Supp. 2d at 300, and subsequently found, after a hearing in July 2010,

13   that the exam was invalid, see United States v. City of New York, No.

14   07-cv-2067, 
2010 WL 4137536
, at *5 (E.D.N.Y. Oct. 19, 2010).             On this

15   appeal, the City does not challenge the grant of summary judgment

16   against the City on the disparate impact claim, nor, as far as we can

17   determine, the District Court’s ruling on the invalidity of Exam 6019.

18      Pending litigation - disparate treatment claim. In addition to

19   reasserting    the   disparate   impact   claim   from   the    Government’s

20   complaint, the Intervenors’ complaint added a discriminatory treatment

21   claim, alleging that the Defendants’ use of the challenged employment

22   procedures    constituted   intentional   discrimination       against    black

23   applicants.    That claim raises one of the central issues on this

24   appeal.

                                        -10-
 1        On July 25, 2008, the District Court denied the Intervenors’ motion

 2   to augment their discriminatory treatment claim by amending their

 3   complaint    to   challenge    “additional    discriminatory     screening     and

 4   selection devices” used from 1999 to the present.               The Court noted

 5   that, at the time that the Intervenors had sought to intervene, they

 6   had represented that they were “taking pleadings as they find them,”

 7   and were simply seeking to add the disparate treatment claim.

 8        On Sept. 18, 2009, the City moved to dismiss the Intervenors’ claim

 9   of   intentional    discrimination,    and,    on   October     320,   2009,   the

10   Intervenors filed a motion for partial summary judgment on the issue

11   of discriminatory intent.         The Government, which had not alleged

12   discriminatory     treatment    in   its     complaint,   did    not   join    the

13   Intervenors’ motion for summary judgment on the disparate treatment

14   claim.

15        On January 13, 2010, the District Court issued a comprehensive

16   opinion granting the Intervenors’ motion for summary judgment on their

17   disparate treatment claim. See United States v. City of New York, 683

18   F. Supp. 2d 225
, 255 (E.D.N.Y. 2010) (“Disparate Treatment Op.”).               We

19   recount the details of that ruling in Part II, infra.                    In that

20   opinion, the Court dismissed the Intervenors’ Title VII claims against

21   Mayor Bloomberg and Commissioner Scoppetta because individuals are not

22   subject to liability under Title VII, 
id. at 243-44, and 
dismissed the

23   discriminatory treatment claim against them on the ground that they

24   were entitled to qualified and official immunity, 
id. at 269-72.

                                           -11-
 1      Pending litigation - relief. On September 10, 2009, after the

 2   Court’s July 22, 2009, Disparate Impact Opinion but before its January

 3   13, 2010, Disparate Treatment Opinion, the Government submitted a

 4   proposed order requesting injunctive and monetary relief to implement

 5   the Disparate Impact Opinion.         On January 21, 2010, eight days after

 6   the Disparate Treatment Opinion, the Court issued the first of four

 7   orders dealing with relief.      The January 21 order primarily alerted

 8   the   parties   to   monetary   and    compliance    issues   that    the   Court

 9   anticipated pursuing, but specifically required the City to develop a

10   new testing procedure for entry-level firefighters.                 It left for

11   future consideration the extent to which the City could continue to

12   use Exam 6019, a test the City first administered in January 2007 and

13   had   used   thereafter   to    generate      its   most   recent    firefighter

14   eligibility list.     The validity of that test had not previously been

15   challenged or adjudicated.

16      On May 26, 2010, the Court issued a second relief order.              In that

17   order, the Court stated that, in the absence of needed materials, it

18   could not then determine the validity of Exam 6019 nor determine to

19   what extent the FDNY could use the results of that exam for entry-

20   level hiring of firefighter.          In view of the complexity of pending

21   relief issues, the Court appointed a Special Master to facilitate the

22

23

24

                                            -12-
 1   Court’s assessment of Exam 6109 and to oversee the City’s development

 2   of a new exam.6

 3      On October 19, 2010, the Court issued a third relief order.            That

 4   order permanently enjoined the City from using Exam 6019, with a

 5   limited exception not relevant to the appeal.

 6      On December 9, 2010, the Intervenors moved for equitable and

 7   monetary relief based on the Court’s previous finding, on motion for

 8   summary judgment, of disparate treatment.           Among other injunctive

 9   relief,   they    requested   the    appointment   of   monitor   to   oversee

10   compliance, enhanced recruitment and advertising to target minority

11   applicants, modification of the FDNY’s post-exam screening process,

12   and prevention of retaliation and workplace discrimination against

13   black firefighters.    On February 28, 2011, the Government submitted a

14   revised proposed relief order, requesting relief based on the Court’s

15   disparate impact finding.     In August 2011, the District Court held a

16   bench trial to determine appropriate injunctive relief for the City’s

17   intentional discrimination.         The Government did not participate in

18   that trial.

19      On September 30, 2011, the Court issued detailed findings of fact,

20   based on the evidence introduced at th bench trial, to support its



          6
            The Court initially appointed Robert M. Morgenthau as Special
     Master. On June 1, 2010, after the City objected to the selection of
     Morgenthau because of the City’s disputes with the New York County
     District Attorney’s Office, which he had headed, Morgenthau asked to
     be relieved, and on the same day the Court appointed Mary Jo White.

                                           -13-
 1   subsequent grant of injunctive relief.       The Court noted that its

 2   “assessment of the evidence” was “influenced” by the factual record

 3   established in earlier stages of the litigation, including the finding

 4   that Exams 7029 and 2043 had a disparate impact on black and Hispanic

 5   firefighter candidates, the finding of intentional discrimination, and

 6   the finding that Exam 6019 was invalid for lack of job validation.

 7   United States v. City of New York, No. 07-CV-2067, 
2011 WL 766158
, at

 8   *1 n.1 (E.D.N.Y. Sept. 30, 2011).   Approximately one week later, the

 9   Court issued a draft remedial order and informed the parties that it

10   intended to appoint a Court Monitor to oversee the City’s compliance

11   with this order.   The Court permitted the City and its Intervenors an

12   opportunity to comment on the draft order.    On December 8, 2011, the

13   Court issued the injunction that is a principal subject of this

14   appeal. See United States v. City of New York, No. 07-CV-2067, 
2011 WL

15 
  6131136 (E.D.N.Y. Dec. 8, 2011) (“Injunction Op.”).    The details of

16   the terms of that injunction will be recounted in Part IV, infra,

17   dealing with the City’s objections to several of those terms.

18      On February 1, 2012, the District Court, pursuant to Rule 54(b) of

19   the Federal Rules of Civil Procedure, certified for entry of partial

20   summary judgment its ruling dismissing the claims against Mayor

21   Bloomberg and Commissioner Scoppetta on grounds of qualified and

22   official immunity.

23      The City filed a timely appeal, and the Intervenors filed a timely

24   cross-appeal, which have been consolidated.   Motions for back-pay and

                                      -14-
 1   damages remain pending in the District Court and are not the subject

 2   of this appeal.

 3                                   Discussion

 4      Before considering any of the issues on appeal, we note that the

 5   City   has   explicitly   declined   to   challenge   the   District   Court’s

 6   disparate impact ruling, the remedy requiring development of a new

 7   entry-level exam, or the appointment of a Special Master.         The City’s

 8   appellate papers also present no challenge to the District Court’s

 9   third relief order substantially enjoining use of Exam 6019. What the

10   City challenges on its appeal is the granting of summary judgment in

11   favor of the Intervenors on their disparate treatment claim and all

12   aspects of the injunction beyond those requiring development of a new

13   entry-level exam.    On the cross-appeal, the Intervenors challenge the

14   District Court’s dismissal of their claims against Mayor Bloomberg and

15   Commissioner Scoppetta on the ground of qualified immunity.

16   I. Appellate Jurisdiction

17      All parties acknowledge our jurisdiction to review the December 8,

18   2011, injunction, see 28 U.S.C. § 1292(a)(1), and the February 1,

19   2012, partial final judgment dismissing the claims against Mayor

20   Bloomberg and Commissioner Scoppetta, see Fed. R. Civ. P. 54(b).           The

21   Intervenors challenge our jurisdiction to review the District Court’s

22   January 13, 2012, ruling granting the Intervenors summary judgment on

23   their disparate treatment claim.      They point out that this ruling is

24   not a final order and has not been incorporated into a final judgment.

                                          -15-
 1   The City responds that we have jurisdiction over the disparate

 2   treatment ruling because it is “inextricably intertwined,” with the

 3   injunction. Lamar Advertising of Penn, LLC v. Town of Orchard Park,

 4   356 F.3d, 371 (2d Cir. 2004) (internal quotation marks omitted); see

 5   also Swint v. Chambers County Commission, 
514 U.S. 35
, 51 (1995).

 6      We agree with the City.        First, the Intervenors themselves focused

 7   almost   exclusively   on   the    disparate      treatment     finding   in   their

 8   proposed order for injunctive relief, and, in summation during the

 9   bench trial on relief, emphasized that broad remedies were needed to

10   counteract    intentional   discrimination.             More   significantly,   the

11   District Court explicitly acknowledged that its findings on which the

12   injunction would later be based were “influenced” by its disparate

13   treatment finding, and some of the more far-reaching provisions of

14   that injunction appear to be grounded, at least partially if not

15   entirely, on that finding.        Sufficient “intertwining” exists between

16   the injunction and the disparate treatment summary judgment ruling to

17   support pendent appellate jurisdiction over the latter ruling.

18   II.   The   Summary   Judgment     Ruling    on   the    Intervenors’     Disparate

19   Treatment Claim

20      In considering the District Court’s grant of summary judgment to

21   the Intervenors on their disparate treatment claim, which requires an

22   intent to discriminate, we note at the outset that questions of

23   subjective intent can rarely be decided by summary judgment. See

24   Harlow v. Fitzgerald, 
457 U.S. 800
, 816 (1982).                The principal issue

                                           -16-
 1   presented by the summary judgment ruling concerns the nature of a

 2   defendant’s obligation to respond to a prima facie case presented by

 3   a plaintiff class in a pattern-or-practice discriminatory treatment

 4   lawsuit.

 5      Initiation of a pattern-or-practice claim. Before considering that

 6   issue, we first consider how a pattern-or-practice claim arises.     A

 7   pattern-or-practice claim under Title VII can be asserted either by

 8   the United States or by a class of plaintiffs, usually current or

 9   prospective employees against whom some adverse employment action has

10   been taken because of an impermissible reason such as race.    Section

11   707(a) of the Civil Rights Act of 1964, 42 U.S.C. § 2000e-6(a),

12   authorizes the Attorney General to bring a civil action whenever that

13   officer “has reasonable cause to believe that any person or group of

14   persons is engaged in a pattern or practice of resistence to the full

15   enjoyment of any of the rights secured by [subchapter VI of chapter

16   21], and that the pattern or practice is of such a nature and is

17   intended to deny the full exercise of the rights herein described

18   . . . .”7   A group of plaintiffs, entitled to be certified as a class,

19   may also initiate a pattern-or-practice suit. See Cooper v. Federal

20   Reserve Bank of Richmond, 
467 U.S. 867
, 876 n.9 (1984) (“[T]he

21   elements of a prima facie pattern-or-practice case are the same [as a


          7
            Section 707 was amended by Section 5 of the Equal Employment
     Opportunity Act of 1972, 42 U.S.C. § 2000e-6(c), to give the EEOC,
     rather than the Attorney General, authority to bring patter-or-
     practice suits against private sector employers.

                                       -17-
 1   Government-initiated suit under Section 707(a)] in a private class

 2   action.”); Franks v. Bowman Transportation Co., 424 U.S., 747, 750-51

 3   (1976) (analyzing class action alleging pattern of discriminatory

 4   employment practices).

 5      Although the pending suit was brought by the United States, the

 6   Government did not allege a pattern or practice of discriminatory

 7   treatment.   Its claim was solely that the City’s use of Exams 7029 and

 8   2043 had a discriminatory impact on minority applicants for the

 9   position of entry-level firefighter.    The Intervenors, once certified

10   as a class, have asserted what amounts to claim of pattern-or-practice

11   discriminatory treatment.8


          8
            The Intervenors’ complaint did not explicitly assert a claim of
     a pattern-or-practice.    That phrase first entered this litigation
     rather unobtrusively as one aspect of the prayer for relief in the
     Intervenors’ complaint, which, in listing the elements of a requested
     injunction, asked the Court to “appoint entry-level firefighters from
     among qualified black applicants in sufficient numbers to offset the
     historic pattern and practice of discrimination against blacks in
     testing and appointment to that position.” Intervenors’ Complaint,
     Prayer for Relief ¶ 3(d). The phrase is not mentioned at all in the
     Intervenors’ extensive memorandum of law in support of their motion
     for summary judgment on the disparate treatment claim. Nevertheless,
     by the time the District Court issued its Disparate Treatment Opinion,
     the phrase had become prominent.      Section IV of that opinion is
     captioned “INTERVENORS’ TITLE VII PATTERN-OR-PRACTICE DISPARATE
     TREATMENT 
CLAIM.” 683 F. Supp. 2d at 246
. And as the litigation has
     reached this Court, the phrase appears repeatedly in the briefs of the
     City and the Intervenors, although it is conspicuously absent from the
     Government’s brief (except for one mention in the description of the
     District Court’s Disparate Treatment Opinion, see Brief for United
     States at 19).

          We surmise that the Intervenors are entitled to assert a pattern-
     or-practice claim because they sought and were granted class action
     status and alleged not only the disparate impact of Exams 7029 and

                                      -18-
 1      Comparison of individual and pattern-or-practice claims.          We next

 2   compare individual and pattern-or-practice claims.            The principal

 3   difference between individual and pattern-or-practice discriminatory

 4   treatment   claims   is   that,   although   both   require   an   intent   to

 5   discriminate, an individual claim requires an intent to discriminate

 6   against one person, see, e.g., McDonnell Douglas Corp. v. Green, 411

 
7   U.S. 792 
(1973), and a pattern-or-practice claim requires that “racial

 8   discrimination was the company’s standard operating procedure[,] the

 9   regular rather than the unusual practice,” International Brotherhood

10   of Teamsters v. United States, 431 U.S., 324, 336 (1977), and that the

11   discrimination was directed at a class of victims, see, e.g., Franks,

12   424 U.S. at 772.9 
   It should be noted that “[a] pattern or practice

13   case is not a separate and free-standing cause of action . . ., but is

14   really merely another method by which disparate treatment can be

15   shown.” Chin v. The Port Authority of New York and New Jersey, 685



     2043 but also a long-standing pattern of discrimination in hiring
     firefighters.   Their complaint alleged, among other things, that
     “[t]he FDNY has a long history of unlawfully discriminating against
     blacks in its hiring process and of maintaining the number of black
     firefighters at its disproportionately low level compared to their
     representation in the population of the City as a whole,” ¶ 31, “the
     FDNY has consistently failed and refused to comply with many of the
     [City’s Equal Employment Practices Commission’s] recommendations,
     particularly with regard to its hiring criteria,” ¶ 32, and “the City
     and the FDNY have repeatedly failed and refused to remedy this
     obviously discriminatory situation,” ¶ 33.
          9
            Cf. EEOC v. Shell Oil Co., 
466 U.S. 54
, 73 (1984) (requiring an
     EEOC charge filed by a commissioner to “identify the groups of persons
     that he has reason to believe have been discriminated against”).

