‘‘Anecdotal reports TTT are ordinarily more

helpful in generating lines of inquiry thanin proving causation.’’ Federal JudicialCenter, Reference Manual on ScientificEvidence 217 (2011). Individual storiessay little, for instance, about the frequencyof an event’s occurrence or the reasons forthat occurrence. Without knowing at leastthose two items, it can hardly be assumedthat the stories reflect a broader trendflowing directly from intentional discrimi-nation. See Wessmann v. Gittens, 160F.3d 790, 805–06 (1st Cir.1998); CoralConstr. Co. v. King Cnty., 941 F.2d 910,919 (9th Cir.1991). Anecdotes are alsomore susceptible to mistaken perception,leading to erroneous conclusions—especial-ly when collections of stories are treatedas quasi-statistics. See Fisher v. VassarColl., 70 F.3d 1420, 1444–45 (2d Cir.1995).And bias can skew anecdotal evidence, aswhen only those who feel most stronglyabout an issue offer anecdotes or when thesoliciting party has a particular objectivein mind. Cf. United States v. Local 560 ofInt’l Bhd. of Teamsters, Chauffeurs, Ware-housemen, & Helpers of Am., 780 F.2d267, 277 (3d Cir.1985) (finding that a sur-vey that was meant to show the ‘‘reputa-tion’’ of a particular organization shouldhave been excluded when it only surveyedpersons known ‘‘to be hostile’’ to the or-ganization). Because ‘‘anecdotes provideno mechanism for assessing truthfulness,typicality, or frequency,’’ courts can andshould question their usefulness, just as‘‘[s]cientists and medical researchers’’ havedone for many years. David A. Hyman,Lies, Damned Lies, and Narrative, 73Ind. L.J. 797, 803 (1998).

2.

The majority finds Plaintiffs’ anecdotalevidence sufficient principally because theratio reflecting the number of affidavitsalleging discrimination compared to thenumber of class members is purportedly


small. Maj. op. at 912–13. As of 2006,Plaintiffs’ experts determined that ‘‘ap-proximately 150 African–Americans’’ com-prised the class. J.A. 1154. Given thatthe class period extends well into 2011, itis reasonable to assume that Nucor hiredadditional black applicants since 2006, con-servatively setting the present class size at160 black employees or more. The sixteenaffidavits that Plaintiffs provide thereforerepresent roughly one affidavit for everyten class members—a weak sample fromthe entire class. ‘‘[A] court must be waryof a claim that the true color of a forest isbetter revealed by reptiles hidden in theweeds than by the foliage of countlessfree-standing trees.’’ Cooper, 467 U.S. at879–80, 104 S.Ct. 2794. When ten percentof a class (or less) complains of mistreat-ment in a discrimination case, a districtcourt does not clearly err in finding thatsuch complaints do not establish a ‘‘stan-dard operating procedure’’ of discrimina-tion, Teamsters, 431 U.S. at 336, 97 S.Ct.1843, ‘‘significant adverse effects’’ on therelevant class, Watson, 487 U.S. at 986,108 S.Ct. 2777, or ‘‘significant proof’’ ofclass-wide discrimination, Wal–Mart, 131S.Ct. at 2553.

3.

What may matter more than the quanti-ty of a plaintiff’s evidence is its quality.If, for instance, the anecdotal evidence isindirect and circumstantial, the districtcourt might justifiably probe whether thatevidence truly gives rise to a necessaryinference of discrimination. After all, ‘‘adistrict court may properly consider thequality of any anecdotal evidence.’’ Rossi-ni v. Ogilvy & Mather, Inc., 798 F.2d 590,604 (2d Cir.1986); accord Eastland v.Tenn. Valley Auth., 704 F.2d 613, 625(11th Cir.1983).

