Central Wholesalers, Inc., EEOC v., 4C Opening Brief 04-08 dtv.doc

TABLE OF AUTHORITIES

Cases

Anderson v. Boeing, 222 F.R.D. 521 (N.D. Okla. 2004)............................ 67

Anderson v. Group Hospitalizations, 820 F.2d 465 (D.C. Cir. 1987)...... 50

Anderson v. Westinghouse Savannah River, 406 F.3d 248 (4th Cir. 2005) 32

Bennett v. Nucor, 656 F.3d 802 (8th Cir. 2011)........................................ 33

Bradley v. Pizzaco, 939 F.2d 610 (8th Cir. 1991)....................................... 59

Bryte ex rel. Bryte v. America Household, 429 F.3d 469

(4th Cir. 2005)................................................................................... 26,27

Burns v. Anderson, F.App’x 543 (4^th Cir. 2005)........................................ 47

Capaci v. Katz & Besthoff, 711 F.2d 647 (5th Cir. 1983)....................... 38,51

Chin v. Port Authority of N.Y. & N.J., 685 F.3d 135 (2d Cir. 2012)......... 54

Coates v. Johnson & Johnson, 756 F.2d 524 (7th Cir. 1985).................... 36

Cochran v. Brinkman, 2009 WL 4823858 (N.D. Ga. Dec. 9, 2009)....... 45

Cook v. Boorstin, 763 F.2d 1462 (D.C. Cir. 1985)...................................... 36

Daubert v. Merrell Dow Pharmaceuticals, 509 U.S. 579 (1993)..... passim

Dothard v. Rawlinson, 433 U.S. 321 (1977)..................................... 31,51,58

Duckworth v. State Admin. Board of Electric Laws, 332 F.3d 769

(4th Cir. 2003)........................................................................................ 26

Dunning v. Bush, 536 F.3d 879 (8th Cir. 2008)........................................ 44

Dwyer v. Smith, 867 F.2d 184 (4th Cir. 1989)........................................... 27

TABLE OF AUTHORITIES (con’t)

EEOC v. Dial, 156 F. Supp. 2d 926 (N.D. Ill. 2001)................................... 65

EEOC v. General Telegraph Co. of NW, 885 F.2d 575

(9th Cir. 1989).......................................................................................... 52

EEOC v. General Electric, 532 F.2d 359 (4th Cir. 1976)........................... 68

EEOC v. Joint Apprenticeship Committee, 186 F.3d 110

(2d Cir. 1999)..................................................................................... 49,51

EEOC v. LA Weight Loss, 509 F. Supp. 2d 527 (D. Md. 2007)................ 65

EEOC v. Mitsubishi, 990 F. Supp. 1059 (C.D. Ill. 1998)............................ 65

EEOC v. Princeton Healthcare, 2012 WL 5185030

(D.N.J. Oct. 18, 2012)............................................................................. 66

EEOC v. Radiator Specialty, 610 F.2d 178 (4th Cir. 1979)................. 59,60

EEOC v. Scolari Warehouse Markets, 488 F. Supp. 2d 1117

(D. Nev. 2007)......................................................................................... 65

EEOC v. Sears, 839 F.2d 302 (7th Cir. 1988)............................................. 51

EEOC v. Sterling Jewelers, 2010 WL 86376

(W.D.N.Y. Jan. 6, 2010)..................................................................... 65,67

Gallagher v. S. Source Packaging, 568 F. Supp. 2d 624

(E.D.N.C. 2008)...................................................................................... 41

General Electric Co. v. Joiner, 522 U.S. 136 (1997).................................. 46

General Telegraph Co. v. EEOC, 446 U.S. 318 (1980).............................. 64

Golden Nugget v. Chesapeake Bay Fishing,

93 F. App’x 530 (4^th Cir. 2004)............................................................. 43

TABLE OF AUTHORITIES (con’t)

Green v. Missouri Pac.R.R., 523 F.2d 1290 (8^th Cir. 1975)................. 28, 59

Gregory v. Litton System, 316 F. Supp. 401 (C.D. Cal. 1970),

aff'd 472 F.2d 631 (9th Cir. 1972)......................................................... 59

Griggs v. Duke Power, 401 U.S. 424 (1971)............................................... 23

Hoyle v. Freightliner LLC, 650 F.3d 321 (4th Cir. 2011).......................... 42

In Re New Bern Riverfront Development,

2013 WL 4834016 (E.D.N.C. Sept. 10, 2013).................................. 42,43

Keener v. United States, 181 F.R.D. 639 (D. Mont. 1998)......................... 41

Kopf v. Skyrm, 993 F.2d 374 (4th Cir. 1993)............................................. 46

Lilly v. Harris-Teeter Supermarket, 720 F.2d 326 (4th Cir. 1983)........ 36

Luke v. Family Care & Urgent Med. Clinics,

323 226 F.R.D. 536 (S.D. W. Va. 2005)................................................. 41

Luma v. Stryker, 226 F.R.D. 536 (S.D. W.Va. 2005)............................... 43

Mayor & City Council of Baltimore v. Unisys, 2013 WL 4784118 (D. Md. Sept. 5, 2013).............................................................................................................. 43

Marvin Lumber & Cedar v. PPG Industrial,

401 F.3d 901 (8th Cir. 2005)........................................................ 48,53,56

McReynolds v. Sodexo Marriott Services,

349 F. Supp. 2d 30 (D.D.C. 2004)............................................. 48,49,50

National Railroad Passenger Corp. v. Morgan,

536 U.S. 101 (2002)........................................................................... 63,66

Nnadili v. Chevron U.S.A., 2005 WL 6271043

(D.D.C. Aug. 11, 2005)............................................................................ 40

TABLE OF AUTHORITIES (con’t)

Paige v. California, 291 F.3d 1141 (9th Cir. 2002)..................... 36,37,54,57

Phillip v. Cohen, 400 F.3d 388 (6th Cir. 2005)......................................... 55

Quiet Tec. DC-8 v. Hurel-Dubois UK Ltd., 326 F.3d 1333

(11th Cir. 2003)....................................................................................... 47

S. States Rack & Fixture v. Sherwin-Williams,

318 F.3d 592 (4th Cir. 2003)............................................................ 27,42

In re Scrap Metal Antitrust Litigation, 527 F.3d 517 (6th Cir. 2008) 47,48

Serrano v. Cintas, 699 F.3d 884 (6th Cir. 2012)....................................... 64

Smith v. City of Jackson, 544 U.S. 228 (2005)......................................... 33

Southwire v. J.P. Morgan Chase, 528 F. Supp. 2d 908

(W.D. Wis. 2007)............................................................................... 48,49

Stagi v. Nat'l R. Passenger Corp., 391 F. App’x 133 (3d Cir. 2010) 31,35,36

TFWS v. Schaefer, 325 F.3d 234 (4th Cir. 2003)...................................... 47

Tucker v. Ohtsu Tire & Rubber, 49 F. Supp. 2d 456 (D. Md. 1999)........ 40

United States v. Allegheny-Ludlum Industrial, 517 F.2d 826

(5th Cir. 1975)......................................................................................... 65

United States v. County of Fairfax, 629 F.2d 932 (4th Cir. 1980).......... 58

United States v. Fresno Unified Sch. District, 592 F.2d 1088

(9th Cir. 1979)................................................................................... 64,65

United States v. Gregory, 871 F.2d 1239 (4th Cir. 1989)......... 58,59,60,63

Wards Cove Packing v. Atonio, 490 U.S. 642 (1989)............................... 27

Watson v. Ft. Worth Bank & Trust, 487 U.S. 971 (1988)......................... 30

TABLE OF AUTHORITIES (con’t)

Williams v. Giant Food, 370 F.3d 423 (4th Cir. 2004)....................... 67,68

Federal Statutes

42 U.S.C. §2000e-2(k)(1)(A)....................................................................... 24

42 U.S.C. §2000e-2(k)(1)(A)(i).............................................................. 27,29

42 U.S.C. §2000e-2(k)(1)(B)............................................................ 25, 29,32

42 U.S.C. §2000e-2(k)(1)(B)(i).................................................................... 32

42 U.S.C. §2000e-5 ("§706")............................................................... passim

42 U.S.C. §2000e-5(e)(1)............................................................................. 68

42 U.S.C. §§2000e-5(f)(3).............................................................................. 1

42 U.S.C. §2000e-5(g)(1)............................................................................. 63

42 U.S.C. §2000e-6 ("§707")............................................................... passim

28 U.S.C. §1291................................................................................................ 1

28 U.S.C. §1331................................................................................................ 1

Rules

Fed.R.App.P. 4(a)(1)&(5)................................................................................ 1

Fed. R. App. P. 32(a)(5).................................................................................. 1

Fed. R. App. P. 32(a)(6).................................................................................. 1

Fed. R. App. P. 32(a)(7)(B)............................................................................. 1

TABLE OF AUTHORITIES (con’t)

Fed. R. App. P. 32(a)(7)(B)(iii)....................................................................... 1

Fed. R. Civ. P. 26(e)(1).................................................................................. 39

Fed. R. Evid. 37(c)......................................................................................... 23

Fed. R. Evid. 702............................................................................. 13,18,19,46

Legislative History

137 Cong. Rec. H9505-01............................................................................. 34

Miscellaneous

EEOC Enforcement Guidance on the Consideration of Arrest and conviction Records in Employment Decisions under Title VII (4/25/12)................... 34

STATEMENT OF JURISDICTION

Plaintiff-Appellant Equal Employment Opportunity Commission filed this action against Defendant-Appellee Freeman under Title VII of the Civil Rights Act of 1964, 42 U.S.C §§2000e et seq. JA23.[1] EEOC alleged Freeman engaged in a pattern or practice of discrimination against Blacks and males because its credit and criminal background check policies had a disparate impact on Blacks, the criminal policy had a disparate impact on men, and neither policy was job-related or consistent with business necessity. The district court had jurisdiction under 42 U.S.C. §§2000e-5(f)(3) and 2000e-6(b), and 28 U.S.C. §1331. On August 9, 2013, the court entered final judgment against EEOC. JA1078. After an extension, EEOC timely appealed on November 6, 2013. JA1303;JA1305. See Fed.R.App.P. 4(a)(1)&(5). This Court has jurisdiction under 28 U.S.C. §1291.