                                         -19-
 
1   F.3d 135
, 148-49 (2d Cir. 2012) (quoting in a parenthetical Celestine

 2   v. Petroleos de Venezuela SA, 
266 F.3d 343
, 355 (5th Cir. 2001))

 3   (internal quotation marks omitted).10

 4      Both types of suits involve a scheme of shifting burdens borne by

 5   the contending sides. In both, the plaintiff bears the initial burden

 6   of presenting a prima facie case. Both McDonnell 
Douglas, 411 U.S. at

 7 
  807, and 
Teamsters, 431 U.S. at 336
, refer to the plaintiff’s initial

 8   burden    as   a    burden   to   establish       “a   prima   facie   case,”   meaning

 9   sufficient     evidence      to   create    a     rebuttable    presumption     of   the

10   existence of the ultimate fact at issue: in McDonnell Douglas, the

11   employer’s intent to discriminate against the plaintiff, and in

12   Teamsters,         the   employer’s    pervasive         practice      of   intentional

13   discrimination against the class. The Supreme Court has noted that in

14   general “[t]he phrase ‘prima facie case’ not only may denote the

15   establishment of a legally mandatory, rebuttable presumption, but also

16   may be used by courts to describe the plaintiff’s burden of producing

17   enough evidence to permit the trier of fact to infer the fact at

18   issue,” Texas Dep’t of Community Affairs v. Burdine, 
450 U.S. 248
, 254

19   n.7 (1981), and has explicitly instructed “that in the Title VII

20   context we use ‘prima facie case’ in the former sense,” 
id.

21




          10
            The 
Supreme Court has criticized the EEOC for not adopting
     “special regulations more closely tailored to the characteristics of
     ‘pattern-or-practice’ cases.” Shell Oil 
Co., 466 U.S. at 67 
n. 19.

                                                -20-
 1       In an individual case, the plaintiff’s initial burden consists of

 2   the now familiar components of showing “(i) that he belongs to a

 3   racial minority; (ii) that he applied and was qualified for a job for

 4   which the employer was seeking applicants; (iii) that, despite his

 5   qualifications, he was rejected; and (iv) that, after his rejection,

 6   the    position    remained   open   and    the   employer   continued   to    seek

 7   applicants from persons of complainant’s qualifications.” McDonnell

 8   
Douglas, 411 U.S. at 802
.       This burden is “not onerous,” Burdine, 
450

 9   U.S. at 253
; indeed, it is “minimal,” St. Mary’s Honor Center v.

10   Hicks, 
509 U.S. 502
, 506 (1993), or “slight,” Wanamaker v. Columbian

11   Rope Co., 
108 F.3d 462
, 465 (2d Cir. 1997).

12       In a pattern-or-practice case, the plaintiff’s initial burden is

13   heavier in one respect and lighter in another respect than the burden

14   in an individual case.        It is heavier in that the plaintiff must make

15   a     prima   facie   showing   of   a     pervasive   policy   of    intentional

16   discrimination, see 
Teamsters, 431 U.S. at 336
, rather than a single

17   instance of discriminatory treatment.              It is lighter in that the

18   plaintiff     need    not   initially      show   discrimination     against    any

19   particular present or prospective employee. See 
id. at 360; Chin, 
685

20   F.3d at 147
.      Although instances of discrimination against particular

21   employees are relevant to show a policy of intentional discrimination,

22   they are not required; a statistical showing of disparate impact might

23   suffice. See Hazelwood School District v. United States, 
433 U.S. 299
,

24   307-08 (1977) (“Where gross statistical disparities can be shown, they

                                              -21-
 1   alone may in a proper case constitute prima facie proof of a pattern

 2   or practice of discrimination.”).          With both types of cases, the

 3   plaintiff’s initial burden is only to present a prima facie case that

 4   will support a rebuttable presumption of the ultimate fact in issue.

 5      Once the McDonnell Douglas plaintiff has established its prima

 6   facie case, the burden then shifts to the employer “to rebut the

 7   presumption of discrimination,” 
Burdine, 450 U.S. at 254
.                  The

 8   employer need only “‘articulate come legitimate, nondiscriminatory

 9   reason for the employee’s rejection.’” 
Id. at 253 (emphasis 
added)

10   (quoting McDonnell 
Douglas, 411 U.S. at 802
).11             In Teamsters, the

11   Supreme Court said that the employer responding to a prima facie case

12   in a pattern-or-practice suit has the burden to “defeat” that case,

13   431 U.S. at 360
.      “[D]efeat” might be thought to imply something

14   stronger that “rebut,” but the Court’s language indicates that the

15   Court means the same thing in both contexts.           In McDonnell Douglas,

16   the court said that the employer may discharge its rebuttal burden by

17   “articulat[ing] some legitimate, nondiscriminatory reason for the

18   employee’s 
rejection.” 411 U.S. at 802
, and in Teamsters, the Court

19   similarly   said   that   the   employer   may   do   so   by   “provid[ing]   a



          11
             In this respect, the rebuttal burden on the employer in a
     discriminatory treatment case is less than the burden in a disparate
     impact case. In the latter case, the employer bears the burden of
     proving that the neutral employment policy, such as an exam, shown to
     have a discriminatory impact, is job-related. See Albemarle Paper Co.
     v. Moody, 
422 U.S. 405
, 424 (1975); Griggs v. Duke Power Co., 
401 U.S.
     424
, 431-32 (1971).

                                         -22-
 1   nondiscriminatory      explanation   for    the   apparently   discriminatory

 2   
result,” 431 U.S. at 360 
n. 46.        Although the Court has explicitly

 3   called the employer’s burden in a McDonnell Douglas case a burden of

 4   “production,” 
Burdine, 450 U.S. at 255
, and has not used that word to

 5   describe the employer’s burden in a pattern-or-practice case, we think

 6   the rebuttal burden in both contexts is one of “production.” See

 7   Reynolds v. Barrett, 
685 F.3d 193
, 203 (2d Cir. 2012) (noting that in

 8   pattern-or-practice case “the burden of production shifts to the

 9   employer”); Robinson v. Metro-North Commuter R.R., 
267 F.3d 147
, 159

10   (2d Cir. 2001) (noting that in pattern-or-practice case “‘the burden

11   [of production] then shifts to the employer’”) (quoting Teamsters, 
431

12   U.S. at 360
) (brackets in Robinson), abrogated on other grounds by

13   Wal-Mart Stores, Inc. v. Dukes, 
131 S. Ct. 2641 
(2011).12

14      A central issue in the pending case is what showing an employer

15   must make to satisfy its burden of production in a pattern-or-practice

16   case.        In Teamsters the Supreme Court stated that the employer’s

17   burden was “to defeat the prima facie showing of a pattern or practice

18   by demonstrating that the Government’s proof is either inaccurate or

19   
insignificant.” 431 U.S. at 360 
(emphasis added).              The emphasized

20   words raise a question as to whether the Supreme Court thought the

21   employer’s rebuttal evidence must be directed at the statistics that



             12
            In Teamsters, the Supreme Court had no need to label the nature
     of the employer’s rebuttal burden because the Court was reviewing a
     case that had been fully tried on the merits.

                                          -23-
 1   often constitute the prima facie case of discrimination or simply at

 2   the rebuttable presumption of discrimination that arises from those

 3   statistics.

 4      We think the Court meant that the employer must produce any

 5   evidence     that   is   relevant     to     rebutting      the     inference     of

 6   discrimination.     No plaintiff can limit the type of evidence that a

 7   defendant must produce to rebut a prima facie case by its selection of

 8   particular    evidence   to   support      that   case.     The    Supreme     Court

 9   explicitly recognized this obvious point in Teamsters when it stated

10   that, although “[t]he employer’s defense must, of course, be designed

11   to meet the prima facie case . . . [,] [w]e do not mean to suggest

12   that there are any particular limits on the type of evidence an

13   employer may 
use.” 431 U.S. at 360 
n.46.          The Court offered an example

14   of an employer whose pattern of post-Act hiring was a product pre-Act

15   hiring, 
id. at 360, an 
example of evidence that would rebut the

16   inference    of   discriminatory    intent    arising     from    the   plaintiff’s

17   statistics, but not dispute the statistics themselves.                  That showing

18   would not demonstrate that the proof of the pattern was inaccurate or

19   insignificant; it would demonstrate that the proof of the pattern was

20   legally irrelevant.

21      Of course, it is always open to a defendant to meet its burden of

22   production by presenting a direct attack on the statistics relied upon

23   to constitute a prima facie case.       A defendant might endeavor to show

24   that the plaintiff’s statistics are inaccurate, for example, infected

                                          -24-
 1   with arithmetic errors, or lacking in statistical significance, for

 2   example, based on too small a sample.   But the rebuttal need not be so

 3   limited.   A defendant may rebut the inference of a discriminatory

 4   intent by accepting a plaintiff’s statistics and producing non-

 5   statistical evidence to show that it lacked such an intent.         In

 6   Teamsters, the Supreme Court recognized this means of rebutting a

 7   prima facie case by stating that “the employer’s burden is to provide

 8   a nondiscriminatory explanation for the apparently discriminatory

 9   
result.” 431 U.S. at 360 
n.46.   Again, such an explanation rebuts the

10   inference from a plaintiff’s statistics, even though it does not

11   directly challenge the statistics themselves.13


          13
            Our dissenting colleague contends that we have “conflate[d] two
     distinct tests set out in our disparate treatment jurisprudence,”
     [slip op. __ (dissent at 1)] and that where a plaintiff presents
     statistics to establish its prima facie case of a pattern or practice
     or pervasive discrimination, “those statistics must necessarily be
     addressed” by the defendant’s rebuttal evidence, [slip op. __ (dissent
     at 5)].

          As to the first contention, we have explicitly recognized the
     crucial difference that a plaintiff endeavoring to present a pattern
     or practice claim of intentional discrimination must prove a pervasive
     pattern of such discrimination whereas a plaintiff endeavoring to
     present only a claim of individual discrimination may succeed by
     showing that a facially neutral policy had a discriminatory impact.

          As to the second contention, our dissenting colleague cites
     
Teamsters, 431 U.S. at 360
, as stating that an employer rebutting a
     prima facie pattern or practice case must demonstrate that “the
     plaintiff’s statistics were inaccurate or insignificant.” [slip op. __
     (dissent at 3)]. But the referenced sentence from Teamsters refers to
     the plaintiff’s “proof,” not the plaintiff’s “statistics.”           A
     defendant, by presenting evidence of its choosing that it lacked a
     discriminatory intent, satisfies its rebuttal burden of showing that
     the plaintiff’s prima facie proof lacked significance. Furthermore,

                                      -25-
 1        Some confusion might have been created on this point by a passage

 2   in    the    late   Professor   Arthur    Larson’s   treatise   on   employment

 3   discrimination that this Court quoted in 
Robinson, 267 F.3d at 159
.

 4   That passage begins by stating, “Three basic avenues of attack are

 5   open to the defendant challenging the plaintiff’s statistics, namely

 6   assault on the source, accuracy, or probative force.” 1 Arthur Larson

 7   et al., Employment Discrimination § 9.03[2], at 9-23 (2d ed. 2001)

 8   (emphasis added).      This sentence, read in isolation, might be thought

 9   to require an employer to challenge the plaintiff’s statistics as

10   such.       But that interpretation is dispelled by Prof. Larson’s later

11   recognition in the same passage, also quoted in 
Robinson, 267 F.3d at

12 
  159, that a defendant may use “other non-statistical evidence tending

13   to rebut the inference of discrimination.” 
Larson, supra
, § 9.03[2],

14   at 9-24 (emphasis added).       Indeed, the current version of Employment

15   Discrimination, compiled by Prof. Larson’s son, has rewritten the

16   sentence quoted in Robinson and, more significantly, includes a

17   subsection making it clear that non-statistical evidence, including an

18   employer’s affirmative action efforts, are “both relevant to and



     although the dissent suggests that we have ignored Teamsters by
     permitting a defendant to rebut a prima facie case without directly
     challenging the plaintiff’s statistics, it is the Teamsters opinion
     itself that says, “We do not mean to suggest that there are any
     particular limits on the type of evidence an employer may use” to meet
     a plaintiff’s prima facie 
case, 431 U.S. at 360 
n.46, and also says
     that the employer’s burden in rebutting a prima facie case “is to
     provide   a   nondiscriminatory   explanation   for   the   apparently
     discriminatory result,” 
id.

                                              -26-
 1   probative 
of absence of intent to discriminate.” 1 Lex Larson,

 2   Employment Discrimination § 9.03[2][c], at 9-20.1 (2d ed. 2011)

 3   (footnote omitted).

 4      We        have   recognized   that   non-statistical    evidence,   such   as   a

 5   defendant’s affirmative action program, is probative of the absence of

 6   an employer’s intent to discriminate. See Coser v. Moore, 
739 F.2d

 7 
  746, 751-52, (2d Cir. 1984); see also EEOC v. Sears, Roebuck & Co.,

 8   
839 F.2d 302
, 314 (7th Cir. 1988) (“[S]tatistical evidence is only one

 9   method of rebutting a statistical case.”).                Although cases such as

10   Coser and Sears, Roebuck were considering evidence available to negate

11   discriminatory intent at trial, we see no reason why a defendant may

12   not proffer such evidence to satisfy its burden of production in

13   advance of trial on the merits.14


             14
             Our recent opinion in Reynolds v. Barrett stated that the
     defendant’s burden of production is to show “that the statistical
     evidence proffered by the plaintiffs is insignificant or 
inaccurate.”
     685 F.3d at 203 
(citing 
Teamsters, 341 U.S. at 360 
(substituting
     “statistical evidence” for proof,” the word used in the relevant
     passage in Teamsters)). This statement in Reynolds is dictum; the
     issue in that case was not the required content of a defendant’s
     rebuttal, but “whether recourse to the pattern-or-practice evidentiary
     framework is appropriate in a suit against individual state officials
     brought pursuant to 42 U.S.C. § 1983 for intentional discrimination.”
     685 F,3d at 197. Even as dictum, we think this sentence in Reynolds
     should be understood to offer one way to rebut a prima facie case, but
     surely not the only way. That broader understanding is required by
     the recognition in Teamsters that (a) an employer may rebut a prima
     facie case by offering a nondiscriminatory 
explanation, 431 U.S. at
     360 
n.46, and (b) the Supreme Court did not intend to limit the type
     of evidence an employer may use, 
id.      It is also 
required by the
     incontestable point that no plaintiff can limit its adversary’s
     responding evidence by the type of evidence that the plaintiff chooses
     to present.

                                               -27-
 1      Teamsters sets a high bar for the prima facie case the Government

 2   or a class must present in a pattern-or-practice case: evidence

 3   supporting a rebuttable presumption that an employer acted with the

 4   deliberate purpose and intent of discrimination against an entire

 5   
class. 431 U.S. at 358
.       An employer facing that serious accusation

 6   must have a broad opportunity to present in rebuttal any relevant

 7   evidence that shows that it lacked such an intent.

 8      Continuing with a comparison of the shifting burdens in individual

 9   and pattern-or-practice cases, we note that a defendant’s burden of

10   production “can involve no credibility assessment,” 
Hicks, 509 U.S. at

11 
  509, and “necessarily precedes the credibility-assessment stage,” 
id.