At least as to the promotions-relatedmatters at issue in this appeal, Plaintiffs

do not present compelling anecdotal evi-dence. Byron Turner, for instance, doesnot address promotions at all. Neitherdoes Walter Joseph Cook. In what mightbe an employment law first, Kenneth Hub-bard complains that Nucor promoted him.See J.A. 1097; cf. Kalamazoo Cnty. Rd.Comm’n v. Deleon, ––– U.S. ––––, 135S.Ct. 783, 784, 190 L.Ed.2d 887 (2015)(Alito, J., dissenting from denial of certio-rari) (‘‘Respondent’s supervisors did notviolate federal law by granting him thetransfer that he sought and that they hadno reason to believe he did not want.’’).And Earl Ravenell testifies about a timethat he applied for a promotion and wasnot selected—because another black em-ployee was selected for that opening. Healso tells us that he chose not to apply forany other positions because of ‘‘the look onhis [supervisor]’s face.’’ J.A. 1111. Theseand other examples are not ‘‘cherrypick[ed],’’ maj. op. at 913, but merely offersome insight into why the district courtcould reasonably decide differently thanthe majority does.

Much of the anecdotal evidence alsoamounts to conclusory and speculativestatements of personal belief. For in-stance, even those employees who do men-tion job qualifications rely almost exclu-sively on their personal, subjective, andunsubstantiated views of their own abili-ties. We usually do not give such testimo-ny much, if any, weight. See Williams v.Giant Food Inc., 370 F.3d 423, 433 (4thCir.2004); Evans v. Techs. Applications &Serv. Co., 80 F.3d 954, 960 (4th Cir.1996).Other employees assume racism in theprocess without identifying an objectivefact to support that view. Named plaintiffRamon Roane declares, for example, thathe applied for a position that was ‘‘sudden-ly cancelled because Nucor was not readyfor an African American to hold a supervi-sory position.’’ J.A. 996. Yet he does notexplain how or why he came to that conclu-


sion, and ‘‘[a] plaintiff’s self-serving opin-ions, absent anything more, are insufficientto establish a prima facie case of discrimi-nation.’’ Mackey v. Shalala, 360 F.3d 463,469–70 (4th Cir.2004).

In addition, Plaintiffs’ evidence is oftenso incomplete that it lacks any probativevalue. For example, Bernard Beaufortdiscusses a promotions decision that hebelieves ‘‘was made unfairly.’’ J.A. 6008.But he does not know who eventually re-ceived the job, what his or her race was,‘‘what [the decision] was based on,’’ orwhether ‘‘it was based on [his] race.’’ J.A.6008. Other employees testify about notreceiving promotions, but many of thesedeclarants do not indicate whether theywere minimally qualified for the positionor whether the selected employee was ofanother race. Without these fundamentalfacts, we cannot know whether particularpromotions decisions raise even a circum-stantial inference of discrimination. SeeCline v. Roadway Express, Inc., 689 F.2d481, 485 n. 4 (4th Cir.1982); accord Tex.Dep’t of Cmty. Affairs v. Burdine, 450U.S. 248, 253, 101 S.Ct. 1089, 67 L.Ed.2d207 (1981).

4.

The majority concentrates on one anec-dotal comment from one supervisor in thebeam mill: ‘‘I don’t think we’ll ever have ablack supervisor while I’m here.’’ J.A.1885–86; see also maj. op. at 899, 917.That comment could be compelling evi-dence in a case hinging on decisions madeby that particular decisionmaker. On theother hand, it might not be, as we havediscounted ‘‘stray or isolated’’ remarks,even at summary judgment. Brinkley v.Harbour Recreation Club, 180 F.3d 598,608 (4th Cir.1999); see also Birkbeck v.Marvel Lighting Corp., 30 F.3d 507, 511–12 (4th Cir.1994) (finding that decision-maker’s singular remark did not evidencediscriminatory practices at company).