STATEMENT OF THE ISSUES

1. Whether the district court erred in concluding EEOC failed to identify the “particular employment practice” under challenge where EEOC isolated out from Freeman’s decisionmaking process two individual sub-elements of Freeman’s background check policy for challenge: credit checks and criminal checks.

2. Whether the court abused its discretion in finding EEOC’s supplemental expert reports untimely and in excluding them as a discovery sanction.

3. Whether the court abused its discretion in excluding the expert reports as unreliable and erred in holding EEOC failed to establish a prima facie case of disparate impact discrimination.

4. Whether the court erred in limiting EEOC’s credit claim to acts within 300 days of the charge and the criminal claim to acts within 300 days of EEOC’s formal notice of expansion of its credit investigation.

STATEMENT OF THE CASE

This race and sex discrimination lawsuit challenges Freeman’s use of credit and criminal checks to make hiring decisions. EEOC alleged the credit checks had a statistically significant disparate impact on Blacks; the criminal check had a statistically significant disparate impact on Blacks and men; and neither policy was job-related or consistent with business necessity. EEOC filed suit under 42 U.S.C. §2000e-5 (“§706”) and 42 U.S.C. §2000e-6 (“§707”) of Title VII seeking injunctive relief, back pay, and a bench trial. The court granted Freeman’s partial motion to dismiss, limiting EEOC’s suit to acts within 300 days of the charge, which alleged the credit checks were discriminatory. JA48. The court later granted Freeman’s motion for partial summary judgment, limiting the criminal claim to acts within 300 days of EEOC’s formal notice to Freeman of its expanded investigation. JA127. Finally, the court granted Freeman’s motion to exclude EEOC’s expert reports as unreliable and/or untimely and granted summary judgment, holding EEOC failed to establish a prima facie case without those reports and failed to identify the “specific employment practice” under challenge. JA1046. EEOC appealed.

STATEMENT OF FACTS

A. Factual Background

In 2007, Katrina Vaughn, who is Black, applied for a logistics coordinator position at Freeman. JA42. After her final interview, Vaughn anticipated a job offer. But when the phone rang, Freeman did not offer her a job. Instead, Freeman told Vaughn it would not hire her because she had three “charge-offs” from eight years ago, which violated the company’s credit check policy.[2] JA45.

Freeman’s Criminal and Credit Check Policies

Freeman adopted a credit and criminal check policy in 2001. JA456-57; JA468. Although Freeman typically ran background checks only after making a conditional job offer, Freeman told applicants upfront about the checks and required applicants to authorize a search by the company’s vendor, PreScreen America (PSA). JA513(¶¶11,12); JA530. Freeman’s policy candidly acknowledges the point of this practice: to “deter individuals with negative information from applying.” JA530. Once an applicant accepted a conditional offer, PSA ran the checks and reported the results. JA513(¶¶ 11-13). Rejected applicants had five “days to dispute the findings with PSA.” JA529.

Freeman’s credit and criminal policies were set out in Freeman’s one-page “Background Check Criteria for a ‘No Hire’” policy, which addressed disqualifying criteria for motor vehicle records, education, employment and professional license. JA468. The policy listed fourteen disqualifying credit events, ranging from recent home or car foreclosure to “defaulted student loans.” JA468. Freeman also disqualified applicants who had pled guilty, been convicted, or pled no contest to a broad list of twenty-five criminal offenses ranging from “murder” to “more than one misdemeanor or drug charge in the last 7 years.” JA468. Twenty listed criminal offenses had no time limit. Outstanding warrants were a disqualifier.[3]

On July 20, 2006, Freeman issued a revised “Background Checks” policy setting out three “levels” of checks. JA470. “General employee” applicants had criminal checks and social security verification; applicants for “credit sensitive” positions had those plus credit checks; and company officers, general managers, and department heads had social security, criminal, credit, and education/certification checks. JA470-71. Some applicants had motor vehicle record checks. JA470. The list of automatically disqualifying criminal offenses remained the same, but Freeman slightly modified the credit criteria, listing twelve disqualifying criteria. JA472-73. Specifically, the revised policy disqualified applicants with more than two accounts of $300 currently past 90 days due (instead of one active account of any amount 90+ days due); more than three active non-medical collection accounts (instead of one); more than two paid charge-offs in the last year (instead of three years); or any unpaid charge-offs in the last year (instead of more than one unpaid charge-off in the last three years). JA473; JA458(¶8).

On August 11, 2011, Freeman again modified its background check policy. Most significantly, it abandoned credit checks altogether. JA649. Consequently, all employees had criminal and social security checks, and “officers, general managers, and corporate branch department heads” had those plus education/certification authentication; some applicants had motor vehicle record checks. JA649. Freeman also modified its criminal background policy. While the policy had previously represented that applicants were automatically excluded for a listed offense, the revised policy stated that criminal history was used “only when such information is job-related and consistent with business necessity” and that the company did a “case-by-case” analysis based on the nature/gravity of the offense, time elapsed, and nature of job sought. JA650. The revised policy also stated that “ordinarily” no one would be excluded for offenses more than seven years old and that applicants had the opportunity to clear up any pending warrants. JA650-51.

Investigation and Lawsuit

After Freeman refused to hire Vaughn, she filed an EEOC charge alleging that Freeman discriminated against her based on her race, Black. Vaughn said Freeman had told her she would be hired “contingent on . . . passing a drug, criminal and credit background check” and that Freeman subsequently told her it denied her hire because of her credit history. JA72. Vaughn added that Freeman “discriminates in this manner against racial minorities, as a class.” JA72. EEOC investigated. On September 25, 2008, EEOC sent Freeman a letter stating, “consistent with the class allegations contained in the Charge . . . and [EEOC’s] previous requests for information,” EEOC was expanding its investigation to include nationwide race discrimination in hiring . . . including credit and criminal checks, as of January 1, 2001. JA74.

On March 27, 2009, EEOC issued a letter of determination finding Freeman’s use of credit and criminal checks violated Title VII. JA547-48. After conciliation failed, EEOC filed suit under §706 and §707 alleging Freeman engaged in a pattern or practice of discrimination because its background checks had an unlawful disparate impact on Blacks (criminal and credit) and males (criminal). [4] JA23.

Freeman and PSA produced voluminous data during the investigation and discovery, including two “investigative spreadsheets” Freeman compiled with credit check information from 1/1/05-10/13/08, and criminal check information from 1/1/07-10/14/08 from 18 of Freeman’s 39 branches; credit and criminal background reports; applicant flow logs; and EEO data sheets. R.957(¶4); JA273(¶¶16,18).

Murphy’s July 26, 2012 Amended Report

EEOC timely disclosed the 7/16/12 expert report of Dr. Kevin Murphy, an industrial psychologist. JA1198. On July 26, 2012, a few days after EEOC’s expert disclosure deadline, EEOC produced Murphy’s amended report, which made corrections. JA318. Murphy explained in the 7/26/12 report that he merged “a large number of data files” to create a master data file of 58,892 applicants. JA325. Murphy discerned from the data the sex of 56,071 applicants and the race of 53,499 applicants. JA325(Table 1). He was “able to verify the outcomes of credit and criminal background checks for a subset of these 58,892 applicants whose background checks were run and reported.” JA325. Table 2 shows that of 1,170 applicants who had a credit check, 116 applicants failed (9.9% of total), and of 1,100 applicants who had a criminal check, 39 applicants failed (3.5% of total). JA326. Table 3 breaks down the “pass/fail” rate by race/sex. As to credit checks, Table 3 shows that 30 out of 127 Blacks failed (23.6% of Blacks) but only 29 out of 695 Whites failed (4.2% of Whites).[5] JA327. As to criminal checks, Table 3 shows that 36 out of 758 men failed (4.7% of men), while just 1 out of 337 women failed (0.3% of women), and 12 out of 171 Blacks failed (7.0% of Blacks) while just 15 out of 623 Whites failed (2.4% of Whites). JA326.