12   (emphasis in 
original).       Nothing in Teamsters suggests that these

13   aspects of the defendant’s production burden do not apply to pattern-

14   or practice claims.    Nor are there differences with respect to the

15   remaining aspects of the burden-shifting scheme, at least at the

16   liability stage of a trial.        If the defendant fails to rebut the

17   plaintiff’s   prima   facie    case,    the   presumption   arising   from   an

18   unrebutted prima facie case entitles the plaintiff to prevail on the

19   issue of liability and proceed directly to the issue of appropriate

20   relief. See 
Hicks, 509 U.S. at 509
.             On the other hand, if the

21   defendant satisfies its burden of production, the presumption arising

22   from the plaintiff’s prima facie case “drops out,” see 
id., 

  at 510-11, and the trier of fact must then determine, after a full

24   trial, whether the plaintiff has sustained its burden of proving by a

                                            -28-
 1   preponderance of the evidence the ultimate fact at issue. See United

 2   States Postal Service Board of Governors v. Aikens, 
460 U.S. 711
, 715

 3   (1983) (individual plaintiff must prove intent to discriminate);

 4   
Teamsters, 431 U.S. at 336 
(Government in pattern-or-practice case

 5   must    prove   that   intentional     discrimination      was   the   defendant’s

 6   standard     operating   procedure.”).        Of   course,    the    evidence   that

 7   originally      supported   the   plaintiff’s      prima     facie   case    remains

 8   available to contribute to the persuasive force of the plaintiff’s

 9   proof   on   the   ultimate   issue.    See   Reeves    v.    Sanderson     Plumbing

10   Products, Inc., 
530 U.S. 133
, 143 (2000); 
Burdine, 450 U.S. at 256
11   n.10.

12      At the relief stage, however, a special rule applies in pattern-or-

13   practice cases.        Once the Government or a class has proven by a

14   preponderance of the evidence a policy of intentional discrimination

15   and seeks relief for individual victims of that policy, “[t]he proof

16   of the pattern or practice supports an inference that any particular

17   employment decision, during the period in which the discriminatory

18   policy was in force, was made in pursuit of that policy. . . . [T]he

19   burden then rests on the employers to demonstrate that the individual

20   applicant was denied an employment opportunity for lawful reasons.”

21   
Teamsters, 431 U.S. at 362 
(citing 
Franks, 424 U.S. at 773 
n.32).                 In

22   Wal-Mart 
Stores, 131 S. Ct. at 2552 
n.7, the inference was explicitly




                                            -29-
 1   called “a rebuttable inference.”15

 2      The Intervenors’ prima facie case.        The statistical disparities

 3   supporting the unchallenged finding that the Exams has a racially

 4   disparate impact also served to establish a prima facie case on the

 5   Intervenors’ claim of a pervasive pattern of discriminatory treatment,

 6   especially in light of the long-standing pattern of low minority

 7   participation in the FDNY. See Hazelwood School 
District, 433 U.S. at

 8 
  307-08 (“Where gross statistical disparities can be shown, they alone

 9   may in a proper case constitute prima facie proof of a pattern or

10   practice of discrimination.”).       The City does not dispute that the

11   Intervenors presented a prima facie case of discriminatory treatment.

12      The City’s rebuttal.        The City produced evidence attempting to

13   rebut the inference that it had acted with a discriminatory intent.

14   It articulated a nondiscriminatory reason for using the challenged

15   exams - the fact that they were facially neutral.           The City also

16   relied on its contention that the exams had been prepared in an

17   attempt   to   comply   with   “acceptable   test   development   methods.”

18   Defendants’ Statement of Disputed Material Facts ¶ 1.       In support of

19   this contention, the City proffered detailed declarations of Matthew

20   Morrongiello, a Tests and Measurement Specialist in the City’s DCAS


          15
            This rebuttable inference arising at the relief stage, after
     proof by preponderance of the evidence at the liability stage of the
     existence of a pattern or practice of intentional discrimination,
     should not be confused with the rebuttable presumption arising at the
     threshold of the liability stage, after presentation of only a prima
     facie case of such a pattern or practice.

                                         -30-
 1   who analyzed Exam 7029, and Alberto Johnson, a DCAS employee who was

 2   primarily responsible for preparing Exam 2043. See Disparate Impact

 3   
Op., 637 F. Supp. 2d at 100
.                Their affidavits detailed the efforts

 4   that they made to develop job-related exams.16                    The City also pointed

 5   to    its    efforts     to   increase       minority         hiring   through    targeted

 6   recruitment.

 7        The    District    Court’s      rejection      of   the    City’s   rebuttal.        The

 8   District Court’s grant of summary judgment to the Intervenors on their

 9   pattern-or-practice discriminatory treatment claim might be thought to

10   mean either of two things.              On the one hand, the Court might have

11   concluded      that    the    City    had    failed      to    satisfy   its     burden   of

12   production.      On the other hand, the Court might have concluded that,

13   on the available record, no reasonable fact-finder at trial could fail

14   to find that the City maintained a pervasive policy of intentional

15   discriminatory treatment.            The Intervenors argued their motion on the

16   latter theory.         One section of their memorandum of law in support of

17   their motion is captioned “THERE IS NO GENUINE ISSUE OF FACT AS TO THE

18   CITY OF NEW YORK’S DISPARATE TREATMENT OF PLAINTIFFS-INTERVENORS UNDER

19   TITLE VII, AND PLAINTIFFS-INTERVENORS ARE ENTITLED TO JUDGMENT AS A

20   MATTER OF LAW.” Memorandum of Law in Support of Motion for Summary

21   Judgment and in Opposition to Individual Defendants’ Motion for



            16
             The District Court’s ruling against the City on the
     Government’s disparate impact claim discussed these efforts at length.
     See Disparate Impact 
Op., 637 F. Supp. 2d at 100
-08.

                                                  -31-
 1   Qualified Immunity at 8.        That memorandum does not contend that the

 2   City failed only to satisfy its burden of production.

 3     We think it is clear that the District Court granted summary

 4   judgment for the Intervenors because it is believed the City had not

 5   satisfied its burden of production.             The Court stated, “If the

 6   employer fails to respond to plaintiffs’ prima facie case, or if it

 7   fails to carry its burden to dispel the prima facie case, then the

 8   court ‘must find the existence of the presumed fact of unlawful

 9   discrimination    and   must,    therefore,     render   a   verdict    for   the

10   plaintiff.’” Disparate Treatment Op., 
683 F. Supp. 2d 252 
(quoting

11   
Hicks, 509 U.S. at 509
-10 n.3 (emphases in Hicks)).               The Court then

12   added:

13     What is important to note is that in either case, although the
14     ultimate question as to the employer’s state of mind is technically
15     left unresolved - since the fact-finder has not found by a
16     preponderance of the evidence that the employer acted with
17     discriminatory purpose - the employer’s failure to discharge the
18     obligation imposed on it by the burden-shifting framework mandates
19     a finding of unlawful discrimination.
20
21   
Id. (Citing Hicks, 509 
U.S. at 506).
22
23     The District Court deemed the City’s rebuttal deficient for four

24   somewhat related reasons.       First, the Court thought that the City’s

25   burden   of   production   required   it     specifically    to   challenge   the

26   Intervenors’ statistics and faulted the City because it did “not

27   attempt to meet or undermine the Intervenors’ statistical evidence.”

28   Disparate Treatment 
Op., 683 F. Supp. 2d at 253
.                  “This failure

29   alone,” the Court stated, was a sufficient reason to grant summary

                                           -32-
 1   judgment to the Intervenors. See 
id.               As we have 
explained above, this

 2   was too narrow a view of how a defendant may rebut a prima facie case.

 3   On the Intervenors’ motion for summary judgment, the issue for the

 4   District    Court    was    not   whether     the    City      had   produced      evidence

 5   sufficiently attacking the Intervenors’ statistics.                           Instead, the

 6   issue was whether the City’s rebuttal was sufficient to satisfy its

 7   burden of producing evidence to challenge the inference of intentional

 8   discrimination arising from the Intervenors’ prima facie case.

 9        Second, the District Court rejected the evidence the City produced

10   to   satisfy   its    burden      of   production        as    “either    incredible       or

11   inapposite.” Disparate Treatment 
Op., 683 F. Supp. 2d at 266
.                             The

12   Court’s    assessment      of   credibility        (an   assessment      of    information

13   supplied in affidavits) was inappropriate.                       Determining whether a

14   defendant has satisfied its burden of production “can involve no

15   credibility assessment.” 
Hicks, 509 U.S. at 509
. Furthermore, “[t]he

16   defendant need not persuade the court that it was actually motivated

17   by the proffered reasons.” 
Burdine, 450 U.S. at 254
.

18        Nor was the City’s rebuttal evidence “inapposite.”                    All of it was

19   properly presented in an attempt to show that the City lacked a

20   discriminatory      intent.       Although     the       Exams    produced     a   racially

21   disparate impact and were determined by the District Court not to be

22   sufficiently job-related to justify their use, see Disparate Impact

23   
Op., 637 F. Supp. 2d at 110-32
, the City was entitled to produce

24   whatever    evidence       it   had    to    rebut       the   prima     facie     case    of

                                                 -33-
1   discriminatory treatment.   That evidence properly included a showing

2   that the Exams were facially neutral, see Raytheon Co. V. Hernandez,

3   
540 U.S. 44
, 51-52 (2003) (Under “the disparate-treatment framework

4   . . . a neutral . . . policy is, by definition, a legitimate

5   nondiscriminatory reason.”), the efforts (albeit unsuccessful) to

6   prepare job-related exams,17 see NYC 
Guardians, 630 F.2d at 112
7   (noting employer’s “extensive efforts to . . . develop a test they

8   hoped would have the requisite validity”).18 and the efforts at

9   minority recruitment, see Washington v. Davis, 
426 U.S. 229
, 246


         17
            The District Court expressed the view that “the subjective
    motives of the people who designed the Exams are only circumstantially
    relevant to the question of whether the City’s decision to use the
    Exams as screening and ranking devices was discriminatory.” Disparate
    Treatment 
Op., 683 F. Supp. 2d at 254 
(emphases in original). But the
    City was entitled to produce evidence with that circumstantial
    relevance to rebut the claim that it used the Exams with
    discriminatory intent.

         The Intervenors contend that the City’s rebuttal evidence
    concerning the preparation of the challenged exams is irrelevant
    because the test-makers’ affidavits do not explain what the
    Intervenors assert is the “adverse action - here, the continued use of
    the challenged exams.” Brief for Intervenors at 128 (internal
    quotation marks omitted). However, the adverse action, for which the
    employer must supply a nondiscriminatory reason, is the failure to
    hire minority firefighters; the use of the Exams is a circumstance
    that the Intervenors contend shows that the city acted with
    discriminatory intent. That contention will be available at trial.
         18
            It defies understanding why the City would think it a virtue
    that “[t]he individuals who were principally responsible for
    developing Examinations 7029 and 2043, did not, prior to developing
    the Examinations[,] consult with counsel or review the [NYC] Guardians
    decision,” Defendants’ Statement of Disputed Material Facts ¶ 2. The
    District Court characterized NYC Guardians as the “governing case in
    this Circuit for assessing the validity of employment tests.”
    Disparate Impact 
Op., 637 F. Supp. 2d at 108
.

                                     -34-
 1   (1976) (“[A]ffirmative efforts [of municipal employer] to recruit

 2   black    officers       .    .    .     negated    any     inference      that   [employer]

 3   discriminated . . . .”).

 4        The District Court also appeared to consider the City’s evidence

 5   inapposite because, in the Court’s opinion, the City was not entitled

 6   to   “construct     a       competing     account     of    its     behavior.”    Disparate

 7   Treatment 
Op., 683 F. Supp. 2d at 253
.                       This view of the City’s

 8   rebuttal burden runs directly counter to the Supreme Court’s statement

 9   in    Teamsters     that         “the    employer’s        burden    is    to    provide   a

10   nondiscriminatory           explanation      for    the     apparently      discriminatory

11   
result.” 431 U.S. at 360 
n.46.

12        Third, the District Court viewed the City’s opposition to the

13   Intervenor’s summary judgment motion as an improper effort to dispute

14   the issue of discriminatory intent that the Court said would arise “at

15   the end of any Title VII disparate-treatment inquiry.” Disparate

16   Treatment 
Op., 683 F. Supp. 2d at 252 
(emphasis in original).                          This

17   was improper, the Court thought, because “if defendants were allowed

18   to sustain or circumvent their burden of production by invoking the

19   ultimate issue of intent, the burden-shifting structure would become

20   a nullity.” 
Id. at 253.

21 
       We disagree.    A defendant seeking to “defeat,” Teamsters, 
431 U.S.

22 
  at 360, a prima facie case of intentional discrimination at the

23   rebuttal stage has every right to produce evidence to show that it did

24   not have such an intent.                  Although a conclusory denial will not

                                                  -35-
 1   suffice,   evidence    that   tends    to    support   a   denial   is   always

 2   permissible.   When the Supreme Court said in Teamsters that the

 3   employer may satisfy its burden of production by “provid[ing] a

 4   nondiscriminatory     explanation     for    the   apparently   discriminatory

 5   
result,” 431 U.S. at 324 
n.46, it was offering an example of evidence

 6   that ws not disqualified as a rebuttal just because such evidence was

 7   also relevant to the ultimate issue of discriminatory intent.19

 8      Producing at the rebuttal stage some evidence bearing on the

 9   ultimate issue of discriminatory intent does not render the burden-

10   shifting structure a nullity.          That structure serves the useful

11   purpose of obliging the employer to identify a nondiscriminatory

12   reason for its challenged action.       If the employer fails to do so or

13   otherwise fails to produce evidence that meets the inference arising

14   from the plaintiff’s prima facie case, the employer loses. See Hicks,

15   509 U.S. at 509
.    On the other hand, producing evidence that meets the

16   prima facie case moves a pattern-or-practice claim on to trial on the

17   merits, at which time the plaintiff has to prove by a preponderance of

18   evidence that the real reason for the challenged action was an intent

19   to discriminate.      The burden-shifting scheme has not been impaired

20   just because the employer’s rebuttal not only meets the prima facie

21   case but is also relevant to the ultimate issue at trial.           Nothing in


          19
             The District Court seems to have recognized this point by
     stating that the presumption arising from the prima facie case
     “obligates the employer to come forward with an explanation or
     contrary proof.” Disparate Treatment 
Op., 683 F. Supp. 2d at 252
.

                                           -36-
 1   Teamsters obliges an employer to withhold its evidence negating a

 2   discriminatory intent until that trial occurs.

 3      Fourth, the District Court faulted the City for “attempt[ing] to

 4   circumvent its burden of production entirely by arguing that the

 5   Intervenors have not proved that the City harbored a subjective intent

 6   to discriminate against black applicants.” Disparate Treatment Op.,

 
7   683 F. Supp. 2d at 251
.     The Court understood the City to be faulting

 8   the Intervenors for “failure to produce direct evidence                 of the

 9   relevant decisionmakers’ culpable mental state.” 
Id. (emphasis added).

10 
  That was not what the City said.            In its memorandum opposing the

11   Intervenors’   motion     for   summary   judgment   on   the   discriminatory

12   treatment claim, the City stated, “Plaintiffs-Intervenors have not,

13   either directly or by inference, provided facts which would prove an

14   intent to discriminate.” Defendants’ Memorandum of Law in Opposition

15   to Plaintiffs-Intervenors’ Motion for Summary Judgment at 2 (emphasis

16   added).   Correctly understanding that a prima facie case requires

17   facts giving rise to an inference of intentional discriminatory

18   treatment, the City was entitled to contend in rebuttal that the

19   Intervenors had failed to present such facts, even though the District

20   Court had found that their prima facie case was sufficient.

21      At trial on the ultimate issue of whether there was a policy of

22   discriminatory intent, the fact-finder will consider, among other

23   things,   whether,   as   the   Intervenors    contend,   the   lack   of   job-

24   relatedness of the Exams should have been apparent to the City and

                                          -37-
 1   whether the City’s use of the Exams, once their racially disparate

 2   impact was known, proves, in light of the history of low minority

 3   hiring, that the City used the Exams with the intent to discriminate.

 4   Prior to that trial, the City provided a sufficient rebuttal to the

 5   Intervenors’ prima facie case, and the granting of the Intervenors’

 6   motion for summary judgment was error.