In the end, the question proves aca-demic. A class-wide claim challenging de-cisions made by many different decision-makers plainly requires something morethan a single comment from just one ofthem. We see this rule—that sparse com-ments are not enough for class treat-ment—illustrated in cases like King v.General Electric Company, 960 F.2d 617(7th Cir.1992). There, the Seventh Circuitfound that the plaintiffs’ anecdotal evi-dence in an age-discrimination case wasnot enough, even though the record con-tained testimony from a higher managerthat the company was ‘‘going to get rid ofthese old farts and get some new blood inhere.’’ Id. at 628 (Cudahy, J., dissenting)(summarizing evidence rejected by the ma-jority). This Court, too, has rejected anec-dotal evidence of a similarly ‘‘damningcharacter,’’ this time in a racial discrimina-tion case. See Coker v. Charleston Cnty.Sch. Dist., No. 92–1589, 1993 WL 309580,at *6 (4th Cir. Aug. 16, 1993). We foundthat the plaintiffs had not established apolicy or practice of discrimination despitetestimony that a black principal was toldthe community would not ‘‘accept’’ him at apredominantly white school. Id. at *4. Allthis goes to illustrate that plaintiffs likelycannot prove a class-wide policy with asingle comment, no matter how bigotedthe comment may be. One comment cer-tainly does not make the showing thatPlaintiffs insist they make here: a com-mon, uniform policy of animus inflicted by55 or more independent supervisors uponmore than 150 employees scatteredthroughout a multi-department plant.Consequently, the district court did notabuse its discretion in refusing to certifyPlaintiffs’ class based on a single comment.

5.

a.

The district court also gave ‘‘limitedweight’’ to almost 80 affidavits from black


employees at the Nucor plant. J.A. 10950.The affidavits consistently rejected theidea of discrimination in the promotionsprocess, and the district court did notabuse its discretion in affording them someminimal value. Repeatedly, the affidavitssuggest that the promotions process wasfair. See, e.g., J.A. 6024, 6042, 6052, 6069,6078. One such employee specifically re-marked that ‘‘[n]ot all African–Americansfeel like they have been discriminatedagainst at Nucor.’’ J.A. 6109. The sameemployee was actually ‘‘upset by this racialdiscrimination issue because it is not some-thing that has happened to me or is hap-pening across the board here at Nucor.’’Id. Another employee explained that ‘‘theway things are done TTT at Nucor are notinfluenced by race.’’ J.A. 6164.

The list goes on: black employees ap-proved of management’s handling of race-related issues in the plant, see, e.g., J.A.6109, 6215, 6480–81, 6943, explained thatthey were treated well, see, e.g., J.A. 6350,6361, and often reasoned that complaintsof racism from other employees were un-justified, see, e.g., J.A. 6566. Even thosewho felt that promotions were not madefairly often blamed factors other than race,such as a ‘‘buddy’’ system in which super-visors promoted friends. See, e.g., J.A.6258, 6299, 6438, 6494. Some affidavitsalso directly contradicted the sixteen dec-larations that Plaintiffs submitted. Infact, Jacob Ravenell, Kenneth Hubbard,Robyn Spann, and Byron Turner all ex-pressly denied that they had been deniedpromotions because of their race, eventhough Plaintiffs cite them as four of theirsixteen key witnesses. See J.A. 6400,6746, 6933, 6964. The district court hadevery right to weigh such self-contradicto-ry testimony and conclude as it did. SeeStevenson v. City of Seat Pleasant, Md.,743 F.3d 411, 422 (4th Cir.2014).

b.

Based on ‘‘[c]ommon sense and pru-dence,’’ however, the majority finds yetagain that the district court clearlyerred—this time by finding that ‘‘potential-ly coercive’’ affidavits supported Nucor tosome small degree. Maj. op. at 913. Themajority’s naked credibility determinationis exactly the sort of decision we are notmeant to undertake on appellate review.‘‘[W]hen a trial judge’s finding is based onhis decision to credit the testimony of [awitness who] TTT has told a coherent andfacially plausible story that is not contra-dicted by extrinsic evidence, that finding, ifnot internally inconsistent, can virtuallynever be clear error.’’ Anderson, 470 U.S.at 575, 105 S.Ct. 1504.

The majority nevertheless adopts a self-contradictory credibility rule: statementsmade in support of an employer must berejected when the employer obtains them,while statements made against the em-ployer will be given ‘‘significant weightgiven the circumstances in which theywere made.’’ Maj. op. at 914. The majori-ty draws this distinction by assuming thatan employer exercises coercive power inmost any interaction with its employees.‘‘However, it is well settled that not everyinterrogation of employees by Companyofficials constitutes coercion[.]’’ NLRB v.Lexington Chair Co., 361 F.2d 283, 289(4th Cir.1966). And one must not losesight of the practical effect of the majori-ty’s novel approach: employers now haveno incentive to investigate and remedyclaims of discrimination. Employers willwell understand that investigations can nolonger benefit them—at most, facts devel-oped during an investigation will only beused against the employer. Even an em-ployer with a supportive workforce will beunable to defend itself with beneficial em-ployee testimony, lest it be accused ofunproven coercion. Informal resolution,


Congress’ preferred course, will thereforebecome even more difficult. See West v.Gibson, 527 U.S. 212, 218–19, 119 S.Ct.1906, 144 L.Ed.2d 196 (1999) (noting Con-gress’s intention that Title VII claimswould be resolved informally).