Statistical tests in Tables 4 and 5 establish that the background checks had a statistically significant impact based on race and/or sex. JA328-29. The “chi squared” and “Fischer’s exact statistics” in Table 5 show that the likelihood of finding such a significant difference between the credit check pass rate for Blacks and Whites in a race-neutral system was 1 in 1,000 (Blacks were 5 times as likely to fail as Whites). JA329. Table 4 shows the likelihood of finding such large gender differences in the criminal pass rate was “less than 1 in 1,000.” JA328. Murphy concluded that criminal checks have “a significant and meaningful adverse effect” on males (16 times more likely to fail than females). JA328. Table 5 similarly shows a “significant [Black]-White difference in the likelihood of failing” the criminal check (the likelihood of finding such a large difference in a race-neutral system was 3 in 1,000; Blacks were three times more likely to fail than Whites). JA328-29.

Murphy’s report included a “Research Literature” section. JA331. After surveying the literature, he concluded there was a “lack of data” showing that individuals who fail background checks pose a threat or perform worse than other employees; to the contrary, “broad societal-level data . . . provide[s] compelling evidence of the invalidity” of credit and criminal checks. JA337 (emphasis added). Murphy discussed numerous studies showing that background checks are notoriously inaccurate and unreliable. JA337-38. Murphy also referenced studies showing Blacks have unfavorable credit and criminal histories compared to Whites and that “arrests and convictions will often discourage otherwise qualified individuals from even applying to certain jobs, thus magnifying the potentially discriminatory effects of differential rates of arrest and conviction across racial/ethnic groups.” JA339. Murphy cited a 2009 report showing that 28% of Black men, versus only 4% of White men, have served time in prison. JA340.

Huebner’s July 17, 2012 Report

EEOC timely produced the 7/17/12 expert report of Dr. Beth Huebner, an associate professor of Criminology. JA383. She reviewed and replicated Murphy’s statistical analyses from his 7/16/12 report, agreeing that the criminal checks had a statistically significant disparate impact on Blacks and males. JA386. Huebner also discussed scholarly literature showing that Blacks and men have disproportionate contact with the criminal justice system, including at the arrest phase and during court processing. JA387-88.

Table 1 of Huebner’s report lists twenty-seven counties from various states, including some counties with Freeman offices (data was not available as to each jurisdiction with a Freeman office). JA388-89. The table uses Bureau of Justice Statistics data to show the gender and race of individuals (by county) arraigned in state courts for felony offenses in 2006. Blacks and males were disproportionately overrepresented as felony defendants compared to the county population and were more likely than Whites to be sentenced to incarceration than probation for property/drug crimes. JA388-90. Huebner concluded that men and Blacks are more likely than women and Whites to be arraigned in state court and are disproportionately overrepresented. JA390. Huebner also stated that men are more than three times as likely as women to be in prison and that based on current estimates, 29% of Black men, but only 4.4% of White men, could expect to spend time in prison. JA390. Table 2 shows state incarceration rates for 2005 by sex, race, and ethnicity in states identified as relevant to this litigation. JA398-99. It reveals “statistically significant” differences in the incarceration rates for men (“at minimum seven times more likely than women to be incarcerated”), JA398, and Blacks, JA399-400.

Stay of Discovery

On October 3, 2012, the court entered an amended scheduling order requiring Freeman to file its Rule 26(a)(2) expert disclosures by November 23, 2012; making EEOC’s rebuttal expert disclosures due January 9, 2013; and setting February 6, 2013, as the deadline for all expert discovery except that related to monetary remedies. JA263. The order did not set a trial date. On October 31, 2012, the Office of Personnel Management (OPM) moved to intervene (to protect the confidentiality of information related to federal government background investigations, which Freeman sought), and a week later the court granted the parties’ joint motion to stay the deadlines until the court ruled on that motion and OPM’s motion for a protective order. JA265-70; R.104. On November 28, 2012, the court granted OPM’s motion to intervene, R.106, but did not reset the discovery deadlines, and Freeman never made its Rule 26 expert disclosures.

Motions to Preclude Expert Testimony and for Summary Judgment

On December 18, 2012—while discovery was still stayed—Freeman filed a motion under Fed.R.Evid. 702 and 104 to exclude Murphy’s and Huebner’s reports. R.108. Relying on declarations from Suzanne Bragg, its Vice President of Benefits and Compliance, JA456, and Dr. Mary Dunn Baker, its labor economics and applied statistics expert, JA490, Freeman argued that Murphy’s report was unreliable because it relied on bad data. According to Freeman, Murphy measured the wrong periods, omitted applicants, and made coding and double-counting errors. JA461-62(¶¶28-29);JA463-64(¶¶34-35, 37-42);JA465(¶45). Freeman additionally argued that Murphy failed to identify the particular employment practice under challenge. R.108. Freeman moved for summary judgment based on these arguments. R.114.

EEOC’s Response and Murphy’s Supplemental Report

EEOC filed a response with a new declaration from Murphy (“1/17/13 report”). JA786. After reminding the court that discovery was stayed and no trial date was set, R.122,p.6, EEOC argued Murphy’s report was reliable. Countering the criticism that he had omitted applicants, Murphy stated that from a “statistical reliability perspective” it was unnecessary to include each and every applicant’s screening outcome. JA788(¶6). Murphy also stated that his “Freeman75” database included all applicants for whom, based on data provided by Freeman, there was a “clear and specific indication” of their background check outcomes and race and/or gender. JA796-97(¶ 17(a)) (referring to his “conservative approach” for identifying outcomes). Murphy explained that he intentionally did not make any inferences about supplied information and did not modify any apparent anomalies in the data provided, such as seemingly incorrect coding of sex. JA797(¶17). He said that the exclusion from his database of all but 19 post-10/14/08 applicants and criminal checks from half of Freeman’s locations was due to limits in Freeman’s data. JA797-98(¶17).

Addressing the criticism he included “pre-limitations” data, Murphy asserted “it is considered reliable methodology to include” such data because the relevant scientific question is simply whether the selection procedure has a significant adverse impact on a particular demographic group, not whether a particular individual was entitled to recover. JA789(¶10). As for the omission of applicants from EEOC’s damages report prepared by Sovan Tun or from EEOC’s interrogatory responses, Murphy said it would be “unreliable and inappropriate statistical methodology” to rely on these applicants, as it “would skew the analysis in favor of a finding of a larger adverse impact” to single out applicants already identified as part of the class. JA795-96(¶¶15-17).

Instead of refuting each and every criticism of his coding or database, Murphy countered Bragg’s criticisms in toto by re-analyzing the data to correct for the purported errors (coding, double-counting, temporal limits, and “cherry-picking” of 19 applicants from after 10/08, of whom 18 failed). JA790(¶13). Murphy’s re-analysis still showed a statistically significant disparate impact on Blacks (criminal and credit) and men (criminal). Addressing Bragg’s criticism he omitted numerous applicants whose information could have been captured, Murphy used “a less conservative methodology” (relying on spreadsheets including data from 2,788 background checks and the background checks themselves), to create an “augmented file” with additional applicants. JA799-800(¶¶19-20). The “augmented file” included 61,079 applicants from 2006-2011, although “most” were “not subject to [a] background check.” JA800(¶19). The augmented analysis again showed a statistically significant adverse impact on Blacks and males; as to credit checks, Table 3 showed that 30 out of 130 Blacks failed (23.1% of Blacks) while 30 out of 732 Whites failed (4.1% of Whites). JA801. As to criminal checks, Table 3 showed that 34 out of 1,068 men failed (3.2% of men) while 5 out of 525 women failed (1.0% of women), and 15 out of 248 Blacks failed (6.0% of Blacks) while 16 out of 928 Whites failed (1.7% of Whites). JA801. Statistical tests in Tables 4 and 5 again established these differences were statistically significant. JA801-02.

Murphy denied that a separate analysis of data by background “level,” job, policy sub-factor, or Freeman’s purported mitigating factors was necessary. JA802-805(¶¶ 21-25). Missing dates precluded him from identifying more than a handful of pre-7/20/06 applicants. JA807(¶27). But, Murphy said, he was able to compare the result of analyzing solely post-7/20/06 credit outcomes versus pooled outcomes (before and after 7/20/06); this “le[]d to exactly the same conclusion”: disparate impact on Blacks. JA807(¶27).

EEOC additionally argued that Freeman had not disputed that external published statistics were admissible and probative of EEOC’s prima facie case.

Freeman’s Reply and Bragg’s Second Declaration

Freeman replied with a second Bragg declaration disputing Murphy’s analysis and adding new exhibits. JA956. Bragg asserted that Murphy’s “corrected analysis” retained many of the errors Murphy purported to fix.

JA961-62(¶¶15-16). Freeman also argued that Murphy’s “augmented analysis” retained errors and made new ones. JA963-69(¶22-39). Finally, Freeman argued that Murphy’s and Huebner’s discussion of external statistics failed to show disparate impact.

Motion to File a Sur-Reply and Supplemental Reports

EEOC moved to file a sur-reply. Because the court did not immediately rule, on April 22, 2013, EEOC served Freeman with Murphy’s 4/16/13 supplemental expert report and Huebner’s 3/11/13 supplemental report, which addressed Freeman’s criticisms. JA1307;JA1325.