 7   III. Dismissal of Claims Against Mayor Bloomberg and Commissioner

 8   Scoppetta

 9      The District Court dismissed the Intervenors’ Title VII claim

10   against Mayor Bloomberg and former Commissioner Scoppetta for failure

11   to state a claim on which relief could be granted, see Fed. R. Civ. P.

12   12(b)(6); dismissed the Section 1981 and Section 1983 claims against

13   these officials on the ground of qualified immunity; and dismissed the

14   state law claims against these officials on the ground of official

15   immunity. See Disparate Treatment 
Op., 683 F. Supp. 2d at 243-45
, 269-

16   72.20        On their cross-appeal, the Intervenors challenge the immunity

17   rulings, but not the Rule 12(b)(6) ruling, which was plainly correct,

18   see Patterson v. County of Oneida, 
375 F.3d 206
, 221 (2d Cir. 2004)

19   (individuals, as distinguished from employing entitles, not liable

20   under Title VII).

21      Qualified immunity for federal law claims.           The standards for

22   qualified immunity are well settled. See Anderson v. Creighton, 483


             20
            The District Court certified its dismissal ruling for immediate
     appeal under Fed. R. Civ. P. 54(b).

                                           -38-
 
1   U.S. 635
, 640 (1987); Mitchell v. Forsyth, 
472 U.S. 511
, 530-36

 2   (1985); Harlow v. Fitzgerald, 
457 U.S. 800
, 818 (1982).

 3      The District Court began its immunity analysis by observing that

 4   “to hold a supervisory official liable for violating § 1981 or the

 5   Equal Protection Clause, a plaintiff must ‘prove that the defendant

 6   acted with discriminatory purpose.’” Disparate Treatment Op., 
683 F.

 7 
  Supp. 2d at 270 (quoting Ashcroft v. Iqbal, 
556 U.S. 662
, 676 (2009)).

 8   The Intervenors’ theory of liability of the Mayor and the Commissioner

 9   was that, although neither had responsibility for preparation of the

10   Exams, they both condoned their use, with awareness of their disparate

11   impact, and did so despite warnings from the City’s Equal Employment

12   Practices Commission to take corrective action.   With respect to this

13   theory, the District Court stated, “The Intervenors have submitted

14   copious evidence from which a reasonable fact-finder could infer that

15   the Mayor and Commissioner harbored an intent to discriminate against

16   black applicants . . . .” 
Id.    Nevertheless, the Court 
upheld the

17   officials’ qualified immunity defense because “it would not have been

18   clear to them from the governing legal precedent that such conduct

19   violated § 1981 or the Equal Protection Clause.” 
Id.

20      The Court 
did not mean that a public official would not have known

21   that the official would violate Section 1981 or the Equal Protection

22   Clause by intentionally taking adverse employment action on the basis

23   of race. That obvious proposition has been clear at least since 1976,

24   see 
Washington, 426 U.S. at 239-41
.    What would not be clear to the

                                     -39-
 1   officials, the District Court stated, was that the “Title VII burden-

 2   shifting analysis” would apply “to determine whether an individual, as

 3   opposed to a governmental employer, is liable for discrimination under

 4   either § 1961 or the Equal Protection Clause.” Disparate Treatment

 5   
Op., 683 F. Supp. 2d at 270 
(emphasis in original).

 6      In grounding qualified immunity on this rationale the District

 7   Court erred.   The knowledge of a standard governing the conduct of

 8   public officials, required to defeat a claim of qualified immunity, is

 9   knowledge of primary conduct - action of an official that would

10   violate constitutional limitations.          It has nothing to do with

11   secondary   conduct   of   litigation   of   a   claim   of   constitutional

12   violation. Cf. Republic of Austria v. Altmann, 
541 U.S. 677
, 722

13   (2004) (Kennedy, J., with whom Rehnquist, C.J. and Thomas, J., join,

14   dissenting) (distinguishing, for purposes of retroactivity, between

15   statutes that “‘regulate the secondary conduct of litigation and not

16   the underlying primary conduct of the parties’”) (quoting Hughes

17   Aircraft Co. v. United States ex rel. Schumer, 
520 U.S. 939
, 951

18   (1997).   If a public official intentionally acts to the detriment of

19   current or prospective public employees on the basis of race, the

20   official is not shielded by qualified immunity simply because the

21   official might have been unaware that at trial a burden-shifting

22   scheme would regulate the conduct of ensuing litigation.             “For a

23   constitutional right to be clearly established, its contours must be

24   sufficiently clear that reasonable official would understand that what

                                        -40-
 1   he is doing violates that right.” Hope v. Pelzer, 536 U.S 730, 739-41

 2   (2002) (emphases added) (internal quotation marks omitted).21

 3      Having rejected the District Court’s stated reason for dismissing

 4   the federal claims on the ground of qualified immunity, we next

 5   consider whether the record supports dismissal of these claims on the

 6   ground that the Intervenors have not shown a violation of a federal

 7   right.    The District Court did not reach that component of qualified

 8   immunity, see Siegert v. Gilley, 
500 U.S. 226
, 232 (1991), accepting

 9   instead the opportunity created by Pearson v. Callahan, 
555 U.S. 223
,

10   236 (2009), to decide first whether the right alleged to have been

11   violated was clearly established. See Disparate Treatment Op., 
683 F.

12 
  Supp. 2d at 270 n. 32.

13      In considering whether the record would have permitted dismissal on

14   the ground that the officials had not violated a federal right, we

15   encounter two conflicting statements in the District Court’s opinion.

16   On the one hand, the Court referred to “copious evidence” from which

17   a reasonable fact-finder could infer that the officials “harbored” an

18   intent    to   discriminate   against   black    applicants.   See   Disparate

19   Treatment 
Op., 683 F. Supp. 2d at 270
.          On the other hand, the Court

20   stated that there was “no evidence that directly and unmistakably



          21
             In any event, Title VII burden-shifting procedures have
     previously been applied to suits under both Section 1981, see Hudson
     v. International Business Machines Corp., 
620 F.2d 351
, 354 (2d Cir.
     1980), and Section 1983, see Annis v. County of Westchester, 
136 F.3d
     239
, 245 (2d Cir. 1998).

                                         -41-
 1   proves that fact.” 
Id.

 2      We question 
both observations.         As to the second one, there is no

 3   requirement that an intent to discriminate must be proved “directly

 4   and mistakably.”    Like any element in a civil case, the element of

 5   discriminatory intent need be proven only by a preponderance of the

 6   evidence. See 
Teamsters, 431 U.S. at 336
.           And intent, like any state

 7   of mind, may be proved by circumstances reasonably supporting an

 8   inference of the requisite intent. See, e.g., Blue v. Koren, 
72 F.3d

 9 
  1075, 184 (2d Cir. 1995) (requiring “particularized evidence of direct

10   or circumstantial facts” bearing on improper motive in order to resist

11   defendant’s motion for summary judgment).

12      At the same time, we cannot agree with the District Court that the

13   record   revealed   “copious    evidence”    of     the    officials’    intent    to

14   discriminate.       As    the   Supreme     Court    has     recently     observed,

15   “[P]urposeful discrimination require more than intent as volition or

16   intent   as   awareness    of   consequences.         It    instead     involves    a

17   decisionmaker’s undertaking a course of action because of, not merely

18   in spite of [the action’s] adverse effects upon an identifiable

19   group.” 
Iqbal, 556 U.S. at 676-77 
(internal quotation marks and

20   citation omitted; brackets in original); see Personnel Administrator

21   v. Feeney, 
442 U.S. 256
, 279 (1979) (noting that “discriminatory

22   purpose” implies not just that the decisionmaker possessed “intent as

23   awareness of consequences” but that he “selected or reaffirmed a

24   particular course of action at least in part because of, not merely in

                                          -42-
 1   spite of, its adverse effects upon an identifiable group”) (internal

 2   quotation marks omitted).         The record contains ample evidence of the

 3   officials’ awareness of the disparate impact of the Exams, and they do

 4   not dispute such awareness.         Whether it contains sufficient evidence

 5   that they undertook a course of action because of that impact requires

 6   further consideration.

 7      Most of the evidence cited by the Intervenors to satisfy their

 8   burden   of   resisting   the     officials’       motion    for     summary      judgment

 9   concerns steps the officials did not take, rather than a “course of

10   action” they undertook.          For example, the Intervenors point to the

11   failure to have the Exams validated prior to their continued use and

12   their failure to move promptly to develop a new valid exam.                       Although

13   we do not doubt that the failure of senior officials to act can

14   support an inference of discriminatory intent in some circumstances,

15   particularly    where     they    are   in     a     position      to     avoid     likely

16   unconstitutional    consequences,       see,       e.g.,    United      States    ex   rel.

17   Larkins v. Oswald, 
510 F.2d 583
, 589 (2d Cir. 1975) (Corrections

18   Commissioner liable for unwarranted solitary confinement of inmate),

19   we do not believe that the cited omissions of the Mayor or the

20   Commissioner suffice to support a reasonable inference that they

21   declined to act because they wanted to discriminate against black

22   applicants.

23      The principal evidence of a course of action arguably undertaken

24   for the purpose of discrimination is the decision to continue using

                                             -43-
 1   the results of the Exams with awareness of their disparate impact.

 2   Although we disagree with the District Court that there was “copious

 3   evidence” of the officials’ intent to discriminate, we cannot say that

 4   a reasonable fact-finder might not infer, from all the evidence, that,

 5   with respect to the Commissioner heading the FDNY, his involvement in

 6   the decision to continue using the results of the Exams indicated an

 7   intent to discriminate.       Were the decision ours to make, we would not

 8   draw such an inference, but our task is the more limited one of

 9   determining whether such an inference could reasonably be made by the

10   fact-finder.    With respect to the Mayor, however, we think the record

11   does   not   suffice    to   permit   a    fact-finder       to   draw    a   reasonable

12   inference of intent to discriminate.               In light of the myriad duties

13   imposed upon the chief executive officer of a city of eight million

14   people, more evidence would be needed to permit a trier to find that

15   the decision of one municipal department to continue using the results

16   of the Exams supports an inference of discriminatory intent on the

17   part of the Mayor.

18      Official Immunity for state law claims. The common-law doctrine of

19   official     immunity   shields   public         employees    “from      liability   for

20   discretionary actions taken during the performance of governmental

21   functions” and “is intended to ensure that public servants are free to

22   exercise their decision-making authority without interference from the

23   courts.” Valdez v. City of New York, 
18 N.Y.3d 69
, 75-76 (2011)

24   (municipal immunity).         Here, as explained by the District Court,

                                               -44-
 1   Disparate Treatment 
Op., 683 F. Supp. 2d at 270
-72, the decision of

 2   two   of   the   City’s   highest-ranking      officials   to   continue   hiring

 3   firefighters from eligibility lists based on the Exams involved

 4   discretionary decisionmaking.

 5      We therefore affirm the District Court’s decision to dismiss

 6   federal and state law claim against Mayor Bloomberg on the grounds of

 7   qualified and official immunity, affirm the decision to dismiss state

 8   law claims against Commissioner Scoppetta on the ground of official

 9   immunity, but vacate the decision to dismiss federal law claims

10   against Commissioner Scoppetta on the grounds of qualified immunity.

11   IV. Scope of the Injunction

12      “[T]he scope of a district court’s remedial powers under Title VII

13   is determined by the purposes of the Act.” 
Teamsters, 431 U.S. at 364
.

14   Congress enacted Title VII to achieve equal employment opportunities

15   and “to eliminate those discretionary practices and devices which have

16   fostered racially stratified job environments to the disadvantage of

17   minority citizens.” McDonnell 
Douglas, 411 U.S. at 800
.               “Congress

18   deliberately gave the district courts broad authority under Title VII

19   to fashion the most complete relief possible.” Local 28 Sheet Metal

20   Workers’ International Ass’n v. EEOC, 
478 U.S. 421
, 465 (1986).              Once

21   liability for racial discrimination has been established, a district

22   court has the duty to render a decree that will eliminate the

23   discretionary     effects   of   past    discrimination     and   prevent    like

24   discrimination in the future. See Albemarle Paper Co., 422 U.S. at

                                             -45-
 1   418; Bridgeport Guardians, Inc. v. City of Bridgeport, 
933 F.2d 1140
,

 2   1149 (2d Cir. 1991).    Although a court’s power to fashion appropriate

 3   relief is not unlimited, see Bridgeport 
Guardians, 933 F.2d at 1149
,

 4   we have held that, when it appear “that the employer has discriminated

 5   prior to the use of the challenged selection procedure, then it may

 6   also be appropriate to fashion some form of affirmative relief, on an

 7   interim    and   long-term   basis,   to     remedy      past    violations,”   NYC

 8   
Guardians, 630 F.2d at 108
.

 9      The    District   Court   expressed     the    view    that    its   conclusion

10   concerning the need for “close and continuing supervision” is “as

11   applicable to City’s violation of the disparate impact provisions of

12   Title VII as it is to the City’s intentional discrimination against

13   black firefighter candidates.” City of New York, 
2011 WL 4639832
, at

14   *11 (E.D.N.Y. Oct. 5, 2011).      The Intervenors endorse this view and,

15   somewhat extending it, suggest that we should uphold all provisions of

16   the injunction solely on the basis of the unchallenged disparate

17   impact ruling. The City contends that the District court exceeded its

18   discretion by entering an injunction that goes beyond the scope of the

19   Title VII disparate impact violation.            In the City’s view, the only

20   provisions of the injunction that may be sustained as relief for its

21   disparate impact liability are those that require a lawful method of

22   testing and a limitation on interim hiring until a valid exam is




                                           -46-
 1   prepared.22   Any more intrusive remedies, the City argues, are not

 2   warranted in the absence of a valid finding of a pattern or practice

 3   of intentional discrimination, and perhaps not even then.

 4      We disagree with the positions of both sides.     We think that in

 5   some respects the injunction contains provisions that go beyond what

 6   would be appropriate to remedy only the disparate impact liability,

 7   and, because we have vacated the ruling granting summary judgment for

 8   the Intervenors on the disparate treatment claim, we will uphold only

 9   those provisions of the injunction that are appropriate as relief for

10   the City’s liability on the Government’s disparate impact claim.   On

11   the other hand, whatever the dimensions of an appropriate remedy for

12   a straightforward case involving only the disparate impact of a hiring

13   exam, considerably more relief is warranted in this case in light of

14   the distressing pattern of limited FDNY minority hiring.    Even after

15   the 1973 determination that a hiring exam was invalid because of a

16   racially disparate impact, see Vulcan Society of New York City Fire

17   
Dep’t, 360 F. Supp. at 1269
, the City’s percentage of black entry-

18   level firefighters has remained at or below 4 percent for several

19   decades, and the current percentage of 3.4 percent compares woefully

20   to the 16.6 percent achieved by the city’s Police Department and the

21   61.4 percent achieved by the City’s Corrections Department.   Although


          22
            We note that the District Court recently approved the City’s
     use of the results of a new exam for entry-level firefighters. See
     United States v. City of New York, No. 07-CV-2067, 
2012 WL 4503253
     (Sept. 28, 2012).

                                     -47-
 1   some provisions of the injunction cannot be justified in the absence

 2   of a finding of discriminatory intent, many provisions are well within

 3   the District Court’s discretion as a remedy for discriminatory impact

 4   liability in view of the history of minority hiring by the FDNY and

 5   the City’s recalcitrance in undertaking remedial steps.