One is further left to wonder where themajority’s new imagined-coercion-basedrule comes from. Generally, the purport-edly ‘‘coercive nature of the employer-em-ployee relationship TTT is insufficient todemonstrate that TTT [employer-employee]interviews were improper.’’ Slavinski v.Columbia Ass’n, Inc., No. CCB–08–890,2011 WL 1310256, at *4 (D.Md. Mar. 30,2011) (collecting cases); accord Maddockv. KB Homes, Inc., 248 F.R.D. 229, 237(C.D.Cal.2007); McLaughlin v. LibertyMut. Ins. Co., 224 F.R.D. 295, 298(D.Mass.2004); cf. Gulf Oil Co. v. Ber-nard, 452 U.S. 89, 104, 101 S.Ct. 2193, 68L.Ed.2d 693 (1981) (‘‘[T]he mere possibilityof abuses does not justify routine adoptionof a communications ban[.]’’). Certainly itcannot be found in the cases the majoritycites, which all raised questions about de-fendants who contacted putative classplaintiffs after a class action had beenfiled. Here, Nucor investigated and ob-tained affidavits before any lawsuit wasfiled, so it could not have been attemptingto break up the class—the class did noteven exist yet. The majority’s cases alsoinvolved a level of egregious misconductnot found in this case, suggesting thatthose cases were directed at a problemthat does not exist here. See, e.g., Kleinerv. First Nat’l Bank of Atlanta, 751 F.2d1193, 1197–98 (11th Cir.1985) (finding uni-lateral contacts improper where counselviolated direct court order and conducted avast ‘‘selling job’’ seeking class opt-outs in‘‘[s]ecrecy and haste’’ during ‘‘the districtjudge’s vacation’’); see also Burrow v. Sy-baris Clubs Int’l, Inc., No. 13 C 2342, 2014WL 5310525, at *4–5 (N.D.Ill. Oct. 17,2014) (summarizing many of the same

cases and concluding that they ‘‘depict[ed]communications so extreme that they actu-ally cut against [the majority’s present]position’’).

We also need not speculate about ‘‘po-tential’’ coercion, as the circumstancesmake plain that Nucor did not coerce itsemployees into making positive state-ments. No employee has claimed that theaffidavits were coercive. No employee hassuggested that Nucor retaliated againstemployees who complained of discrimina-tion. And the contents of the affidavits donot imply coercion either. Employees evi-dently felt free to speak honestly, as theaffidavits were not universally favorable toNucor. See, e.g., J.A. 10950 (district courtnoting that the affidavits ‘‘actually bol-stered the plaintiffs’ claims of a commonhostile work environment’’). Some em-ployees also chose not to give statementsat all. See, e.g., J.A. 6911. And still otheremployees made handwritten correctionsto their typed affidavits, indicating that theemployees had complete control over theirstatements. See, e.g., J.A. 6120.

What is more, Nucor gave each employ-ee a written notice explaining that theinterview was voluntary, that the inter-views were being taken on behalf of thecompany, that employees could decline toparticipate, and that they would not faceany retaliation for what they said. See,e.g., J.A. 6003. In other contexts, theCourt has said that disclosures like theseprevent coercion. See, e.g., OverniteTransp. Co. v. NLRB, 280 F.3d 417, 434(4th Cir.2002). Each employee who choseto participate then signed an acknowledge-ment and noted in his or her affidavit thatNucor did not coerce the employee. See,e.g., J.A. 6003.

The majority nevertheless condemnsNucor for not informing the employeesthat the company might use their state-


ments in litigation. This novel require-ment—a sort of ‘‘civil Miranda rule’’—seems an odd one given that litigation hadnot been filed. Instead, interviewees wereaccurately informed that ‘‘[t]here ha[d]been a few charges of discrimination filedby African–American employees at Nu-cor,’’ and the interview was meant to ‘‘de-termine what happened.’’ J.A. 6003.