In the 4/16/13 supplement, Murphy elaborated on the challenge he faced in assembling a database from the 40 separate data files produced to him, the 3,237 PSA background reports, and the 12,000 pdf documents. JA1309-11. The only common identifier across files was name, although even that varied by file (e.g., some with middle names, some without). JA1309(¶7). His primary sources were four large applicant flow log files, one large EEO report file, and a large set of PSA checks. JA1310(¶10). These files overlapped substantially with the investigative spreadsheets Bragg claimed Murphy relied exclusively on, Murphy said, but he stated that he did not copy them to create his “Freeman75” database. JA1310(¶12).

Murphy also disputed changing gender/race codes, as Bragg claimed, and said Bragg possibly “had access to other information” suggesting Murphy made coding errors, but that this information had “not [been] provided” to him. JA1311(¶12). Inevitably, Murphy said, some screening decisions could not be reliably captured. Id. He explained that “Freeman75” looked at hiring decisions from 2005-2008 because those were the only applicants for whom he could conclude with “scientific certainty” their race, sex, and outcome. JA1311(¶13). However, to address Bragg’s criticism regarding the temporal scope, he re-reviewed “over 12,000 pages of documents” and identified the gender/race for several hundred additional applicants. JA1316(¶17(b)(14)). Based on his re-review, and using a less conservative approach, Murphy was able to extract 2,911 individuals coded for hireability; 127 (4.4%) were coded “not hireable.” JA1312(¶15). Murphy emphasized this data included “the entire population of individuals for whom the outcomes of credit checks or criminal checks and relevant demographic traits c[ould] be determined.” JA1318(¶19).

Although Murphy repeated that any coding inaccuracies came from Freeman’s data, Murphy incorporated Bragg’s suggestions and still found a statistically significant disparate impact. JA1313-16(¶17) (meticulously listing all coding errors and omissions identified by Bragg). As to credit history, Revised Table 5 shows that 149 Blacks passed, while 36 failed (19.5% of Blacks) and 937 Whites passed and 36 failed (3.7% of Whites). JA1318(¶19). Revised Table 5 also shows that 393 Blacks passed their criminal checks and 13 failed (3.2% of Blacks); in contrast, 1,908 Whites passed, while 23 failed (0.8% of Whites). Id. Revised Table 4 shows that 3,206 men passed the criminal checks and 87 failed (2.6% of men), while 1,769 women passed and five failed (0.3% of women). JA1319-20(¶ 22). Murphy stated that just like his previous analyses, these outcomes showed that the checks had a statistically significant disparate impact on Blacks and men. JA1319(¶¶20-22).

Finally, Murphy analyzed credit check outcomes separately for applicants checked before and after July 20, 2006. He found a statistically significant adverse impact on Blacks during both periods. JA1320-21(¶24). In fact, the racial disparity grew under the new policy. Under the old policy, 14.7% of Blacks and 4.8% of Whites failed, but under the new policy, 19.5% of Blacks and just 3.4% of Whites failed. JA1320-21(¶24).

EEOC served Huebner’s supplemental report along with Murphy’s. JA1023;JA1325. Responding to Freeman’s criticisms, this report focused on felony convictions. JA1328. It showed that 83% of state court felons are male and 38% are Black, although Blacks make up just 12% of the adult population. JA1328. Men and Blacks were more likely to be sentenced to a term of incarceration, and a longer term, than women or Whites. JA1328-29. Federal courts results were similar. JA1330-31. Huebner stated that no publicly available data allowed her to isolate which convicted offenders might qualify for various jobs at Freeman. JA1331.

B. District Court Decisions

1. Motion to Dismiss

Early in the case, the court granted Freeman’s motion to dismiss all claims relating to hiring decisions made before March 23, 2007, which was 300 days before Vaughn’s charge. JA48. The court rejected EEOC’s argument that the limitation period of §706 does not attach to the government’s §707 pattern-or-practice actions. JA54. The court also rejected EEOC’s argument that the continuing violation doctrine applies to pattern-or-practice discrimination, rendering inapplicable the 300-day limitation period. JA55-58.

2. Motion for Partial Summary Judgment

The court granted Freeman’s motion for partial summary judgment, limiting the criminal claim to acts after November 30, 2007, which was 300 days before EEOC’s letter formally notifying Freeman of its expanded investigation. JA127-37.

3. Motions to Exclude and for Summary Judgment

The district court granted Freeman’s motions to preclude expert testimony and for summary judgment and denied EEOC’s motion to file a sur-reply. JA1046. The court did not begin with the facts or the law. Instead, the court began by highlighting EEOC’s recently-filed lawsuits in other cases in other courts challenging criminal checks and a Wall Street Journal article—which neither party had ever cited—discussing those cases and Freeman. JA1047. The court acknowledged that “[b]ecause of the higher rate of incarceration of African-Americans than Caucasians, indiscriminate use of criminal history information might have the predictable result of excluding African-Americans at a higher rate than Caucasians.” JA1047. Nevertheless, the court said, criminal checks remain “an important, and in many cases essential, part of the employment process.” JA1047.

The court next recited the facts and procedural history. But the court made an error that seemed to reflect a fundamental misunderstanding of the case’s procedural posture. Citing the scheduling order that had been stayed, the court stated that “the parties completed their expert disclosures in accordance with the court’s amended scheduling order,” including that “Defendant’s 26(a)(2) expert disclosures be completed by November 23, 2012, and [that] EEOC’s rebuttal expert disclosures be completed by January 9, 2013.” JA1054(citing R.100). This statement was incorrect, as the stay of discovery meant Freeman never filed its expert disclosure and EEOC never filed a rebuttal report.

The court went on to conclude that Murphy’s “inaccurate database” rendered his conclusions unreliable under Daubert v. Merrell Dow Pharmaceuticals, 509 U.S. 579 (1993). Focusing on the 7/26/12 report, the court criticized Murphy’s analysis as a non-random sample from the wrong time period that included numerous errors. JA1060-64. Murphy’s supplements did not save his analysis, the court said. JA1065. The court stated that the “first supplemental report” failed to remove the “cherry-picked” applicants or correct all coding errors; added only a handful of new applicants from after October 2008; failed to exclude the pre-7/20/06 credit checks; and introduced more errors. JA1065. The court did not consider the reliability of the 4/16/13 supplement. JA1065.

The court additionally found the supplemental reports untimely. JA1066-68. While the court seemed to accept the 7/26/12 amended report as timely, the court found the 1/17/13 and 4/16/13 reports untimely. The court rejected EEOC’s argument they were admissible as timely Rule 26(e) supplements. JA1066-68. Concluding the belated disclosures were neither “substantially justified or harmless,” the court excluded the “supplemental reports” under Fed.R.Evid. 37(c). JA1068.

The court also found Huebner’s expert report unreliable because it replicated Murphy’s original 7/16/12 report. JA1068. The non-applicant statistics in Murphy’s and Huebner’s reports could not establish disparate impact, the court said, because they were not representative of the relevant applicant pool and concerned arrest and incarceration rates. JA1068-69. Thus, EEOC failed to establish a prima facie case. JA1069-70. The court added that Freeman had no burden to rebut EEOC’s reports; rather, Freeman could establish unreliability by pointing out errors. JA1069.

Summary judgment was also warranted, the court said, because EEOC failed to isolate the specific policy causing the disparate impact. JA1070. The court suggested Murphy should have broken down his analysis by policy sub-criteria, objective versus subjective factors, and “specific job.” JA1070-71. Finally, the court refused to allow EEOC to file a sur-reply. JA1072-74. Concluding that “no facts” showed a disparate impact and EEOC had put employers in a “Hobson’s choice” by bringing these actions, the court dismissed the case. JA1076.

SUMMARY OF ARGUMENT

Over forty years ago, the Supreme Court declared that Title VII prohibits “not only overt discrimination but also practices that are fair in form, but discriminatory in operation.” Griggs v. Duke Power, 401 U.S. 424, 431 (1971). The 1991 Civil Rights Act codified this principle by prohibiting neutral employment practices that have a disparate impact, unless an employer can demonstrate the practices are “job related and consistent with business necessity.” 42 U.S.C. §2000e-2(k)(1)(A). This lawsuit challenges two such practices: Freeman’s use of credit and criminal checks to make hiring decisions.

Study after study has shown that Blacks have worse credit histories than Whites and minorities and men are disproportionately convicted of crimes and incarcerated. Indeed, the district court itself acknowledged that Blacks have higher incarceration rates than Whites. It is therefore no surprise that Freeman’s own applicant flow data established that its use of credit and criminal checks had a statistically significant disparate impact on Blacks (criminal and credit) and men (criminal). The court’s exclusion of EEOC’s expert report showing this impact constituted reversible error.

The court made a threshold error in holding EEOC failed to identify the “specific employment practice” at issue. Contrary to the court’s conclusion, Title VII does not require disparate impact plaintiffs to break down a credit or criminal check policy by each individual sub-factor or by “job category.” Rather, 42 U.S.C. §2000e-2(k)(1) requires only that plaintiffs identify the “particular employment practice” within an employer’s decisionmaking process that causes the disparate impact. EEOC complied with this requirement by isolating out two elements of Freeman’s multi-step hiring process for challenge: credit checks and criminal checks. Freeman’s “divide and conquer” approach, endorsed by the court, is simply contrary to Title VII’s text, legislative history, and purpose, as well as the case law.