 6      The “General Terms” of the Injunction enjoin the use of the

 7   challenged exams and prospectively prohibit discrimination against

 8   black        or   Hispanic      applicants   for    the    position       of    entry-level

 9   firefighter.        See    Injunction    Op.,   
2011      WL   6131136
,    at    *4.    The

10   “Specific Remedial Measures” section of the Injunction focuses on five

11   substantive areas: Firefighter Test Development and Administration,

12   Firefighter Candidate Recruitment, Attrition Mitigation Plan and

13   Reassessment of Entry-Level Firefighter Selection, Post-Examination

14   Firefighter Candidate Screening, and EEO Compliance Reform. 
Id. at *4-

15 
  *13.

16      We        describe     the     specific   provisions        of   the   injunction    in

17   abbreviated form.23             Paragraph 1424 bars the use of Exams 7029, 2043,

18   and 6019,25 and paragraph 15 bars the use of any exam with a disparate

19   impact against blacks or Hispanics that is not job-related.



             23
            The wording of these summary statements is not to be understood
     as varying the specific terms of the injunction.
             24
                  Paragraphs 1-13 define terms used in the injunction.
             25
            The injunction’s prohibition of the use of Exam 6019 replaces
     the interim permission previously given to use that exam.

                                                  -48-
 1      Paragraph 16 requires approval of the Monitor before taking any

 2   step in the hiring process.

 3      Paragraph     17     bars     retaliation      for     complaining      against

 4   discrimination.       Paragraph 18 bars discrimination against black or

 5   Hispanic firefighter candidates.         The first sentence of paragraph 19

 6   requires   the      elimination     of    all     vestiges      of   intentional

 7   discrimination; the second sentence requires the elimination of all

 8   policies and practices that have a discriminatory impact on black and

 9   Hispanic firefighter candidates.

10      Paragraph   20     requires   compliance     with    the   instuction   of   the

11   Monitor.

12      Paragraph 21 specifies, with certain exceptions, that all required

13   submissions be signed by the fire Commissioner and the Corporation

14   Counsel and be reviewed and approved by the Mayor.

15      Paragraphs 22-24 require prior notice to the Monitor and the

16   parties concerning new hiring and details of the preparation of new

17   eligibility lists.

18      Paragraphs 25-30 require recruitment efforts, including the hiring

19   of a recruitment consultant, the preparation of a recruitment report,

20   and either compliance with the consultant’s recommendations or an

21   explanation for not following them.

22      Paragraphs 31-36 require steps to mitigate attrition during the

23   selection process.

24      Paragraphs 37-46 require various steps to be taken after exams are

                                           -49-
 1   administered.     Paragraphs 37 and 38 require a detailed written record

 2   of any oral conversations that concern a candidate.                 Paragraph 39

 3   requires designation of a senior official to enforce the writing

 4   requirement.      Paragraphs 40 and 41 require written procedures for

 5   conducting background investigations of candidates.

 6      Paragraphs 47-51 require various steps, including appointment of an

 7   EEO consultant, to assure compliance with equal employment opportunity

 8   requirements.

 9      Paragraphs 52 and 53 require development of and compliance with a

10   document retention policy.      Paragraphs 54 and 55 require discovery

11   through document production and deposition availability to assure

12   compliance with the injunction.

13      Paragraph 56 authorizes sanctions for noncompliance.

14      Paragraphs 57-77 appoint Mark S. Cohen as Monitor, specify his

15   duties, and authorize necessary staff.

16      Paragraphs 78-80 provide for the retention of jurisdiction until at

17   least January 1, 2022.      Paragraphs 81 and 82 require the City to pay

18   costs, attorney’s fees, and all expenses.

19      After reviewing these provisions in light of the unchallenged

20   disparate impact finding, the absence (as yet) of a proper disparate

21   treatment finding, the FDNY’s record of minimal minority hiring, and

22   the   District    Court’s   broad,     but    not   limitless,     discretion   in

23   fashioning   appropriate     relief,    we     conclude   that     the   principal

24   components   of   the   injunction   are      appropriate,   but    that   several

                                            -50-
 1   modifications must be made.

 2      In addition to proscribing use of the invalid exams and preparation

 3   of valid exams, the District Court was entirely warranted in ordering

 4   significant affirmative relief (although declining to order any hiring

 5   quota), including appointing a Monitor to oversee the FDNY’s long-

 6   awaited progress toward ending discrimination, ordering development of

 7   policies to assure compliance with anti-discrimination requirements,

 8   requiring efforts to recruit minority applicants, ordering steps to

 9   lessen minority attrition, ordering a document retention policy, and

10   requiring comprehensive review of the entire process of selecting

11   entry-level firefighters. However, we believe several provisions must

12   be modified or deleted, primarily because of our vacating the grant of

13   summary judgment on the disparate treatment claim.

14      Paragraph 19 must be modified to delete the first sentence, which

15   is based on a finding of intentional discrimination that we have

16   vacated subject to further proceedings. The second sentence generally

17   barring policies and practices with a disparate impact must also be

18   modified to bar only those policies and practices not job-related or

19   required by business necessity.

20      Paragraph 21 must be modified to eliminate approval of submissions

21   by the Corporation Counsel, who is not a party to this litigation, and

22   the   Mayor,   whose   dismissal   we    have   affirmed.   Although   we   can

23   understand the District Court’s concern in litigation against the City

24   to have the City’s chief executive officer and chief legal officer

                                             -51-
 1   assume direct responsibility for all submissions, these requirements

 2   are an excessive intrusion into the duties of officials charged with

 3   citywide responsibilities, in the absence of either their liability or

 4   an indication that imposing requirements on the head of the relevant

 5   department will be inadequate.

 6        Paragraphs 26-29 must be modified to eliminate the requirement of

 7   an   outside   recruitment        consultant       and,   instead,   to    assign   the

 8   consultant’s tasks to appropriate City employees. Although the record

 9   warrants performance of these tasks, it does not require burdening the

10   City with the extra expenses of an outside consultant.                    In the event

11   that the Monitor determines that designated City employees are not

12   adequately performing their functions, he may apply to the Court for

13   designation of an outside consultant.                Paragraph 29 must be further

14   modified, for the same reason applicable to paragraph 21, to eliminate

15   the requirement of the Mayor’s approval.

16        Paragraphs 34-36 must be modified, for the same reason applicable

17   to paragraph 21, to eliminate the Mayor’s obligations and substitute

18   those of the Fire Commissioner.

19        Paragraphs   37-39      must    be    eliminated.        The    requirement     of

20   contemporaneous written records of all communications concerning

21   hiring is far too intrusive, at least in the absence of a finding of

22   intentional discrimination.

23        Paragraphs   40   and   41     must   be     modified   to   eliminate,   as   too

24   intrusive, the detailed requirements for CID and PRB policies and

                                                -52-
 1   procedures; the requirement of developing written procedures that are

 2   subject to the Monitor’s approval remains.

 3      Paragraph 42 must be eliminated, for the same reason applicable to

 4   paragraphs 26-29.

 5      Paragraph 43 must be eliminated as imposing too great a burden on

 6   the Monitor, although the Monitor will remain eligible to attend any

 7   PRB meeting.

 8      Paragraph 45 must be modified, for the same reason applicable to

 9   paragraph 21, to eliminate the requirement of the Mayor’s signature

10   and certification.

11      Paragraphs 47-51 must be modified, for the same reason applicable

12   to Paragraphs 26-29, to eliminate the requirement of an outside EEO

13   consultant and to assign the consultant’s tasks to the FDNY’s EEO

14   Office. In the event that the Monitor determines that designated City

15   employees are not adequately performing their functions, he may apply

16   to the Court for designation of an outside consultant.           Paragraph 50

17   must be further modified, for the same reason applicable to Paragraph

18   21,   to   eliminate   the   requirement   of   the   Mayor’s   signature   and

19   certification.

20      Paragraph 54 must be modified to change “any additional document”

21   to “any non-privileged documents.”

22      Paragraphs 66 and 68 must be modified to change “short notice” to

23   “reasonable notice,” and paragraph 67 must be modified to change “one

24   week” to 30 days.”

                                         -53-
 1      Paragraph 71 must be modified to add “The City may apply to the

 2   Court, upon reasonable notice to the parties, to end the employment of

 3   some    or   all    of    the    Monitor’s   staff   and   consultants   upon     a

 4   demonstration that the City has satisfied its burden of proof as

 5   specified in modified Paragraph 78.”

 6      Paragraph 78 must be modified to change “and nor” to “nor” in

 7   subparagraph       (a),   and    to   eliminate   subparagraphs   (e)    and    (f)

 8   concerning intentional discrimination.

 9      Paragraph 79 must be modified in subparagraph (a) to change “2022"

10   to “2017,”, and subparagraph (b) must be modified to change “second of

11   the City’s next two civil service hiring lists” to “City’s next civil

12   service hiring list.”           An extended retention of jurisdiction is not

13   warranted in the absence of a finding of intentional discrimination.26

14      Paragraph 80 must be eliminated. The City is entitled to undertake

15   to satisfy its burden of proof to be relieved of the injunction’s

16   prospective requirements whenever it believes it can do so.

17      Paragraph 83 must be modified to change “and disparate treatment

18   claims that were” in line 3 to “claim that was,”; to change “and

19   Disparate Treatment Opinions” in lines 4-5 to “Opinion”; and, to



            26
            We note that the District Court previously stated, “If after
     the bench trial the court concludes that the City has shown that,
     among other things, it has ended its discriminatory hiring practices
     and taken sufficient affirmative measures to end the policies and
     practices that have perpetuated the harmful effects of those
     discriminatory hiring practices and procedures, the court will
     relinquish jurisdiction.” City of New York, 
2011 WL 4639832
, at *15.

                                              -54-
 1   change “those claims” in line 8 to “that claim.”

 2        Although we have made several modifications, primarily in view of

 3   the fact that a proper finding of intentional discrimination has not

 4   been made, we leave in place the many provisions that the District

 5   Court has wisely required in order not only to remedy the disparate

 6   impact of the challenged exams and but also to put the FDNY on a

 7   course toward future compliance with Title VII.

 8        As modified, the injunction is affirmed.27

 9   V. The City’s Claim for Reassignment to a Different Judge

10        The City contends that, in the event of a remand, the case should

11   be reassigned to a different district judge because of what it alleges

12   is bias on the part of Judge Garaufis.               That is an extreme remedy,

13   rarely imposed, see United States v. Jacobs, 
955 F.2d 7
, 10 (2d Cir.

14   1992) (reassignment is an “extraordinary remedy” reserved for the

15   “extraordinary         case”)   (internal     quotation    marks     omitted),        but

16   occasionally warranted, even in the absence of bias, to avoid an

17   appearance       of    partiality,   see     Hispanics    for   Fair     &    Equitable

18   Reapportionment v. Griffin, 
958 F.2d 24
, 26 (2d Cir. 1992) (“firmness”

19   of     judge’s    views    warranted   reassignment       on    remand       to   assure

20   “appearance of justice”); United States v. Robin, 
553 F.2d 8
, 10 (2d

21   Cir.    1977)    (in    banc)   (“Absent     proof   of   personal     bias       .   .   .

22   reassignment is advisable to preserve the appearance of justice

23   . . . .”).

            27
             We assume the District Court will enter a new injunction
     reflecting the modifications we have required.

                                                -55-
 1      Although the District Judge expressed several criticisms of the

 2   FDNY, we see no basis to require reassignment of the entire case to a

 3   different judge.     However, one aspect of the Judge’s handling of the

 4   case thus far warrants a limited form of reassignment.              In granting

 5   summary judgment ro the Intervenors on their pattern-or-practice

 6   discriminatory treatment claim, Judge Garaufis stated that the City’s

 7   rebuttal evidence in opposition to that claim was “either incredible

 8   or inapposite.” Disparate Treatment 
Op., 683 F. Supp. 2d at 266
.

 9      This assessment is cause for concern for two reasons.             First, in

10   considering the sufficiency of the City’s rebuttal evidence, the

11   District Court’s task was only to determine whether the City’s

12   rebuttal evidence satisfied the City’s burden of production.            But the

13   Court went beyond that task and granted summary judgment to the

14   Intervenors. Although summary judgment at the preliminary stage might

15   be proper in a rare case, the Intervenors have not cited any case, and

16   we have found none, in which an employer’s rebuttal evidence in a

17   discriminatory treatment case resulted in a summary judgment for the

18   plaintiff.     Second, and more important, it was improper for the

19   District Court to make any assessment of credibility in considering

20   the sufficiency of the City’s rebuttal to the Intervenors’ prima facie

21   case. See 
Hicks, 509 U.S. at 509 
(determining whether a defendant has

22   satisfied    its   burden   of   production   “can   involve   no   credibility

23   assessment”). The Court not only assessed credibility but did so after

24   considering only affidavits.

                                           -56-
 1      We have no doubt that Judge Garaufis is an entirely fair-minded

 2   jurist who could impartially adjudicate the remaining issues in this

 3   case, but we think a reasonable observer would have substantial doubts

 4   whether the judge, having branded the City’s evidence “incredible,”

 5   could thereafter be impartial in assessing the truth of conflicting

 6   evidence at a bench trial, the parties having waived a jury trial.   Of

 7   course, if any judge were to find a witness’s testimony incredible

 8   when appropriately acting as a bench trial finder of fact, that would

 9   not prevent that judge from determining the facts at future bench

10   trials at which that same witness will testify, even though a similar

11   assessment of the witness’s credibility would be likely.    Defendants

12   relying on the same witness in a succession of separate bench trials

13   are not entitled to a succession of different trial judges just

14   because their witness was disbelieved at the first trial.   But where,

15   as here, a judge makes an unwarranted venture into fact-finding at a

16   preliminary stage and brands a party’s evidence as “incredible”

17   without hearing any witnesses, an objective observer would have a

18   reasonable basis to question the judge’s impartiality in assessing

19   that evidence at trial.28 See Pescatore v. Pan American World Airways,

20   Inc., 
97 F.3d 1
, 21 (2d Cir. 1996) (“To reassign a case on remand, we

21   need not find actual bias or prejudice, but only that the facts might


          28
             We note that this is the unusual case where the risk of an
     appearance of partiality is identified on an interlocutory appeal at
     a preliminary stage of the litigation and can be avoided prospectively
     without undoing a proceeding already concluded.

                                     -57-
 1   reasonably cause an objective observer to question [the judge’s]

 2   impartiality.”) (internal quotation marks omitted).

 3      This conclusion, however, does not warrant the City’s requested

 4   relief of reassigning the entire case to a different judge.        The

 5   appearance of impartiality would be limited to Judge Garaufis’s

 6   conduct of a bench trial on the liability phase of the Intervenors’

 7   remanded disparate treatment claim, and it is only that phase of the

 8   future proceedings that needs to be conducted by a different judge.

 9      This reassignment of a portion of the case to a different judge

10   will potentially create an issue as to implementation of injunctive

11   relief.   The District Court will need to (1) supervise implementation

12   of the portions of the injunction we have affirmed with respect to the

13   Government’s disparate impact claim and, if the Intervenors pursue and

14   prevail on their disparate treatment claim, (2) fashion any additional

15   relief that might be warranted and supervise the implementation of any

16   such relief.   We leave to the District Court the task of determining

17   the appropriate supervision role or roles of Judge Garaufis and/or

18   whichever judge is assigned to preside at the trial of the liability

19   phase of the disparate treatment claim.    In the unlikely event that

20   these two judges cannot agree on their appropriate roles, any party

21   may apply to this Court for further relief.   Pending a ruling in favor

22   of the Intervenors on their disparate treatment claim, if pursued,

23   Judge Garaufis may continue supervising implementation of the portions

24   of the injunction we have affirmed.