The district court did not clearly err inaffording some weight to these many con-trary affidavits.

6.

In addition to the affidavits supportingNucor’s view, Plaintiffs’ affidavits mustalso be weighed against the company’s an-nounced anti-discrimination policy. InWal–Mart, the Supreme Court found thata ‘‘general policy of discrimination’’ washarder to find given the company’s ‘‘an-nounced policy forbid[ding] TTT discrimina-tion and TTT impos[ing] penalties for deni-als of equal opportunity.’’ Id. at 2553.The same holds true here. Nucor is anequal-opportunity employer with an ex-press anti-discrimination policy that harsh-ly penalizes employees engaging in dis-criminatory conduct. Nucor policies evenpunish supervisors who fail to put an endto their subordinates’ discriminatory con-duct. The record also contains accounts ofinstances in which Nucor’s general manag-er condemned discriminatory acts andpunished employees for using offensivelanguage. This countervailing evidencesupports the district court’s conclusionthat, as a whole, the anecdotal evidencefavored Nucor rather than Plaintiffs.

7.

a.

Aside from the qualitative and quantita-tive deficiencies in Plaintiffs’ anecdotal evi-dence, it also does not tell a plant-widestory. In Wal–Mart, plaintiffs’ anecdotal

evidence failed in part because ‘‘[m]orethan half of the[ ] reports [we]re concen-trated in only six States.’’ 131 S.Ct. at2556. As a result, even if one assumedthat ‘‘every single one of these accounts[were] true, that would not demonstratethat the entire company operate[d] undera general policy of discrimination.’’ Id.

The lack of dispersion that proved fatalto the class in Wal–Mart presents itselfhere. Eleven of the sixteen declarations—again, more than half—come from employ-ees in a single department: the beam mill.No cold mill or maintenance employees arerepresented, while only one shipping em-ployee and one melt shop employee ap-pear. And as the district court recog-nized, when one examines the individualinstances of discrimination alleged inPlaintiffs’ declarations, most of them con-cern just one manager and three supervi-sors who all worked in the beam mill. SeeJ.A. 10951. As one black employee put it,‘‘Whatever [wa]s happening in the beammill [wa]s not a plant wide problem.’’ J.A.6109.

b.

The majority somehow finds clear errorin the district court’s finding that Plain-tiffs’ accounts were concentrated in thebeam mill. But it proves easy to see whythe district court found what it did: Plain-tiffs do not cite useful, relevant evidencefrom outside the beam mill. Some anec-dotes fall outside the class period. See,e.g., J.A. 1085. Others involve promotionsthat did in fact go to a black employee.See, e.g., J.A. 1110–11. Some involvetransfers, not promotions. See, e.g., J.A.1063. Still others trace back to beam millsupervisors, not supervisors in other de-partments. See, e.g., J.A. 1079–80. Plain-tiffs count six other instances twice. SeeAppellant’s Br. 9–10. And some of thecited ‘‘instances of alleged promotions dis-


crimination’’ amount to no evidence at all.See, e.g., id. at 9 (citing J.A. 7237—anapplication for transfer—as one instance of‘‘promotion discrimination’’). Most incred-ibly, Plaintiffs’ argument—which the ma-jority appears to adopt—assumes that onecan find evidence of discrimination in ev-ery single instance where a black employ-ee does not receive a promotion for whichhe applies. That concept finds no supportin any part of our jurisprudence. Indeed,it turns the Teamsters framework into acircular absurdity. Plaintiffs presume thateach denied promotion evidences a dis-criminatory policy or practice, eventhough—under Teamsters—Plaintiffs mustprove that a discriminatory policy or prac-tice existed before the court may presumethat a particular denied promotion wasdiscriminatorily made. See Teamsters,431 U.S. at 362, 97 S.Ct. 1843.