The court abused its discretion in excluding as untimely the supplemental reports of EEOC’s experts. They constituted admissible “supplements” under Rule 26. In any event, their exclusion under Rule 37(c) was an abuse of discretion because the belated disclosures were “substantially justified” and/or “harmless,” as discovery was stayed and no trial date had ever been set.

The court’s exclusion of the supplements as unreliable under Daubert was an abuse of discretion. Daubert states that “[p]ertinent evidence based on scientifically valid principles” should be admitted. 509 U.S. at 597. Freeman did not argue Murphy used “junk science,” and Freeman did not challenge the scientific principles or methodology Murphy used to conduct his analysis. Rather, Freeman argued that Murphy used bad data. But attacks on an expert’s database go to the weight of the evidence, not to its admissibility.

Creating a classic “battle of the experts,” Murphy also disputed that many of the purported errors were indeed errors (such as temporal limits of his database). Additionally, Freeman never contended that correcting the purported errors would eliminate the disparate impact. To the contrary, Murphy’s 4/16/13 supplement corrected for all purported errors and omissions and still found a statistically significant disparate impact. In short, any way this pie is sliced—and no matter how many times this pie is sliced—there is a statistically significant disparate impact. The applicant flow analysis therefore established a prima facie case, which the external published statistics corroborate.

Finally, the court erred by applying §706’s individual charge-filing limitation period to this suit challenging a discriminatory policy. The continuing violation doctrine renders that limitation inapplicable to such pattern-or-practice discrimination. The 300-day individual charge-filing limit of §706 is also inapplicable to §707 actions. Even if the 300-day limit applies, however, the court erred by concluding that the criminal claim runs from the date EEOC formally notified Freeman it had expanded its credit investigation to include criminal checks.

STANDARD OF REVIEW

This Court reviews de novo grants of motions to dismiss and for summary judgment. Duckworth v. State Admin. Bd. of Elec. Laws, 332 F.3d 769, 772 (4th Cir. 2003); Bryte ex rel. Bryte v. Am. Household, 429 F.3d 469, 475 (4th Cir. 2005). A district court’s exclusion of expert evidence under Daubert is reviewed for abuse of discretion, as is a court’s imposition of Rule 37(c) discovery sanctions. Bryte, 429 F.3d at 475; S. States Rack & Fixture v. Sherwin-Williams, 318 F.3d 592, 595 (4th Cir. 2003). A court abuses its discretion “if its conclusion is guided by erroneous legal principles.” Bryte, 429 F.3d at 475.

ARGUMENT

I....... EEOC established a prima facie case of disparate impact discrimination.

“[T]he evidentiary requirements of a prima facie case of disparate impact are not onerous.” Dwyer v. Smith, 867 F.2d 184, 189 (4th Cir. 1989). A plaintiff need only identify a “particular employment practice” and show it “causes a disparate impact on the basis” of a protected characteristic. 42 U.S.C. §2000e-2(k)(1)(A)(i); see also Wards Cove Packing v. Atonio, 490 U.S. 642, 656-57 (1989) (“a plaintiff must demonstrate that a specific or particular employment practice . . . created the disparate impact under attack), superseded in part, 42 U.S.C. §2000e-2(k). A plaintiff need not “prove discrimination with scientific certainty; rather . . . [plaintiff’s] burden is to prove discrimination by a preponderance of the evidence.” Bazemore v. Friday, 478 U.S. 385, 400 (1986) (Brennan, J., concurring in part). Once a plaintiff establishes a prima facie case, the burden shifts to the employer to show its employment practice was job related and consistent with business necessity. 42 U.S.C. § 2000e-2(k)(1)(A)(i). Thus, a background check that “operates to disqualify [B]lacks for employment at a substantially higher rate than [W]hites and is not job related” or consistent with business necessity violates Title VII. Green v. Missouri Pac. R.R., 523 F.2d 1290, 1292 (8th Cir. 1975); see generally EEOC Enforcement Guidance on the Consideration of Arrest and Conviction Records in Employment Decisions under Title VII, (4/25/12), 2012 WL 14998833 (April 25, 2012).

Here, the court held that EEOC failed to establish a prima facie case, meaning Freeman will never be called upon to show how its credit checks (which it has lived without for more than two years) and criminal checks were job-related and consistent with business necessity. The court erred, and it erred by limiting the temporal scope of EEOC’s suit.

A. EEOC identified a particular employment practice.

A threshold requirement for establishment of a prima facie case is identifying the “particular employment practice.” 42 U.S.C. §2000e-2(k)(1)(A)(i). EEOC met this basic requirement by identifying Freeman’s credit and criminal background check policies. The court’s holding to the contrary, JA1070-72, was rooted in a fundamental misunderstanding of Title VII’s requirements and a misapplication of the summary judgment standard.

The court interpreted 42 U.S.C. §2000e-2(k)(1)(B) (“§703”) as requiring EEOC to identify the specific policy sub-criteria causing the disparate impact and to separately analyze the impact by “specific job.” [6] It does not. Section 703(k)(1)(B)(i) states that a plaintiff “shall demonstrate that each particular challenged employment practice causes a disparate impact, except that if the” plaintiff demonstrates “the elements of [the] decisionmaking process are not capable of separation for analysis, the decisionmaking process may be analyzed as one employment practice.” Here, EEOC did not challenge Freeman’s “decisionmaking process” as a whole. EEOC did not even challenge Freeman’s background check policy—which included social security checks, education checks, and motor vehicle checks—as a whole. Instead, as Title VII requires, EEOC isolated for challenge two discrete “elements” of Freeman’s decisionmaking process: credit checks and criminal checks. This complied with §703.

In reaching a contrary conclusion, the court misunderstood what a plaintiff must show to establish a disparate impact. The Supreme Court long ago made clear that a plaintiff must do more than show “statistical disparities in the employer’s work force.” Watson v. Ft. Worth Bank & Trust, 487 U.S. 971, 994 (1988) (plurality opinion). Instead, a “plaintiff must . . . identify[] the specific employment practice that is challenged.” Id. Thus, Watson held, when an employer uses subjective criteria and a standardized rule or test, the plaintiff must “isolat[e] and identify[] the specific employment practices” causing the disparity. Id. Significantly, the Supreme Court did not say it was incumbent upon the plaintiff to break down the specific employment practice by sub-criteria, such as breaking down a test by individual question; to the contrary, Watson suggested that a standardized test is itself a “specific employment practice.” Id. (saying it has been “relatively easy” to identify a “specific employment practice” in standardized test cases). This is because the purpose of the Watson rule is to avoid impact liability based on “bottom line” workforce disparities, not to require the maximum possible disaggregation of data relating to a single step in the decisional process.

Congress later codified Watson’s principle that plaintiffs must identify a specific employment practice into §703(k)(1). Congress softened the effect of Watson by creating a safe harbor for plaintiffs confronted with a decisionmaking process not capable of separation into discrete elements, but EEOC is not arguing here that Freeman’s decisionmaking process is incapable of separation. To the contrary, EEOC separated out a single element from Freeman’s decisionmaking process—its background check policy—and then within that element isolated out for challenge two discrete sub-elements: credit checks and criminal checks. This satisfied Title VII’s particularity requirement. Contrary to the court’s conclusion, Title VII does not require slicing and dicing a discrete employment practice into its smallest sub-particles. See Stagi v. Nat’l R. Passenger Corp., 391 F. App’x 133, 147 (3d Cir. 2010) (unpublished) (“There is no legal requirement to use the smallest possible unit of analysis.”); see also Dothard v. Rawlinson, 433 U.S. 321, 329-30 (1977) (in pre-1991 Act case, finding disparate impact based on the “combined” height and weight requirements for prison guards).

The court’s misreading of §703(k)(1)(B)(i) seemed to be based on a misinterpretation of the provision’s use of the term “elements.” JA1070.

The court appeared to view the prohibition on challenging a generalized “decisionmaking process” unless “the elements . . . are not capable of separation” as a mandate to break down every “particular employment practice” into its individual sub-elements. But “elements” here refers merely to the “particular employment practice.” 42 U.S.C. §2000e-2(k)(1)(B)(i) (If a plaintiff demonstrates “that the elements of a respondent’s decisionmaking process are not capable of separation for analysis, the decisionmaking process may be analyzed as one employment practice.”) (emphasis added). Thus, Congress’ use of “elements” in §703(k)(1)(B)(i) does not mandate breaking down every credit or criminal background check into its smallest anatomical part, just as the statute does not require tests be broken down by individual question.

Precedent from this Court confirms this interpretation of the statute. In Anderson v. Westinghouse Savannah River, 406 F.3d 248, 265-66 (4th Cir. 2005), this Court assumed the plaintiff had identified a specific employment practice where she challenged two aspects of a particular hiring process: selection for interview, and the interview panel’s selection of a candidate. This Court did not require the plaintiff to disaggregate the selection-for-interview process, for instance, by the “six core competencies” (teamwork, leadership, etc.) used by the hiring managers and interview panel. Id. at 256.