                                      -58-
 1      The federal rules permit separate trials of separate issues, see

 2   Fed. R. Civ. P. 42(b), and we see no obstacle to having a second judge

 3   try a separate issue where a bench trial of that issue by the first

 4   judge risks an appearance of partiality.29

 5                                 Conclusion

 6      The grant of summary judgment to the Intervenors on their disparate

 7   treatment claim is vacated; the dismissal of the federal and state law

 8   claims against Mayor Bloomberg is affirmed, as is the dismissal of the

 9   state law claims against Commissioner Scoppetta; the dismissal of the

10   federal law claims against Commissioner Scoppetta is vacated; the

11   injunction is modified and, as modified, is affirmed; and the case is

12   remanded with directions that the bench trial on the liability phase

13   of the Intervenors’ disparate treatment claim against the City will be

14   reassigned to a different district judge.

15      Affirmed in part, vacated in part, and remanded.




          29
            Dividing aspects of a single case between two judge of the same
     court is doubtless unusual, but our Court has taken the even more
     unusual course of sending bifurcated issues in a single case to two
     different courts. When the Temporary Emergency Court of Appeals
     (“TECA”) existed for handling appeals concerning issues arising under
     the Economic Stabilization Act (“ESA”), our Court divided appeals
     containing such issues and sent the ESA issues to the TECA court and
     kept the remaining issues (often antitrust issues) in our Court. See
     Coastal States Marketing, Inc. v. New England Petroleum Corp., 
604
     F.2d 179
, 186-87 (2d Cir. 1979).

                                     -59-
POOLER, Circuit Judge, dissenting:

       I concur in the majority opinion in its statements of the controlling law and conclusions

in Parts I, III, IV, and V. I respectfully dissent however, as to Part II, because I believe that the

majority incorrectly applies the standard for cases involving an individual Title VII violation to

this pattern-or-practice case. By requiring the City merely “to articulate some legitimate,

nondiscriminatory reason” for the adverse action as set out in McDonnell Douglas Corp. v.

Green, 
411 U.S. 792
, 802 (1973), rather than demand it “demonstrat[e] that the [plaintiff’s] proof

is either inaccurate or insignificant” as required by International Brotherhood of Teamsters v.

United States (Teamsters), 
431 U.S. 324
, 360 (1977), the majority conflates the two distinct tests

set out in our disparate treatment jurisprudence.1

I. Two Different Burden-Shifting Frameworks: McDonnell Douglas and Teamsters

       The Supreme Court has “devised a series of shifting evidentiary burdens” for the

different types of disparate treatment claims under Title VII “that are intended progressively to

sharpen the inquiry into the elusive factual question of intentional discrimination.” Watson v.

Fort Worth Bank & Trust, 
487 U.S. 977
, 986 (1988) (internal citation and quotation marks

omitted). In McDonnell 
Douglas, 411 U.S. at 802
, the Court set out the burdens for individual

claims, and in 
Teamsters, 431 U.S. at 336
, it established the analysis for pattern-or-practice

claims. See Robinson v. Metro-North Commuter R.R. Co., 
267 F.3d 147
, 157 n.3 (2d Cir. 2001).

“The McDonnell Douglas and Teamsters frameworks differ in important respects.” Reynolds v.

Barrett, 
685 F.3d 193
, 204 (2d Cir. 2012). Although their structures are essentially the same,


       1
        Concluding the district court correctly applied Teamsters in its grant of summary
judgment, I believe reassignment to a different district judge is unnecessary, however, I
nonetheless concur with the majority opinion’s conclusion in Part V.

                                                  1
“the content of the specific stages” of these frameworks are “different.” Coates v. Johnson &

Johnson, 
756 F.2d 524
, 532 (7th Cir. 1985). These differences exist to address the distinct

purposes of the separate types of claims.

       Individual suits aim to uncover discrimination where a plaintiff “has been subjected to

‘disparate treatment’ because of his race.” Furnco Constr. Corp v. Waters, 
438 U.S. 567
, 582

(1979) (Marshall, J., concurring in part, dissenting in part). Under the McDonnell Douglas

framework the individual plaintiff is required to show that “(1) he is a member of a protected

class; (2) he was qualified for the position he held; (3) he suffered an adverse employment

action; and (4) the adverse action took place under circumstances giving rise to [an] inference of

discrimination.” Ruiz v. Cnty. of Rockland, 
609 F.3d 486
, 492 (2d Cir. 2010); see also

McDonnell 
Douglas, 411 U.S. at 802
. If the plaintiff establishes its prima facie case, the burden

then shifts to the employer who need only “articulate some legitimate, nondiscriminatory reason

for the employee’s rejection.” McDonnell 
Douglas, 411 U.S. at 802
. If the employer is able to

satisfy that burden, the inquiry then returns to the plaintiff, to demonstrate that the proffered

reason is a pretext for discrimination. 
Id. at 804.

       In 
contrast, the Supreme Court announced in Teamsters that “pattern-or-practice” claims

brought under Section 707 of Title VII2 aim to thwart widespread procedures that fall harshly on

one racial group rather than mere isolated instances. 
Teamsters, 431 U.S. at 336 
n.16. Unlike

McDonnell Douglas, the plaintiff must make a prima facie showing of pattern-or-practice under


       2
         Pattern-or-practice claims originate from section 707(a) of the Civil Rights Act of 1964,
42 U.S.C. § 2000e-6(a), which states where “persons [are] engaged in a pattern or practice of
resistance to the full enjoyment of any of the rights secured by this subchapter, and that the
pattern or practice is of such a nature and is intended to deny the full exercise of the rights” the
Attorney General may bring a civil action.

                                                  2
Teamsters “that unlawful discrimination has been a regular procedure or policy followed by an

employer or group of employers.” 
Teamsters, 431 U.S. at 360
. If the plaintiff meets its initial

burden, the employer must then “defeat” that showing by “demonstrating that the Government’s

proof is either inaccurate or insignificant.” 
Id. (emphasis added). If 
the employer fails to satisfy

its burden of production, the suit then concludes the liability phase and enters the remedial

phase. 
Id. at 361. If 
the defendant satisfies the burden of production, the district court “must

then determine, by a preponderance of the evidence, whether the employer engaged in a pattern

or practice” in order for it to “fashion class-wide injunctive relief.” 
Reynolds, 685 F.3d at 203
(internal quotation marks and citations omitted).

       Teamsters’ detailed scheme differs from that of McDonnell Douglas in two important

respects with regards to this case. First, unlike a plaintiff bringing an individual claim under

McDonnell Douglas, a plaintiff establishing a prima facie case of pattern-or-practice is often

required to use statistics. 
Id. at 339 (stating 
statistics have and will continue to serve an

important role in pattern-or-practice cases); see also 
Robinson, 267 F.3d at 157 
n.3. Second, in a

pattern-or-practice case governed by Teamsters, the defendant must also satisfy a more difficult

burden by responding to the plaintiffs’ prima facie case. Thus, where a plaintiff used statistical

evidence to prevail, a Teamsters defendant may not merely “articulate some legitimate,

nondiscriminatory reason” for the alleged discrimination, as required under McDonnell 
Douglas,

411 U.S. at 802
, instead the employer must “meet” the Government’s proof, 
Teamsters, 431 U.S.

at 360 
n.46, by demonstrating the plaintiff’s statistics were inaccurate or insignificant. 
Id. at




                                                  3
360.3 
Thus, the burdens of proof for both plaintiff and defendant are heavier under Teamsters.4

       To understand the reasoning underlying Teamsters’ more demanding standards for

plaintiffs and defendants, in comparison to that under McDonnell Douglas, we need only observe

the separate types of discrimination they aim to end. In an individual claim, the targeted

behavior involves instances of discrimination against a particular plaintiff. 
Reynolds, 685 F.3d at

203
-04. Thus, statistical evidence is not necessary or, in fact, helpful in establishing these

particular occurrences. Id.; see also Hudson v. Int’l Bus. Mach. Corp., 
620 F.2d 351
, 355 (2d

Cir. 1980). In general, the plaintiff’s prima facie burden under McDonnell Douglas has been

characterized as “not onerous.” Tex. Dep’t of Cmty. Affairs v. Burdine, 
450 U.S. 248
, 253


       3
           The majority claims this sentence incorrectly substitutes the word “statistics” for
“proof.” Maj. Op. at 25-26 n.13. However, I agree that Teamsters requires the defendant meet
the Government’s “proof” and have stated this point above. Rather than incorrectly replace
“proof” with the word “statistics” this sentence only reveals the further logical deduction that the
defendant must meet the plaintiff’s statistical proof. It explains this point in the context of the
paragraph’s syllogism. The syllogism proceeds from Teamsters as follows: 1) Under Teamsters
a defendant has the “rebuttal burden of showing that the plaintiff’s prima facie proof lacked
significance,” a point with which the majority agrees, Maj. Op. at 25-26 n.13; 2) In pattern-or-
practice cases the plaintiff’s proof is often composed of statistics, as was the case here. Ergo, 3)
in such cases the defendant’s obligation is to meet that proof, by defeating the meaning of those
statistics. Thus the use of “statistics” rather than “proof” in this context is a far cry from a
conflation and instead is the result of a logical conclusion.
       4
          In addition to these two main differences, to remedy the particular harm at which they
are aimed, pattern-or-practice claims authorize a specific form of injunctive relief against an
employer “to the full enjoyment of any of the rights secured by the statute.” Chin v. Port Auth.
of N.Y. & N.J., 
685 F.3d 135
, 146-48 (2d Cir. 2012) (citing 42 U.S.C. § 2000e-6) (internal
quotation marks omitted). If the district court can fashion class-wide prospective relief,
including “an injunctive order against continuation of the discriminatory practice, an order that
the employer keep records of its future employment decisions and file periodic reports with the
court, or any other order necessary to ensure the full enjoyment of the rights protected by Title
VII.” 
Teamsters, 431 U.S. at 361 
(internal quotation marks omitted). Thus, pattern-or-practice
plaintiffs seek different relief, unlike plaintiffs seeking individualized relief who assert “back
pay, front pay, or compensatory recovery.” 
Reynolds, 685 F.3d at 203 
(internal citation and
quotation marks omitted).

                                                  4
(1981). Similarly, a McDonnell Douglas defendant’s burden is less demanding—being able to

merely present a legitimate neutral reason—where there are no statistics to controvert. For

example, it is sufficient if the defendant’s evidence simply sets forth “the reasons for the

plaintiff’s rejection.” 
Id. at 255.

        In 
contrast, the “essence” of pattern-or-practice claims is different because they aim to

end discrimination of a different kind. These claims address cases where “the employer treated

all members of the class in the same way,” so “the manner of proving [and defeating] the claim

usually relies heavily on statistical evidence, bolstered as necessary with anecdotal evidence.” 1

Arthur Larson et al., Employment Discrimination § 8.02, at 8-15 (2d ed. 2012) (hereinafter

“Larson”). As the Supreme Court underscored in Teamsters, statistical evidence is often

necessary in proving pattern-or-practice cases because plaintiffs must prove more than mere

“sporadic acts of discrimination; rather, they must establish that intentional discrimination was

the defendant’s ‘standard operating procedure.’” 
Robinson, 267 F.3d at 158 
(quoting 
Teamsters,

431 U.S. at 336
). Once the plaintiff has used statistical evidence to make out its prima facie

case, the employer must then “defeat the prima facie showing” by addressing those statistics.

Teamsters, 431 U.S. at 360
. Although there are no “limits on the type of evidence an employer

may use,” because the employer must “meet” the plaintiff’s case and demonstrate that it is

inaccurate or insignificant, 
id. at 360 n.46, 
if statistics were used, those statistics must

necessarily be addressed.

        The Teamsters burden on the employer is carefully calibrated to identify widespread

discriminatory acts; thus, an employer’s noncompliance with this test is an attempt to evade its

purposes. The burden shifting framework “‘is a procedural device, designed . . . to establish an


                                                   5
order and proof of production,’ and, like other procedural rules, [it] subjects noncompliant

parties to default regardless of the objective merit of their claims or defenses.” Dist. Ct. Op. at

36 (quoting St. Mary’s Honor Ctr. v. Hicks, 
509 U.S. 502
, 522 (1993) (emphasis in original)).

Similarly, as it fails to address the pattern-or-practice concerns, using the “ill-suited” McDonnell

Douglas pattern of proof is likewise inappropriate. 
Robinson, 267 F.3d at 157 
n.3 (quoting 1

Arthur Larson et al., Employment Discrimination § 8.01[4], at 8-13 (2d ed. 2001)). As one Court

put it: “In a complex class action, utilizing statistical proof and counterproof, the value of the

[McDonnell Douglas] sequence—to highlight the issues in contrast—is about as relevant as a

minuet is to a thermonuclear battle.” Vuyanich v. Republic Nat’l Bank, 
521 F. Supp. 656
, 661

(N.D. Tex. 1981), vacated on other grounds, 
723 F.2d 1195 
(5th Cir. 1984). Thus, an employer

who attempts to displace the Teamsters requirement with McDonnell Douglas proof must fail.

        Courts have continuously acknowledged these differences between Teamsters and

McDonnell Douglas. For example, the year following its opinion in Teamsters, the Supreme

Court reiterated the distinction between McDonnell Douglas and Teamsters in Title VII cases.

Waters, 438 U.S. at 575
. In Waters, a suit brought by three individuals the Court found that it

was necessary to apply the “approach . . . contained in McDonnell Douglas” because it “was not

a ‘pattern or practice’ case like Teamsters v. United States . . . .” 
Id. at 575 & 
n.7; see also

Cooper v. Fed. Reserve Bank of Richmond, 
467 U.S. 867
, 876 (1984) (distinguishing Teamsters

and McDonnell Douglas in a case contemplating additional proceedings to determine the scope

of individual relief).




                                                  6
       This Circuit has authoritatively held that Teamsters establishes the approach for pattern-

or-practice cases, distinct from that of McDonnell Douglas.5 In Robinson, this Circuit

enunciated that “Title VII disparate treatment claims are of two types,” either individual claims

to be analyzed under McDonnell Douglas or pattern-or-practice claims brought under Teamsters.

Reynolds, 685 F.3d at 202 
(citing 
Robinson, 267 F.3d at 157 
n.3). We have since reiterated

Robinson’s holding that Teamsters established the distinct test for pattern-or-practice claims in

subsequent cases. See 
Reynolds, 685 F.3d at 202
; Chin v. Port Auth. of N.Y. & N.J., 
685 F.3d

135
, 146-50 (2d Cir. 2012). Lower courts in this Circuit have also diligently adopted our

approach, see, e.g., Lomotey v. Conn., Dep’t of Transp., 
2012 WL 642763
, at *12 (D. Conn. Feb.

28, 2012) (stating that Teamsters is distinguishable from McDonnell Douglas framework);

EEOC v. Bloomberg L.P., 
778 F. Supp. 2d 458
, 468 (S.D.N.Y. 2011) (same); United States v.

N.Y.C. Transit Auth., 
2010 WL 3855191
, at *23 (E.D.N.Y. Sept. 28, 2010) (applying Teamsters

to this pattern-or-practice case and stating that “the McDonnell Douglas framework does not

apply in Section 707 cases”); United States v. City of New York, 
631 F. Supp. 2d 419
, 427

(S.D.N.Y. 2009) (collecting cases for the proposition that individuals cannot bring pattern-or-

practice cases), and at least one other Circuit has recognized Robinson’s change to our law, see,

e.g., Hohider v. United Parcel Serv., Inc., 
574 F.3d 169
, 179 n.11 (3rd Cir. 2009) (noting that

more recently “courts of appeals have used the Teamsters two-stage framework to analyze


       5
         Although our earlier cases immediately following Teamsters applied both the
McDonnell Douglas and Teamsters tests to pattern-or-practice cases without distinguishing
them, Ottaviani v. State Univ. of N.Y. at New Paltz, 
875 F.2d 365
, 370 (2d Cir. 1989); Woodbury
v. N.Y.C. Transit Auth., 
832 F.2d 764
, 768-69 (2d Cir. 1987); Ste. Marie v. E. R.R. Ass’n., 
650
F.2d 395
, 397 (2d Cir. 1981), this Court subsequently allayed any concerns over confusion
between the two by definitively holding that the Teamsters framework sets distinct standards for
pattern-or-practice claims. 
Robinson, 267 F.3d at 159
-60.