The district court recognized, as itshould have, that the anecdotal evidencewas more substantial when it came to thebeam mill. For that reason, the districtcourt explained that it was willing to certi-fy a class of those applying out of and intothe beam mill. J.A. 10953–54 & n.16.Plaintiffs never accepted the invitation, sothey remain responsible for proving plant-wide commonality. That effort requires asubstantial showing beyond a single de-partment. See, e.g., Bennett, 656 F.3d at816 (holding that the district court proper-ly declined to certify a hostile work envi-ronment class where anecdotal evidencewas concentrated in a single department).

Outside the beam mill, Plaintiffs at bestpresent a few scattered anecdotes in eachdepartment. That’s not enough. ‘‘[A]class plaintiff’s attempt to prove the exis-tence of TTT a consistent practice within agiven department[ ] may fail even thoughdiscrimination against one or two individu-als has been proved.’’ Cooper, 467 U.S. at878, 104 S.Ct. 2794; accord Ste. Marie v.

E. R.R. Ass’n, 650 F.2d 395, 406–07 (2dCir.1981). The district court might verywell have clearly erred had it acceptedsuch evidence. One can hardly say that itclearly erred in doing just the opposite.

8.

In a last effort to save their class-wideclaim, Plaintiffs make much of other factsthat do not relate directly to promotions.They seem to give special attention to thefacts underlying their already-certifiedhostile work environment claim. The ma-jority agrees that such evidence provides a‘‘cultural backdrop’’ that renders an ‘‘equi-table promotions system’’ essentially im-possible. Maj. op. at 912. Notably, thatview never appeared in Brown I, but refer-ences to Plaintiffs’ hostile work environ-ment claims now appear at least a dozentimes in the majority opinion. The majori-ty also finds evidence of a ‘‘culture’’ in thealleged fact that Nucor hired only oneblack supervisor before the EEOC investi-gation, even though ‘‘[t]he mere absence ofminority employees in upper-level posi-tions does not suffice to prove [even] aprima facie case of discrimination withouta comparison to the relevant labor pool.’’Carter, 33 F.3d at 457.

We have never held that class plaintiffsmay establish a common, classwide policyof discrimination with mere evidence ofcompany ‘‘culture.’’ Other decisions, in-cluding Wal–Mart, reject the notion that‘‘culture’’ is enough. See Wal–Mart, 131S.Ct. at 2553; Davis, 717 F.3d at 487–88.The majority would nevertheless ‘‘sweepmany individual plaintiffs and sets of factsinto one class on the premise that all re-flect illegal conduct by the defendant inpractice and culture if not in policy’’—eventhough that is ‘‘precisely the sort of classthat the Supreme Court recently rejectedin [Wal–Mart ].’’ Jamie S. v. MilwaukeePub. Schs., 668 F.3d 481, 504 (7th Cir.


2012) (Rovner, J., concurring in part).Furthermore, simply saying that a compa-ny has a ‘‘cultural problem’’ does not iden-tify any particular employment policy orpractice, McClain v. Lufkin Indus., Inc.,519 F.3d 264, 274 (5th Cir.2008), let alone acommon, uniform policy spanning theclass.

We have also never held that facts es-tablishing a hostile work environment una-voidably relate to all other employmentdecisions made in the same company.Such a connection would be hard to justify,as acts giving rise to a hostile work envi-ronment are only distantly related to thediscrete acts that underlie disparate treat-ment and impact claims. ‘‘The probativevalue of other discriminatory acts dependsTTT on the nature of the discriminationcharged.’’ Hunter v. Allis–ChalmersCorp., Engine Div., 797 F.2d 1417, 1424(7th Cir.1986), abrogated on other groundsby Patterson v. McLean Credit Union, 491U.S. 164, 109 S.Ct. 2363, 105 L.Ed.2d 132(1989). And ‘‘[h]ostile environment claimsare different in kind from discrete acts.’’Nat’l R.R. Passenger Corp. v. Morgan, 536U.S. 101, 115, 122 S.Ct. 2061, 153 L.Ed.2d106 (2002). In contrast to acts creating ahostile work environment, discriminatoryemployment decisions ‘‘inflict[ ] direct eco-nomic harm.’’ Burlington Indus., 524U.S. at 762, 118 S.Ct. 2257. They willoften require ‘‘the imprimatur of the enter-prise and the use of its internal processes.’’Id.