The district court cited no cases holding that a criminal/credit check policy must be disaggregated by sub-elements or job category, and the cases the court did cite are either inapposite or support EEOC’s position. For example, in Smith v. City of Jackson, 544 U.S. 228, 241 (2005), the Supreme Court merely reiterated the Wards Cove principle that a plaintiff must do more than “simply allege that there is a disparate impact on workers, or point to a generalized policy.” Applying that principle, the Court held the plaintiffs’ generalized challenge to a pay plan was insufficient because plaintiffs failed to identify “any specific test, requirement, or practice within that pay plan that has an adverse impact.” Id. Unlike the Smith plaintiffs, EEOC did not make a generalized assault on Freeman’s multi-step hiring practice as a whole, or even on Freeman’s background check policy as a whole, but instead on two elements of it: credit checks and criminal checks. Similarly, Bennett v. Nucor, 656 F.3d 802 (8th Cir. 2011), is inapposite because the employer’s selection process there was capable of separation into “objective criteria like experience, training, disciplinary history, and test scores, and subjective criteria such as interview performance and the opinion of the candidate’s current supervisor.” Unlike the Bennett plaintiffs, EEOC separated out two sub-elements of a single step (background checks) of Freeman’s selection process for challenge: credit and criminal checks. Id. at 817-18.

Legislative history also shows that Congress did not intend §703(k) to require slicing and dicing a specific employment practice into its smallest sub-elements and then analyzing it again by job position. The bipartisan interpretive memorandum prepared by Senators Danforth, Kennedy, Hatch, and Minority Leader Dole states that when “‘[a] decision-making process includes particular, functionally-integrated practices which are components of the same criterion, standard, method of administration, or test, such as’” the height/weight requirement of Dothard, “‘the particular functionally-integrated practices may be analyzed as one employment practice.’” 137 Cong. Rec. H9505-01 (daily ed. Nov. 7, 1991) (statement of Mr. Fried, quoting memorandum). Similarly, Rep. Hyde called the “particularity requirement” “not unduly burdensome” and explained, for instance, that a “100-question intelligence test may be challenged” as a whole. Id. (statement of Rep. Hyde).

The district court also erred in applying the “particular employment practice” requirement to this summary judgment record. Murphy opined that disaggregating the policies by sub-factors, timeframe for convictions, background level, or job was “contrary” to the facts and “scientific standards within [his] profession.” JA802-03(¶¶21,22). He observed that while Freeman’s expert said it was possible to disaggregate the analysis by policy sub-criteria, she did “not state that such an analysis is necessary or appropriate to reliably analyze the adverse impact.” JA803(n.4)(emphasis added). As for Freeman’s purported use of “mitigating circumstances,” Murphy stated that “under the [applicable] scientific standards . . . mitigation would not constitute a separate selection procedure.” JA805(¶25). Moreover, he said, the pass/fail data he reviewed would presumably have already taken into consideration any mitigation (i.e., an applicant was hired, or not hired), and, in any event, “no reliable data” allowed him to isolate mitigation. Id. Thus, summary judgment was inappropriate on this record.

Summary judgment was particularly inappropriate here because Freeman never offered evidence that disaggregating the data would negate Murphy’s conclusion of disparate impact. Cf. Stagi, 391 F. App’x at 142 (employer offered evidence that disaggregating 716 “Feeder Pools” leading to specific promotion positions disproved disparate impact on women). The absence of a genuine issue of material fact as to the relevance of disaggregating the data further underscores the district court’s error in granting summary judgment. See id. at 145 (expert’s decision to “aggregate the data, although not obviously correct, i[s] also not obviously incorrect, and so there remains a genuine issue of material fact” as to whether one-year promotion rule caused disparate impact).

Even if the data could have been disaggregated as Freeman proposed, this Court and others have endorsed aggregating data to enhance its reliability. As this Court explained, aggregation increases the reliability of statistics because “by increasing the absolute numbers in the data, chance will more readily be excluded as a cause of any disparities found.” Lilly v. Harris-Teeter Supermarket, 720 F.2d 326, 336 n.17 (4th Cir. 1983) (in pattern-or-practice case, stating that aggregating data by year is “highly preferable”); see Coates v. Johnson & Johnson, 756 F.2d 524, 541(7th Cir. 1985) (“[P]ooling data is sometimes not only appropriate but necessary, since statistical significance becomes harder to attain as the sample size shrinks.”); Cook v. Boorstin, 763 F.2d 1462, 1468-69 (D.C. Cir. 1985) (rejecting defendant’s argument to restrict statistical analysis to particular job categories). Aggregation “is particularly appropriate where small sample size may distort the statistical analysis and may render any findings not statistically significant.” Paige v. California, 291 F.3d 1141, 1148 (9thCir. 2002) (approving aggregation of results of different exams for different supervisory positions in impact case).

Here, only a few thousand applicants had credit or criminal checks during the time period deemed relevant by the district court. Freeman then proposed sub-dividing that sample size even further, into at least 25 smaller pools for the criminal check (for each sub-criteria) and at least eleven for credit (for each sub-criteria), and then dicing even those pools again by timeframe, mitigation, outstanding warrants, lying, and again by background “level.” Given that Freeman and the court suggested disaggregating the data by “job position”—and that Freeman had 154 job titles (R.130,pp.24-25;JA285-88), the sum total of the slicing and dicing Freeman proposed would lead to such an absurdly small sample pool that it would undoubtedly “distort the statistical analysis and . . . render any findings not statistically significant.” Paige, 291 F.3d at 1148.

Freeman’s approach to interpreting §703—which is not “divide and conquer” so much as “divide and divide ad infinitum and conquer”— would, in fact, effectively foreclose disparate impact challenges to background checks of any kind. As this Court recognized in Lilly, 720 F.2d at 336 n.17, a large data pool is needed in order to decrease the likelihood that chance produced the statistical disparity. To require plaintiffs to sub-divide the analysis by sub-elements of a background check policy, and then to sub-divide those results again by job category or position, would leave such small raw numbers of data that it would be impossible to show a statistically significant disparity—which seems to be Freeman’s transparent goal in making its “specific employment practice” argument. See Capaci v. Katz & Besthoff, 711 F.2d 647, 654 & n.4 (5th Cir. 1983) (defendant’s attempts to disaggregate data by year and city was an “unfair and obvious attempt” to make it “difficult to demonstrate statistical significance”; observing that employer’s “divide and conquer” approach would make it “impossible to demonstrate significance . . . since even a record of hiring . . . zero women would not yield statistically significant results”).

B. The district court abused its discretion in excluding the supplemental reports as untimely.

The district court’s exclusion of the 1/17/13 and 4/16/13 reports as untimely constituted an abuse of discretion because they were admissible supplements under Rule 26(e). Even if they were not, the court abused its discretion in excluding them under Rule 37(c) given that no expert depositions had occurred, Freeman never produced its expert report, EEOC never filed a rebuttal report, and no trial date had ever been set. Left to stand, the court’s ruling would impose a new, draconian rule for parties offering expert testimony: get it perfect the first time, or go home. Such a ruling cannot be left to stand, as it cannot be reconciled with the letter or the spirit of Rules 26 and 37.

1. The reports were Rule 26(e) supplements.

Rule 26(e)(1) requires a party to supplement its disclosures “in a timely manner,” or “as ordered by the court,” upon learning the disclosures are “incomplete or incorrect.” Fed.R.Civ.P. 26(e)(1). Rule 26(e)(2) extends this duty to expert reports and depositions, stating, “any additions or changes . . . must be disclosed” before that party’s pre-trial disclosures are due. Thus, Rule 26(e) clearly anticipates that experts will modify their testimony up until the pre-trial disclosure deadline. The district court’s 5/11/12 scheduling order, which had set EEOC’s 7/18/12 expert disclosure deadline, incorporates this expectation, as it had given the parties a full month after EEOC’s 11/19/12 rebuttal deadline to “complete Rule 26(e)(2) supplementation.” JA257.

Despite the fact both Rule 26(e) and the court’s own scheduling orders contemplated time for supplementation of expert reports, the district court ruled that Murphy’s and Huebner’s reports did not count as Rule 26(e) supplements. The court erred. Murphy’s 1/17/13 and 4/16/13 supplements fell under Rule 26(e) because he merely made modifications to his database—modifications proposed by Freeman—while affirming his core conclusions of disparate impact based on race/sex. Courts have permitted supplementation under Rule 26(e) even in cases—unlike this one—where an expert offered new, previously undisclosed opinions. See Tucker v. Ohtsu Tire & Rubber, 49 F. Supp. 2d 456, 460 (D. Md. 1999) (stating it was “clear” that expert’s testimony as to “two new opinions” “was a form of supplementation permitted by Rule 26(e)(1)” and permitting belated report based on new testing). Here, Murphy was not even offering new opinions; he merely refined his data—making his reports admissible Rule 26(e) supplements. See Nnadili v. Chevron U.S.A., 2005 WL 6271043, at *2 (D.D.C. Aug. 11, 2005) (affidavit that “refine[d]” expert’s estimate of economic loss constituted Rule 26 supplement); IP Innovation v. Red Hat, 2010 WL 1027479, at *1 (E.D. Tex. March 11, 2010) (denying motion to strike untimely supplemental expert report attached to surreply to Daubert motion). Similarly, Huebner’s 3/11/13 report constituted a supplement because she merely refined the kind of data she evaluated to affirm her core finding—that criminal checks have a disparate impact on Blacks and males.