                                                 7
pattern-or-practice claims brought as private-plaintiff class actions under Title VII” and citing to

Robinson, 267 F.3d at 158
-60).

        In addition to recognizing these general differences, this Court went even further in

Reynolds v. Barrett and underscored that Robinson also adopted one of the most predominant

differences between the tests: unlike the McDonnell Douglas burden-shifting framework, the

employer under Teamsters must “show that the statistical evidence proffered by the plaintiffs is

insignificant or inaccurate.” 
Reynolds, 685 F.3d at 203
. “Typically, this is accomplished by

challenging the source, accuracy, or probative force of the plaintiffs’ statistics.” 
Id. (internal

quotation marks 
and citation omitted). Thus, the “prudent defendant will follow all three routes

if possible, presenting its own version of the numbers game, attempting to undermine the

plaintiffs’ version with specific attacks on the validity of the plaintiffs’ statistics, and garnering

non-statistical evidentiary support as well.” 
Robinson, 267 F.3d at 159 
(internal citation and

alterations omitted).

        Robinson thus emphasized the two differing aspects of the employer’s burden under

Teamsters that are at issue today. First, the requirement that the employer show “insignificance

or inaccuracy” is a much more difficult bar than merely proffering some non-pretextual reason

for discrimination as required by McDonnell Douglas. Second, also unlike McDonnell Douglas,

this more difficult task under Teamsters must often be achieved by responding to the plaintiff’s

statistical proof. In essence, there is a statistical bind inherent in pattern-or-practice cases.

Because the plaintiff’s proof will often be composed of statistics, and the employer must

“demonstrat[e] that the [prima facie] proof is either inaccurate or insignificant,” Teamsters, 
431




                                                   8
U.S. at 360
, the employer is often required to defeat the plaintiff’s statistical evidence.6

       Consistent with our interpretation in Robinson and Reynolds, our sister Circuits have

similarly upheld Teamsters’ distinct analysis. Indeed, every Circuit, save the First and Federal

Circuits, has decided cases specifically quoting that the employer’s burden under Teamsters is to

demonstrate that the plaintiff’s offered proof is “inaccurate or insignificant.” See 
Hohider, 574

F.3d at 177-78
; Morgan v. United Parcel Serv. of Am., Inc., 
380 F.3d 459
, 463-64 (8th Cir.

2004); Beck v. Boeing Co., 60 F. App’x 38, 39 (9th Cir. 2003); Thiessen v. Gen. Elec. Capital

Corp., 
267 F.3d 1095
, 1106 (10th Cir. 2001); EEOC v. Joe’s Stone Crab, Inc., 
220 F.3d 1263
,

1287 n.22 (11th Cir. 2000); Anderson v. Douglas & Lomason Co., 
26 F.3d 1277
, 1285 (5th Cir.

1994); EEOC v. Sears, Roebuck & Co., 
839 F.2d 302
, 308 (7th Cir. 1988); Ardrey v. United

Parcel Serv., 
798 F.2d 679
, 683 (4th Cir. 1986); Segar v. Smith, 
738 F.2d 1249
, 1300-01 (D.C.

Cir. 1984); Alexander v. Aero Lodge No. 735, Int’l Ass’n of Machinists and Aerospace Workers,

565 F.2d 1364
, 1382 n.11 (6th Cir. 1977), superceded by statute on other grounds.

II. The District Court Correctly Applied Teamsters

       Thus, adhering to this unanimous recognition of Teamsters by this Circuit and others, the

district court in this case faithfully applied the distinct pattern-or-practice test and found the City

did not meet its more difficult burden. To begin its discussion, the district court announced that


       6
         This Court’s prescriptions in Reynolds and Robinson acknowledge the statistical bind
that most employers will face. In these cases, the Court stated that challenging the statistical
proof would be “[t]ypical[]” or the “basic” form of rebuttal. 
Reynolds, 685 F.3d at 203
;
Robinson, 267 F.3d at 159 
(internal citation omitted). Thus, Reynolds and Robinson recognize
that Teamsters allows for non-statistical evidence, but these cases also recognize that the
employer’s proof will rarely, if ever, be exclusively non-statistical. They notice that in most
pattern-or-practice cases the plaintiff’s proof will need to consist of statistics to show widespread
discrimination, 
Teamsters, 431 U.S. at 339-40 
n.20, and therefore the employer will have to
answer that evidence with more than just non-statistical proof.

                                                   9
the Intervenors clearly met their burden by presenting “sufficient undisputed statistical evidence

to support a prima facie case that the City had a pattern or practice of discriminating against

black applicants.” Dist. Ct. Op 28-29.7 Not only did the statistics bear out that the City’s use of

its two Exams on a pass/fail basis and the rank order processing yielded “statistically significant

adverse effects on black candidates,” 
id. at 29, but 
the Intervenors also “supplemented their

statistical showing with extensive historical, anecdotal, and testimonial evidence that the

intentional discrimination was the City’s standard operating procedure” for several decades, 
id.

at 31 (internal 
quotation marks omitted).

        Finding the prima facie burden had been met, the district court then moved on to the

Teamsters requirement enunciated in Robinson that “the City bears the burden of ‘demonstrating

that the [Intervenors’] proof is either inaccurate or insignificant’ by attacking its ‘source,

accuracy, or probative force.’” 
Id. at 33 (citing 
Robinson, 267 F.3d at 159
). Having failed to

“dispute either the accuracy or practical significance of Plaintiffs’ statistical analyses,” 
id. at 7,

the 
district court found the City to have “abjured [its Teamsters] responsibility entirely,” 
id. at

33. The 
court stated, “[t]he City has not offered a competing ‘statistical summary treatment of

the protected class,’ has not attempted to undermine the Intervenors’ statistics with ‘specific

attacks on their validity,’ and has garnered no ‘anecdotal or other non-statistical evidence

tending to rebut the inference of discrimination.’” 
Id. (internal alterations and 
citations omitted).


        7
         For example, the court found plaintiffs proved their burden by showing that (1) the
application of the pass/fail policy to Exam 7029 and Exam 2403 had a net effect of eliminating
“between 607 and 684 black applicants who would not have failed the Exams but for the
disparity” and also kept 144 black firefighters from gaining appointments they would have
otherwise received, and (2) the rank-ordering of the examinations had the net effect of denying
112 black applicants “approximately 34 years’ worth of wages and seniority that they would
have received absent the policy’s disparate effects.” Dist. Ct. Op. at 29-30.

                                                  10
Thus, despite Robinson’s suggestion that “the prudent defendant will follow all three 
routes,”

267 F.3d at 159 
(internal citation omitted), here, the lower court correctly found that the City

attempted to “circumvent its burden of production entirely.” Dist. Ct. Op. at 33.

       Rather than responding to the statistical evidence, the City only “argu[ed] that the

Intervenors ha[d] not proved that the City harbored a subjective intent to discriminate against

black applicants.” 
Id. In essence, the 
City ignored the inevitable conclusion of the statistics and

tried to focus on intent. But, “[a]t this stage, lack of direct proof regarding the employer’s

mental state is simply immaterial to the question of whether the City can rebut the presumption

of unlawful discrimination created by the Intervenors’ prima facie showing.” 
Id. at 37. Despite
the City’s correct assertion that what “actually motivate[s] the employer’s decision” is relevant,

Hazen Paper Co. v. Biggins, 
507 U.S. 604
, 610 (1993), departure from the Teamsters framework

is “fatal” where the motivation did not address the statistical evidence. Dist. Ct. Op. at 35.

       In the face of the district court’s faithful application of the law, the majority not only

overrules that well-reasoned decision but also confuses the distinct pattern-or-practice standard

by incorrectly conflating the employer’s burden in Teamsters with the one set out in McDonnell

Douglas. See Maj. Op. at 23-25. While the majority asserts that the City need only provide “a

nondiscriminatory explanation” under Teamsters, 
id. at 22, 25, 
30, I think the City was required

both under the law of this Circuit and the Supreme Court to show more. At the outset, the

majority recognizes that “[i]n a pattern-or-practice case, the plaintiff’s initial burden is heavier,”

id. at 21, however, 
it fails to recognize that a similarly heavy burden also exists for the

defendant-employer under Teamsters.




                                                  11
III. The Majority Failed to Notice Teamsters’ “Insignificant or Inaccurate” Burden is

Higher than that in McDonnell Douglas.

       First, the majority incorrectly tries to pare down the employer’s higher bar in Teamsters

to the one set out in McDonnell Douglas. Under Teamsters, the employer may only satisfy its

heavy burden if its defense “meet[s] the prima facie case” and proves it is “insignificant or

inaccurate.” 
Teamsters, 431 F.3d at 360 
n.46. The majority tries to evade Teamsters’ more

difficult “insignificant or inaccurate” burden by characterizing the language in Reynolds as

dictum. Maj. Op. at 27 n.14. I find this characterization confounding. There is no reasonable

reading of Reynolds, which can classify this language as dicta, especially because, as the

majority admits, the Supreme Court in Teamsters clearly stated that the employer has the burden

to “defeat” the prima facie case. 
Id. at 22-23. The 
employer’s burden to defeat the prima facie

case as insignificant or inaccurate can certainly not be said to be the same as McDonnell

Douglas’ requirement that the employer simply rebut the plaintiff’s evidence with the offer of a

nondiscriminatory reason. Still, the majority tries to lessen Teamsters’ more stringent language

by stating that although “‘[d]efeat’ might be thought to imply something stronger than

[McDonnell Douglas’ requirement to] ‘rebut,’” the Court’s language “means the same thing in

both contexts.” 
Id. at 23. This 
argument clearly disregards the plain language of not only

Teamsters, but also the law of this Circuit that is binding on subsequent panels. Not only were

the separate meanings of “defeat” and “rebut” intended by the Supreme Court’s separate use of

these two distinct words, but the separate meanings are underscored in the different standards

requiring a demonstration that the plaintiff’s proof is “inaccurate or insignificant” versus




                                                 12
showing some neutral legitimate reason.8

       Additionally, the majority tries to convince us a lower bar can be met by quoting

Teamsters’ recitation of the McDonnell Douglas’ standard. Insofar as stating that Teamsters

may recognize the McDonell Douglas standard is relevant, the majority is in fact correct.

Teamsters does state that the employer may “provid[e] a nondiscriminatory explanation for the

apparently discriminatory result.” 
Id. at 23. However, 
Teamsters does not state this will fully

satisfy the employer’s duty to meet the “inaccurate or insignificant” standard. As explained by

the D.C. Circuit,

       [t]he bare articulation of a nondiscriminatory explanation, while sufficient to rebut an
       individual plaintiff’s low-threshold McDonnell Douglas showing, generally will not
       suffice as a rebuttal to a typical class-wide showing of pervasive discrimination. . . . in
       both individual and class action contexts the defendant faces the same rebuttal burden; it
       must present sufficient evidence to permit the trier of fact to decline to draw the inference
       of discrimination from the plaintiffs’ proof. But in the class action pattern or practice
       case the strength of the evidence sufficient to meet this rebuttal burden will typically
       need to be much higher than the strength of the evidence sufficient to rebut an individual
       plaintiff’s low-threshold McDonnell Douglas showing.

Segar, 738 F.2d at 1269-70 
(discussing applicability of Teamsters as opposed to McDonnell

Douglas to pattern-or-practice claims). In other words, the majority ignores the fact that in cases

with substantial statistical proof in the prima facie case, a McDonnell Douglas offering of a

nondiscriminatory reason will not satisfy the employer’s burden under the higher Teamsters bar.

See Hazelwood School Dist. v. United States, 
433 U.S. 299
, 308-12 (1977) (detailing how



       8
          Beyond disregarding the clear distinction in the separate instances of these words and
suggesting that the Supreme Court doesn’t intend to use the word “defeat” when it uses it, the
majority seems to misconstrue the law even when the plain meaning of these words reveals
differences. The Oxford English Dictionary defines “defeat” as “To unmake, undo, do away
with; to ruin, destroy” in comparison to “rebut” as “to provide a counter-argument to.” Oxford
English Dictionary (online ed. March 2012).

                                                13
alternate statistics were necessary to rebut statistical prima facie showing). Instead, the

McDonnell Douglas standard is relevant only to the extent that Teamsters encompasses

McDonnell Douglas, but Teamsters’ burden still requires more beyond it.

IV. The Majority Failed to Recognize Teamsters’ Method of Proof is More Difficult than

McDonnell Douglas’

       In addition to paring down the employer’s burden under Teamsters, the majority also

inaccurately describes the method of proof required. The majority is correct that Teamsters

states the employer’s proof need not be limited to statistics alone. However, as previously

explained, because pattern-or-practice suits usually focus on widespread discrimination, statistics

will often be necessary to make such a case.9 Where a plaintiff presents such evidence as part of

its prima facie case, Teamsters requires the employer to counter with a different method of proof

than that which would be sufficient under McDonnell Douglas; it requires the employer to

address the cold hard numbers. Larson, § 9.03[2][a], at 9-18 to -19.

       The majority disagrees. It admits the passage from Larson stating, the employer has

three avenues of attack to challenge “the plaintiff’s statistics,” quoted by this Court in Robinson,

“might be thought to require an employer to challenge the plaintiff’s statistics,” but the majority

abandons that interpretation. Maj. Op. at 26. Instead, the majority cites to the portion of

Larson’s treatise, which states that a defendant may use “other non-statistical evidence tending


       9
          “Statistics showing racial or ethnic imbalance are probative in [disparate treatment
cases],” 
Teamsters, 431 U.S. at 340 
n.20, and although anecdotal evidence may be useful to
bring “the cold numbers convincingly to life,” 
id. at 339, statistical 
evidence is often necessary
and often sufficient to establish a prima face case, id. See also 
Reynolds, 685 F.3d at 203
(stating “statistics alone can make out a prima facie case of discrimination in a pattern-or-
practice suit if the statistics reveal a gross disparity in the treatment of workers based on race”)
(internal citation, quotation marks and alterations omitted).

                                                 14
to rebut the inference of discrimination.” 
Id. As stated above, 
this quotation does not present a

contradiction; while use of non-statistical evidence is relevant, it is often not sufficient if it does

not meet the prima facie evidence. Although Larson’s treatise goes on to state, in a section

called “Other Rebuttal Techniques,” that one Seventh Circuit case found exclusive non-

statistical evidence of “an employer’s affirmative action efforts” may be relevant, Larson, §

9.03[2][c], at 9-20.1, this case is atypical. Larson’s presentation of this case as an exception

suggests that the defendant’s rebuttal burden will often not be sufficient unless it defeats the

plaintiff’s statistical evidence or at the very minimum attacks those statistics. See Larson, §

9.03[2][a], [b].

        As Larson states, defendant’s burden is to “come forward with evidence sufficient to

rebut the prima face case based on statistics.” Larson, § 9.03[2], at 9-16. This means the

employer’s rebuttal will often be composed of targeting “flaws in the plaintiff’s [statistical]

evidence.” Larson, § 9.03[2][a], at 9-17. In fact, Larson goes even further and states that the

Supreme Court’s decision in Bazemore v. Friday, 
478 U.S. 385 
(1986), has “been interpreted to

imply that the defendant must do more than merely point out possible flaws in the plaintiff’s

statistical analysis, the defendant must also show that consideration of the missing factors would

make a difference in the analytical outcome.” Larson, § 9.03[2][b], at 9-19. Thus the defendant

is often required to “present[] its own statistics to the court” which “a defendant will wish to

argue . . . are more ‘finely tuned’ and have more probative force because they take into

consideration a variety of relevant factors that the plaintiff’s statistics ignore.” Larson, §

9.03[2][b], at 9-19.