The ‘‘probativeness’’ of items like com-ments, jokes, and other acts ‘‘is [also] cir-cumscribed if they were made [or done] ina situation temporally remote from thedate of the employment decision[s], or ifthey were not related to the employmentdecision[s] in question or were made bynondecisionmakers.’’ McMillan v. Mass.Soc’y for Prevention of Cruelty to Ani-mals, 140 F.3d 288, 301 (1st Cir.1998).

Here, Plaintiffs’ evidence suffers to somedegree from all three of these defects.For instance, Plaintiffs’ statements oftendo not tell us when the offensive conductoccurred, so we have no way of assessingtemporal proximity. None of the ‘‘cultur-al’’ evidence pertains specifically to pro-motions. And most all of the relevanthostile-work-environment conduct camefrom non-decisionmakers, even though it‘‘is the perception of the decisionmakerthat is relevant’’ in claims like Plaintiffs’.Smith v. Flax, 618 F.2d 1062, 1067 (4thCir.1980); accord Mateu–Anderegg v. Sch.Dist. of Whitefish Bay, 304 F.3d 618, 623(7th Cir.2002) (‘‘[S]tatements are only rele-vant if they come from a decisionmaker,someone involved in the adverse employ-ment decision[s].’’). Lastly, to the limitedextent that supervisors did involve them-selves in the incidents that Plaintiffs de-scribed, those supervisors chiefly workedin the beam mill—undermining any infer-ence of a common, plant-wide policy.

At bottom, the majority concludes thatwe should permit Plaintiffs to pursue twoclass claims pertaining to promotions be-cause they have successfully establishedtheir right to pursue a separate, distin-guishable hostile-work-environment claim.Title VII does not work that way, and,rhetoric aside, the majority is unable toidentify a single decision to support thatkind of proposition. ‘‘In the law, the ab-sence of precedent is no recommendation.’’Dukes v. Wal–Mart, Inc., 509 F.3d 1168,1200 (9th Cir.2007) (Kleinfeld, J., dissent-ing). Moreover, to assume that a plaintiffestablishes a right to class treatment forhis discrete-act class merely because hehas established such a right as to a hostile-work-environment class is to reinstate asuspect revision of the ‘‘across-the-board’’rule that the Supreme Court rejectedthree decades ago. See Falcon, 457 U.S.at 153, 157–59, 102 S.Ct. 2364 (rejectingthe idea that ‘‘an employee complaining of


one employment practice’’ may automati-cally ‘‘represent another complaining ofanother practice’’ merely because both al-leged discrimination based on the sameprotected trait). The district court did notabuse its discretion in refusing to exhumethat long-dead idea.

The district court did not clearly err indeclining to give dispositive weight to evi-dence going to Plaintiffs’ hostile-work-en-vironment claim when deciding whether tocertify Plaintiffs’ separate promotions-re-lated classes.

* * *

When closely examined, Plaintiffs’ anec-dotal evidence proves to be just as uncon-vincing as their statistical proof. ‘‘Because[Plaintiffs] provide no convincing proof of acompanywide discriminatory TTT pro-motion policy, TTT they have not estab-lished the existence of any common ques-tion.’’ Wal–Mart, 131 S.Ct. at 2556–57.The district court therefore did not abuseits discretion in declining to certify theclass because of its lack of commonality.

IV.

On the road to its desired result, themajority undermines well-established judi-cial processes, causes a rift between thisCourt and a co-equal circuit court withoutexplanation, and brings substantial uncer-tainty to an area of law that begs forclarity.

As to judicial processes, the majorityopinion evidences little respect for the roleof the district court and the standard ofreview. The district court has lived withthis matter for several years now, and itbest understands how the case has devel-oped. Its actions bespeak a court strivingto scrupulously apply Rule 23’s require-ments. The district court complied withour mandate, rejected more than one re-quest to decertify from Nucor, and contin-

ually endeavored to respect findings thatthis Court has (actually) made. Yet themajority shows no concern for that effort.And it shows just as little concern for thisCourt’s well-established waiver rule, whichshould plainly apply here.

As to our sister circuits, the majorityopinion begets a circuit split. The EighthCircuit affirmed the denial of class certifi-cation in a case involving the same claims,the same experts, and the same defendant.As should be clear by now, that decisioncannot be reconciled with this one. Themajority never even tries to do so.