This Court’s opinion in Campbell v. United States, 470 F. App’x 153 (4th Cir. 2012), does not require a contrary result. Campbell held the plaintiff’s supplemental expert report inadmissible under Rule 26(e)(1). Id. at 157. But, unlike here, the plaintiff’s initial expert report was untimely and it failed to comply with Rule 26(a). Id. at 156-57. Thus, there was no compliant expert report to “supplement.” See id. at 157 (supplement “did not simply add or correct information, but rather attempted to recast” initial opinion to comply with Rule 26(a)). The four other cases relied upon by the district court are also distinguishable because the belated reports either offered new or dramatically different opinions than the initial reports. See Luke v. Family Care & Urgent Med. Clinics, 323 F. App’x 496, 500 (9th Cir. 2009) (untimely declaration offered “new theory of causation”); Gallagher v. S. Source Packaging, 568 F. Supp. 2d 624, 629 (E.D.N.C. 2008) (new report changed expert’s conclusion concerning lost revenue); Keener v. United States, 181 F.R.D. 639, 641 (D. Mont. 1998) (belated report was “dramatic[ally]” different); Cochran v. Brinkman, 2009 WL 4823858, at *6-7 (N.D. Ga. Dec. 9, 2009) (belated report and deposition testimony concerned “new testing” and “new opinions”).

2. The supplements were not excludable under Rule 37(c).

Even assuming, arguendo, the reports were inadmissible under Rule 26, the district court’s exclusion of them under Rule 37(c) as a discovery sanction constituted an abuse of discretion. Rule 37(c)(1) provides that if a party fails to comply with Rule 26, the evidence is excluded “unless the failure was substantially justified or is harmless.” According to this Court, this determination “should be guided” by five factors: (1) surprise to the opposing party; (2) ability to cure the surprise; (3) disruption of trial; (4) importance of the evidence; and the (5) explanation for the non-disclosure. Southern States, 318 F.3d at 597. The first four factors speak mainly to the harmless exception, while the fifth factor “relates primarily to the substantial justification” exception. Id.

Here, the district court failed to cite, much less apply, the Southern States factors. Cf. Hoyle v. Freightliner LLC, 650 F.3d 321, 330 (4th Cir. 2011) (holding district court’s failure to expressly mention Southern States does not constitute an abuse of discretion, but reaffirming importance of Southern States factors to analysis). Reversal is required here because application of this Court’s Southern States factors establishes that the disclosures were both “substantially justified” and “harmless.”

First, there was no “surprise” to Freeman, as EEOC timely identified Murphy and Huebner and served their original reports, and the supplemental reports merely refuted Freeman’s own criticisms of Murphy’s and Huebner’s analyses while confirming the original findings of disparate impact. See In Re New Bern Riverfront Dev., 2013 WL 4834016, at *6 (E.D.N.C. Sept. 10, 2013) (surprise was “minimal” where plaintiff timely identified the witness and produced his report; witness had not been deposed; and discovery deadline was four months off). Second, even if Freeman somehow was surprised by a report that made no new conclusions, Freeman had “both the ability and the opportunity to cure” it through depositions. Id. (surprise could be cured by continuation of expert’s deposition); see Golden Nugget v. Chesapeake Bay Fishing, 93 F. App’x 530, 534 (4th Cir. 2004) (second factor weighed against exclusion of expert testimony disclosed for first time at trial where other party “had the ability to cure the surprise but failed to capitalize on opportunities to do so”); Luma v. Stryker, 226 F.R.D. 536, 545 (S.D. W.Va. 2005) (belated expert reports admissible where reports were produced before the experts’ depositions and did not disrupt trial).

As to the third Southern States factor, admitting the supplemental reports could not have disrupted the trial, as no trial date had ever been set. Fourth, the supplemental reports were extremely “important,” as they went to the heart of EEOC’s prima facie case and showed the scientific insignificance of Freeman’s criticisms of Murphy’s database. See Mayor & City Council of Baltimore v. Unisys, 2013 WL 4784118, at *5 (D. Md. Sept. 5, 2013) (fourth factor weighed against exclusion because doing so “would utterly hamstring the City’s ability to prove its case”). Thus, application of the first four Southern States factors shows that the belated disclosures were exceedingly “harmless,” making their exclusion an abuse of discretion. See Dunning v. Bush, 536 F.3d 879, 889-90 (8th Cir. 2008) (court abused its discretion in excluding expert report disclosed after summary judgment motion but before end of discovery and three months before trial).

Application of the fifth Southern States factor—EEOC’s explanation for the belated disclosures—additionally establishes that the disclosures were “substantially justified.” It was Freeman who opted to file a motion to exclude while discovery was stayed—before Freeman’s expert disclosure deadline, EEOC’s rebuttal deadline, or expert depositions. Consequently, EEOC was substantially justified in filing Murphy’s 1/17/13 report (rebutting Bragg’s and Baker’s declarations); his 4/16/13 report (rebutting Bragg’s second declaration, which made new criticisms and attached new exhibits); and Huebner’s 3/11/3 supplement (responding to criticisms raised for the first time in Freeman’s response brief).

Moreover, the two cases the district court relied upon to exclude the reports are inapposite. In Campbell, the plaintiff’s initial non-compliant expert report was five days late, and her supplement was filed a month later. Stating that “trial was scheduled to begin in less than thirty days” after the government’s motion in limine hearing and that plaintiff’s expert had yet to be deposed, this Court concluded that the “surprise” to the government was “great” and affirmed the exclusion of the report. 470 F. App’x at 157. This case is not Campbell. EEOC timely produced Murphy’s and Huebner’s compliant original reports, Freeman had plenty of time to depose both experts, and no trial date had ever been set.

Cochran v. Brinkman Corp., 2009 WL 4823858, at *13-15 (N.D. Ga. Dec. 9, 2009), is similarly inapposite. There, the plaintiff’s expert timely filed a report about the cause of a turkey-fryer accident. Two days before his deposition, he conducted new testing, and he offered new opinions during his deposition and subsequent reports as to the fryer’s inadequacies. Id. at *6. Not surprisingly, the court held the belated disclosures were not substantially justified or harmless. Consistent with that ruling, the court also struck the part of the expert’s affidavit offered in response to the defendant’s Daubert motion that relied on the belated reports and three other affidavits attached to the plaintiff’s response to the Daubert motion because they had never been disclosed by the expert deadline and/or were irrelevant. Id. at *14-15. Unlike Cochran, EEOC timely identified both experts and produced the supplements prior to depositions, and the supplements all confirmed the core findings of disparate impact in the original reports.

C. The district court abused its discretion in excluding the supplemental reports as unreliable and erred in granting summary judgment.

“Rule 702 is broadly interpreted.” Kopf v. Skyrm, 993 F.2d 374, 377 (4th Cir. 1993). Consequently, exclusion of expert testimony is the exception, not the rule. Fed.R.Evid. 702, Advisory Committee’s Note (2000). Rule 702 provides that expert testimony involving scientific knowledge that will assist the trier of fact to understand the evidence or determine a fact in issue is admissible when the testimony is based on “sufficient facts or data” and is the “product of reliable principles and methods,” and when the expert reliably applied those principles and methods.

In Daubert, the Supreme Court explained that “the inquiry envisioned by Rule 702 is . . . a flexible one” focused on the twin objectives of “relevance and reliability.” Daubert, 509 U.S. at 594-95. “[R]eliability” refers not to “‘accuracy’ [or] validity’” but to “evidentiary reliability.” Id. at 590 n.9 (emphasis in original). “In a case involving scientific evidence, evidentiary reliability will be based on scientific validity.” Id. “The focus” is “solely on principles and methodology, not on the conclusions that they generate.” Id. at 595. Daubert, then, protects juries from “junk science.” Gen. Elec. Co. v. Joiner, 522 U.S. 136, 153 (1997) (Breyer, J., concurring).