                                                  15
       In addition to arguably misinterpreting Professor Larson’s treatise, the majority

disregards this Circuit’s actual case law by claiming Reynolds’ language on the method of proof,

recognized in Robinson, is “dictum.” Maj. Op. at 27 n.14. Moreover, the majority looks over

the logical conclusion in Robinson and Reynolds that statistics are not always but often required

because “[s]tatistics proffered during the ‘liability phase’ of a pattern-or-practice suit purport to

demonstrate that a pattern of discrimination exists at an entity.” 
Reynolds, 685 F.3d at 204
(emphasis in original). As the Court stated, “In a Title VII case, these statistics can make out a

prima facie case that the employer was engaged in a pattern or practice of discrimination. This is

because [statistics or] an analysis of the collective acts of those who do the employer’s bidding

bespeak the employer’s motivation.” 
Id.

       This Court’s prescriptions 
in Robinson and Reynolds acknowledge the most “basic” or

common situation that the rebuttal will be statistical—not because they have misread Teamsters

as stating the evidence must be exclusively statistical but because statistics will most often be

necessary for the plaintiff to show the entity engaged in a pattern-or-practice. 
Teamsters, 431

U.S. at 339-40 
n.20. Thus, as stated by Larson in his treatise, non-statistical proof may be

relevant to disproving those statistics but it most often will not be sufficient. See Larson, §

9.03[2][a], [b], at 9-16 to -20. Regardless, whatever the plaintiff’s proof is—statistical or

otherwise—if the defendant does not show that proof to be inaccurate or insignificant it has not

carried its burden of proof and summary judgment must be granted in favor of the plaintiffs.

       In undermining our caselaw on the employer’s method of proof, the majority also ignores

the law of our sister Circuits that have held the same rule. It is well recognized among other

Circuits that the employer must defeat the Government’s statistical proof. See Hohider, 
574 F.3d


                                                  16
at 183 (“The Teamsters framework was judicially promulgated as a method of proof for

pattern-or-practice claims brought by the government under Title VII, as that statute

authorizes—it provides a means by which courts can assess whether a particular form of

statutorily prohibited discrimination exists, just as the McDonnell Douglas framework does for

individual claims of disparate treatment.”); Celestine v. Petroleos de Venezuella SA, 
266 F.3d

343
, 356 (5th Cir. 2001) (indicating the “Teamsters method of proof [is] an independent method

of proof”); Davoll v. Webb, 
194 F.3d 1116
, 1148 (10th Cir. 1999) (noting the Supreme Court has

“recognized that the specifics of the McDonnell Douglas framework are inapplicable in certain

factual situations, including when the government has brought a broad-based pattern and practice

action” and citing to 
Teamsters, 431 U.S. at 358-60
); 
Segar, 738 F.2d at 1267-69 
(discussing

applicability of Teamsters’ method of proof as opposed to McDonnell Douglas to pattern-or-

practice claims).

       The Intervenors’ proof in this case, as in most pattern-or practice cases, was in fact

statistical. Thus, the City’s method of proof was required to be directed toward those statistics.

The majority’s classification of these differences in method proof as dictum and its choice not to

apply Teamsters’ and thus be satisfied by the City’s facially neutral evidence, give no response

to Intervenors’ statistics that show a clear pattern and practice of discrimination. Whatever the

proof, under Teamsters, the city was obligated to defeat it. Any other argument is contrary to the

law of this Circuit and the Supreme Court.

V. Applying Teamsters’ the City Failed to Meet Its Burden

       In this case, there is no dispute that the Intervenors’ claim alleges that the New York City

Fire Department engaged in a pattern-or-practice for decades that led to black and other minority


                                                17
firefighters being severely underrepresented in the Department. See Dist. Ct. Op. at 2; see also

Maj. Op. at 6-8. As the district court stated, at the time this claim was filed, only 3.4% of the

Department’s force was composed of black firefighters despite black residents making up 25.6%

of New York City’s population. Dist. Ct. Op. at 2. In other words, in a city of over eight

million people, and out of a force with 8,998 firefighters, the Department only employed 303

black firefighters. 
Id. Perhaps most egregious 
is that in an otherwise esteemed Department, this

“one persistent stain” has remained an essentially unchanged practice since at least the 1960s.10

Id.

       In accord with 
this pattern-or-practice case, the Intervenors presented copious statistical

evidence. 
Id. at 6, 28. 
The district court found that the Intervenors met their burden,11 
id. at 28,

but 
regardless, we need not reach that issue here, where, in fact, all parties agree the Intervenors


       10
        In fact, as the district court found, the FDNY’s rate of black firefighters has actually
gone down:

       At the time of the [original] Vulcan Society litigation, blacks and Hispanics constituted
       32% of the City’s population, but only 5% of the Department. In 1990, almost two
       decades later, blacks made up 29% of the City’s population, but only 4% of firefighters.
       In 2002, 25% of the City’s residents were black, compared to only 2.6% of its
       firefighters. Between 1991 and 2007, black firefighters never constituted more than 3.9%
       of the force, and by the time this case was filed in 2007, the percentage of black
       firefighters in the FDNY had dropped to 3.4%.

Dist. Ct. Op. at 16 (internal citations omitted).
       11
            As previously stated, the Intervenors met their burden by showing that black candidates
disproportionately failed Written Exams 7029 and 2043 because of the pass/fail and rank
ordering policies. Dist. Ct. Op. at 29-30. Additionally, the district court noted that this was not
the first time the City had been brought to federal court for its discriminatory use of firefighter
examinations. 
Id. at 11 (citing 
Vulcan Soc’y of N.Y.C. Fire Dep’t, Inc. v. Civil Serv. Comm’n,
360 F. Supp. 1265
, 1269 (S.D.N.Y. 1973), affirmed in relevant part by 
490 F.2d 387 
(2d Cir.
1973)). This showing was more than adequate to make out a prima facie case. Dist. Ct. Op. at 30
(citing Waisome v. Port Auth. of N.Y. & N.J., 
948 F.2d 1370
, 1376 (2d Cir. 1991)).

                                                    18
met their prima facie case under Teamsters. Maj. Op. at 30. The only question that remains is

whether the City defeated the statistical evidence proffered by the plaintiffs as insignificant or

inaccurate. See 
Teamsters, 431 U.S. at 360
.

       To carry its burden, the City proffered what the majority characterized as “a

nondiscriminatory reason” for using the test. Maj. Op. at 30-31. The City’s response was that

the test designers “did not intend to discriminate against any protected group,” Dist. Ct. Dkt. 359

at 5, and that there is no evidence that the Mayor and the Fire Commissioner intended to

discriminate, 
id. at 6-7. The 
City also stated that the FDNY made recruitment efforts to increase

minority hiring. 
Id. at 9; see 
also Maj. Op. at 31.12 In essence, the City stated that it had

prepared the tests with “acceptable test development methods.” 
Id.

       Applying Teamsters’ more 
difficult burden to the City’s offering, the district court

deemed it deficient because the City’s evidence did not sufficiently attack the Intervenors’

statistics. Maj. Op. at 33. Thus, the lower court found that even if the City offered evidence

sufficient for a McDonnell Douglas method of proof, it didn’t defeat the Intervenors’ evidence



       12
           The City points us to more detailed evidence about its recruiting, Appellants’ Br. at 13-
14, as well as an attrition plan meant to “reduce minority candidate dropouts during the
application process. It also points to evidence in the record that “that both exams were
constructed in accordance with “standard job analytic and test development procedures.” 
Id. at
19. The 
City also points us again to the increased EMT promotional exams. 
Id. at 20.
Additionally, 
the City “also enlisted Columbia University’s School of International and Public
Affairs to analyze and recommend improvements to its diversity initiative (“the Columbia
Study”). The Columbia Study arose in conjunction with the FDNY’s “Strategic Plan” to rebuild
and strengthen after 9/11, which identified enhancing diversity as one of its top six priorities.”
Id. Finally, the City 
points us to proof that it created an FDNY High School for Fire and Life,
which is a “99% minority high school [that] incorporates firefighting and EMT studies with a
traditional educational curriculum.” 
Id. As Intervenors note, 
none of this evidence was called to
the attention of the district court. Intervenors’ Br.at 118-20. Even considering these facts, the
City’s argument still fails.

                                                 19
under Teamsters. The district court therefore granted summary judgment to the Intervenors on

their pattern-or-practice case.

       On appeal, the City tried to evade Teamsters’ heavy burden by arguing that its offer of

proof was sufficient according to Raytheon Co. v. Hernandez, 
540 U.S. 44 
(2003). In Raytheon,

the Supreme Court held that an employer’s “proffer of [a] neutral . . . policy plainly satisfied its

obligation under McDonnell Douglas to provide a legitimate, nondiscriminatory reason for the

challenged action.” 
Id. at 53. However, 
Raytheon, unlike the present case, involved an

individual’s non-class claim. Hence the City, in this pattern-or-practice case, tried to evade the

demanding burden by claiming Teamsters is but a mere gloss for McDonnell Douglas. Having

established that Teamsters is distinct and applies a more difficult burden for employers, I

conclude the City was required to defeat the Intervernors’ proof as inaccurate or insignificant.

Here it did not.

       Moreover, where the City merely offered non-statistical proof, in the face of

overwhelming statistics proving its discriminatory intent, the City failed to “meet” the prima

facie case. Here the City rebutted with affidavits, non-statistical proof, that its employees

created the Exams without a discriminatory motive. Although intent is crucial in Title VII cases,

here where statistics are proof of the entity’s discriminatory intent12 the employer’s burden is to

defeat those statistics. The employer must show that the statistics, which show intent, are

incorrect. The employer may not merely counter that it had no ill-motives.


       12
          See Larson § 9.03[1], at 9-13 (stating that “[t]he ultimate burden in any disparate
treatment case is to convince the fact finder that the defendant’s actions were discriminatorily
motivated” and that in a “pattern or practice suit” the plaintiff may depend on two forms of
circumstantial evidence: 1) statistical proof “aimed at establishing the defendant’s past treatment
of the protected group” and 2) “testimony from protected class members”).

                                                 20
       Moreover, it may not disprove the entity’s intent by merely stating the employees making

the exam did not have a discriminatory motive. Instead, the statistics must be challenged. Even

if non-statistical evidence is probative, the City did not present sufficient non-statistical

evidence. Offering declarations that the designers of the Exams did not intend to create a

discriminatory test in no sense goes to whether use of that test was discriminatory, as

Intervernors proved. Even if the tests were not designed to discriminate against minorities, the

use of them undisputedly did through the pass/fail and rank-ordering policies. Still, the majority

tries to argue that the City’s method of proof was sufficient because it challenged “the inference

of intentional discrimination arising from the Intervenors’ prima facie case.” Maj. Op. at 33.

       As already stated, although intent is key in Title VII cases when dealing with pattern-or-

practice cases, mere proof of nondiscriminatory motivations in the creation of a test are not the

end of the intent analysis. As this Court has previously stated, “[p]ersistent use of exams with

disparate racial effects would support an inference of intentional discrimination if proper test

construction were not even attempted.” Guardians Ass’n of the N.Y.C. Police Dep’t, Inc. v. Civil

Serv. Comm’n, 
630 F.2d 79 
(2d Cir. 1980). Thus, if wishing to rely merely on non-statistical

proof, the City was still obligated to raise some facts explaining why its knowing continued use

of those tests was not evidence of discrimination. As the district court stated, the trouble is not

the existence of the Exams per se but the City’s policy or practice of subsequently using those

Exams as pass/fail and rank-ordering devices. Dist. Ct. Op. at 38.

       For this reason, I fully agree with the district court’s conclusion that “the subjective

motives of the people who designed the Exams are only circumstantially relevant to the question

of whether the City’s decision to use the Exams as screening and ranking devices was


                                                  21
discriminatory.” 
Id. As hypothetically stated 
by the district court, a showing that the Exams

were applied and reapplied with a nondiscriminatory intent would be “highly relevant to the

City’s defense because it would support an inference that the City’s actual intent in enforcing

[and reenforcing] the pass/fail and rank-ordering policies was to select the best candidates.” 
Id.

at 38-39. But 
the mere initial motivation of the test designers doesn’t reach the issue of whether

the City’s ex post use of those tests was discriminatory.

       Additionally, although intent may be relevant to the question of imposing injunctive

relief under a disparate impact theory, see Berkman v. City of New York, 
705 F.2d 584
, 595 (2d

Cir. 1983) (stating affirmative relief may be required where, for example, “the defendant has

intentionally or egregiously engaged in a practice of discrimination”), that issue is not relevant at

this initial stage under disparate treatment.13 Where persistent use of the tests bear out evidence

of discriminatory intent and without more proof addressing the Intervenors’ statistics, the City

cannot be said to have defeated the prima facie case.

       Similarly, I find the fact that the City engaged in minority recruitment does not help to

defeat the Intervenors’ case because all the potential black and Hispanic firefighters that were

recruited would still have been subjected to the challenged procedures. As the district court

stated, once the Intervenors made out the prima facie case that the City’s Exams discriminated

against black applicants, proof of recruitment only proved “more blacks were taking the exam”

and thus “more blacks were being illegally harmed, and the City’s evidence is relevant only to


       13
          Moreover, our Circuit has already determined in Ass’n Against Discrimination in
Emp’t, Inc. v. City of Bridgeport, 
647 F.2d 256
, 280 n.22 (2d Cir. 1981) that “the requirement
that an employer have discriminated ‘intentionally’ in order for the provisions of 706(g) to come
into play means not that there must have been a discriminatory purpose, but only that the acts
must have been deliberate, not accidental.”

                                                 22
the scope of the injury, not its source.” Dist. Ct. Op. at 39. The majority cites to Washington v.

Davis, 
426 U.S. 229
, 246 (1976), to argue affirmative efforts to recruit black officers negates the

inference of intent, Maj. Op. at 34-35, but recruitment efforts simply do not satisfy the

defendant’s burden to “meet” the plaintiff’s case under Teamsters.

       Instead, what matters under Teamsters is “(1) whether the City has policies of screening

and ranking applicants based on how well they perform the required task, (2) what effect those

policies have on black applicants, and (3) why the City decided to adopt those policies.” Dist.

Ct. Op. at 38. Because the evidence of recruitment has no bearing on the challenged use of the

Exams, because the City does not provide a sufficient nondiscriminatory explanation where

statistics showed a discriminatory result, and because any recruits would ultimately face the

discriminatory Exams, this evidence cannot serve to show that Intervernors’ evidence was

inaccurate or insignificant.

       In this context, the City completely failed to meet Intervenors’ proof, as required by

Teamsters, and it did not even approach raising a material question of fact about whether the

Intervenors’ proof was inaccurate or insignificant on appeal. Thus, the City’s apparent

disinterest in fixing what it has known to be a problem for more than four decades and its

apparent disinterest in defending itself now according to the Teamsters standard is precisely the

type of disregard the Teamsters test was meant to protect against. For the foregoing reasons, I

am unable to agree with the majority’s conclusion that the City met its burden because the City

may not meet its requirements by supplanting Teamsters with the McDonnell Douglas standard,

and it is unreasonable to understand the language in Robinson and Reynolds as mere dicta.

Accordingly, the Intervenors were entitled to summary judgment on their pattern-or-practice

claim and I DISSENT from that part of the majority’s opinion that holds otherwise.


                                                23