And as to cases to come, the majority’sdecision will offer far more questions thananswers. What standard of review reallyapplies in this context? How much evi-dence must a plaintiff summon to complywith Rule 23? Does appellate waiver mat-ter? Does class treatment of one cause ofaction necessarily warrant class treatmentfor another? Must statistical evidenceprove to be reliable? Does Wal–Martreach only nationwide class actions? Cana sufficiently ‘‘common’’ policy result frominaction? These are only some of thequestions that the majority opinion leavesunresolved.

We should hardly take this troubledroad in the name of ‘‘simple justice.’’ Maj.op. at 922. ‘‘ ‘Simple justice’ is achievedwhen a complex body of law developedover a period of years is evenhandedlyapplied.’’ San Remo Hotel, L.P. v. City &Cnty. of San Fran., Cal., 545 U.S. 323, 345,125 S.Ct. 2491, 162 L.Ed.2d 315 (2005).Evenhandedness is nowhere to be foundhere, so justice remains unserved.

Perhaps the Supreme Court will act torectify the problems that are sure to followfrom today’s opinion. One can only hopethat it will do so soon. In the meantime, Irespectfully dissent. The district court did

957U.S. v. BAJOGHLICite as 785 F.3d 957 (4th Cir. 2015)

not abuse its discretion, and its judgmentto decertify should be affirmed.

,

UNITED STATES of America,Plaintiff–Appellant.

v.

Amir A. BAJOGHLI, Defendant–Appellee.

No. 14–4798.

United States Court of Appeals,Fourth Circuit.

Argued: March 25, 2015.

Decided: May 11, 2015.

Background: Dermatologist charged with‘‘knowingly and willfully executing ascheme to defraud any health care benefitprogram,’’ by routinely diagnosing patientswith skin cancer that they did not haveand then obtaining payment from healthcare benefit programs for performing med-ically unnecessary Mohs surgery on benigntissue, and by directing ‘‘unlicensed andunqualified medical assistants’’ to performwound closures on Mohs surgery patientsand then billing health care benefit pro-grams as if he had personally performedprocedure, filed motions in limine to pre-vent government from introducing certainevidence. The United States District Courtfor the Eastern District of Virginia, GeraldBruce Lee, J., granted motions, and gov-ernment filed interlocutory appeal.

Holdings: The Court of Appeals, Niemey-er, Circuit Judge, held that:

(1) district court abused its discretion inlimiting government’s proof to thatwhich was directly relevant to one or

more of the 53 executions charged inindictment;

(2) district court abused its discretion inexcluding, as ‘‘other crimes’’ evidence,evidence that, after being interviewedby law enforcement, dermatologist im-mediately stopped performing Mohssurgery without biopsy and deletedscheduling data for past wound re-pairs;

(3) district court abused its discretion inexcluding this evidence, alternatively,on ground that it was unfairly prejudi-cial; and

(4) district court abused its discretion, incase in which intent was at issue, inlimiting government’s ability to showevidence of profit that dermatologistwas able to earn by his conduct.

Reversed and remanded.

1. Health O979, 989

When government charges defendantwith ‘‘knowingly and willfully executing ascheme to defraud any health care benefitprogram’’ and elects to charge only someof the executions of that scheme, its elec-tion does not limit its proof to only thecharged executions; it may also introduceevidence of other uncharged executions inorder to prove the scheme, which is ele-ment of the charged misconduct. 18U.S.C.A. § 1347.

2. Health O979

While health care fraud can be com-mitted simply by engaging in isolatedtransaction, ‘‘scheme to defraud,’’ withinmeaning of criminal statute, requires aplot, plan, or arrangement that is executedby fraudulent transaction. 18 U.S.C.A.§ 1347.

See publication Words and Phras-es for other judicial constructionsand definitions.

Documents

BROWN v. NUCOR CORP. - Thomson Reutersstatic.legalsolutions.thomsonreuters.com/pdf/785_F3d_895.pdf · BROWN v. NUCOR CORP. 895 ... NUCOR CORPORATION; Nucor Steel Berkeley, Defendants