1. Murphy’s applicant flow analysis was reliable and establishes a prima facie case.

Freeman did “not mount a true Daubert challenge, for it does not argue that [Murphy’s] methods have not been tested, have not withstood peer review and publication, have excessive rates of error, have no standards for their application, or have not been accepted in their field.” TFWS v. Schaefer, 325 F.3d 234, 240 (4th Cir. 2003) (citing Daubert’s non-exhaustive factors and holding challenge to expert’s calculations goes to “the proper weight to be given” expert’s testimony, “not to its admissibility”). Freeman never argued that Murphy used “junk science.” Freeman never even quibbled about Murphy’s statistical techniques; to the contrary, Freeman’s expert conceded Murphy used “appropriate methods.” JA493(¶12n.11). Rather than challenge Murphy’s methods, Freeman argued Murphy used bad data. Id. But attacks on an expert’s purported use of “unreliable data” go to “the proper weight to afford testimony, not its admissibility.” Burns v. Anderson, 123 F. App’x 543, 549 (4th Cir. Dec. 15, 2004) (affirming admissibility of expert report) (emphasis added); see also Quiet Tec. DC-8 v. Hurel-Dubois UK Ltd., 326 F.3d 1333, 1344-45 (11th Cir. 2003) (Complaints of incorrect data “are of a character that impugn the accuracy of [the expert’s] results, not the general scientific validity of his methods.”); In re Scrap Metal Antitrust Litig., 527 F.3d 517, 529 (6th Cir. 2008) (affirming admissibility of expert report attacked as relying on erroneous data); Marvin Lumber & Cedar v. PPG Indus., 401 F.3d 901, 916 (8th Cir. 2005) (“Generally, even post-Daubert, the factual basis of an expert opinion goes to the credibility of the testimony, not the admissibility.”) (quotation marks and citation omitted); McReynolds v. Sodexo Marriott Servs., 349 F. Supp. 2d 30, 40 (D.D.C. 2004) (purported expert errors go to “weight, not . . . admissibility” of analysis).

Thus, although “the alleged errors and inconsistencies [may provide] grounds for impeaching [Murphy’s] credibility” and the reliability of his ultimate finding, any “mistakes and calculations are not grounds for excluding evidence.” Southwire v. J.P. Morgan Chase, 528 F. Supp. 2d 908, 935 (W.D. Wis. 2007) (emphasis added). Rather, the issue of whether Murphy made errors such as double counting “can be explored fairly through cross-examination.” Id.; see Daubert, 509 U.S. at 596 (“Vigorous cross-examination, presentation of contrary evidence, and careful instruction on the burden of proof are the appropriate means of attacking” evidence alleged to be shaky). Because Freeman alleged only that Murphy made “mistakes . . . in implementing [his] chosen methods,” Daubert did not provide a basis for excluding Murphy’s testimony from trial altogether. Southwire, 528 F. Supp. 2d at 935 (denying Daubert challenge based on double-counting and other purported errors).

But even assuming Freeman made a true Daubert challenge, the court erred in barring Murphy’s testimony altogether because the 4/16/13 supplement corrected every purported database error, temporal error, or geographic error but still found a statistically significant disparate impact based on race (credit and criminal) and sex (criminal). Significantly, the court never even considered the reliability of Murphy’s 4/16/13 report. JA1065 (referring to “first supplemental report”). The court erred, as the 4/16/13 report was admissible, as discussed above, and it was reliable.

To the extent the district court implicitly deemed the 4/16/13 report tainted by errors in Murphy’s earlier reports, the court erred. “The fact that [Murphy] was open to and in fact did correct deficiencies in his preliminary reports argues for the reliability of his testimony, not for its exclusion.” McReynolds, 349 F. Supp. 2d at 39. Moreover, courts have not made perfection the standard of reliability, as virtually every expert report synthesizing voluminous data from sources in varying formats will contain errors. See generally EEOC v. Joint Apprenticeship Comm., 186 F.3d 110, 116 (2d Cir. 1999) (“EEOC’s statistics, while not flawless, are nonetheless sufficient to establish a prima facie case . . . .”).

The court’s exclusion of Murphy’s testimony was also an abuse of discretion because Freeman never even argued that correcting any purported errors would erase the finding of disparate impact. It is not enough for a Daubert challenge to stand and throw darts. See Anderson v. Group Hospitalizations, 820 F.2d 465, 471 (D.C. Cir. 1987) (party challenging statistical data as flawed “bears the burden of showing that the errors or omissions . . . are likely to change the result”) (quotation marks and citation omitted); McReynolds, 349 F. Supp. 2d at 39 (stating that “beyond enumerating alleged errors, defendant fails to explain how these errors had any substantial bearing on the reliability of his reports” and that “absent such a showing,” and in light of the expert’s explanations for purported deficiencies, expert report was admissible). Freeman never explained how the purported errors undermined the reliability of Murphy’s ultimate conclusion of disparate impact.

Nor can such errors negate EEOC’s prima facie case. Once EEOC established a colorable prima facie case, as it did, Freeman needed to show that correcting the errors would negate the disparate impact, which Freeman never did. Consequently, the district court erred in granting summary judgment. See generally Dothard, 433 U.S. at 331 (in discussion of prima facie case, stating that “[i]f the employer discerns fallacies or deficiencies in the data offered by the plaintiff, he is free to adduce countervailing evidence of his own”); Joint Apprenticeship, 186 F.3d at 119-20 (rejecting employer’s attempts to discredit EEOC’s disparate impact statistics as based on stale data where the employer “makes no showing that more recent census data would produce significantly different results;” also rejecting argument that minorities’ lack of interest in apprenticeships undermined EEOC’s statistics and stating, “JAC was obliged to demonstrate that when the alleged missing factors” were accounted for, “no significant disparity exist[ed]”) (emphasis added); Capaci, 711 F.2d at 653-54 (“The defendant must do more than raise theoretical objections to the data . . . instead, the defendant should demonstrate how the errors affect the results . . . .”).

In concluding Freeman had no burden to rebut EEOC’s statistical evidence showing a prima facie case, R.149,p.24, the district court relied on EEOC v. Sears, 839 F.2d 302, 313 (7th Cir. 1988). Sears was a disparate treatment pattern-or-practice case. On appeal, EEOC argued that once it established its prima facie case with statistical evidence, the employer had to produce statistical evidence to rebut EEOC’s statistical analysis. The Seventh Circuit disagreed, saying other categories of evidence can suffice. Id. at 313-14. The Seventh Circuit actually distinguished the cases relied upon by EEOC as “involv[ing] disparate impact claims” and declined to decide whether they applied to disparate treatment cases. Id. at 314. Other courts have also disagreed with Sears. See EEOC v. Gen. Tel. Co. of NW, 885 F.2d 575, 582-83 (9th Cir. 1989) (in pattern-or-practice case, stating that employer “cannot defeat” statistical showing of sex-based disparity in hiring “simply by pointing out possible flaws in EEOC’s data;” instead, employer must “produce credible evidence that curing the alleged flaws would also cure the statistical disparity”) (emphasis added).

In any event, many of the flaws identified by the district court were not flaws at all, while others were not sufficiently serious to render Murphy’s conclusions unreliable under Daubert or to grant summary judgment. Initially, the district court criticized Murphy’s 7/26/12 report as including fewer than 2,014 applicants although Murphy had access to 58,892 applicants. But Freeman ran background checks only after giving a conditional offer; Freeman never suggested anywhere near 58,892 background checks were run, and most applicants did not have checks.[7] JA800(¶19). Criticisms that a “sample size was too small” go to the weight of an expert’s testimony, not its admissibility. Marvin, 401 F.3d at 916.

As for Murphy’s failure to include all applicant outcomes (including post-10/08 applicants), and his exclusion of criminal applicants from half of Freeman’s branch offices, Murphy opined that from a statistical reliability perspective inclusion of this data was unnecessary. JA788,JA796-99(¶¶6,17,18). He further explained that his exclusion of most post-10/08 applicants—which, again, did “not affect the reliability of [his] analysis” or change his conclusions—resulted from the state of Freeman’s files and his “conservative approach” to identifying outcomes and gender/race information. JA797(¶17(a)). He stated that, “in [his] professional experience there is reliable reason to conclude” the checks had a “significant adverse impact during all periods” and that the time period he analyzed was “more than sufficient in size and composition to draw reliable scientific conclusions.” JA798(¶17(c)). And even if Murphy erred in excluding most post-10/08 applicants, this did not render his analysis of pre-10/08 applicant data unreliable.

Freeman also presented no evidence that the omitted applicants differed in any relevant way from those Murphy analyzed, or that including them would erase the disparate impact. To the contrary, Murphy’s 1/17/13 “augmented” analysis (and his 4/16/13 supplement), showed that adding extra applicants did not erase the disparate impact, it confirmed it. See JA800(¶20) (“less conservative methodology” with additional applicants still showed “a significant adverse impact” on Blacks and males); JA1307-22 (4/16/13 supplement).

As to the inclusion of pre-limitations period applicants in the amended report, this criticism is moot if this Court agrees that the 300-day individual charge-filing limitation period does not apply. See, infra, at Section II.A-B. In any event, courts have held “it is appropriate to admit pre-liability data into evidence in a disparate impact case if . . . practices remain similar over a long period of time.” Paige, 291 F.3d at 1149 (approving of inclusion of pre-liability data “taken from examinations and corresponding eligibility lists that expired before the start of the liability period”) (emphasis added); see also Chin v. Port Auth. of N.Y. & N.J., 685 F.3d 135, 152 (2d Cir. 2012) (in disparate impact promotion case, approving reliance on pre-limitations data); McReynolds, 349 F. Supp. 2d at 43 (same). Here, the 2006 credit policy changes were minor. The disqualifying criminal offenses remained the same under the 2001 and 2006 policies. That makes inclusion of the pre-limitations data appropriate under the case law. Inclusion is especially appropriate because Freeman has offered nothing but speculation about whether the ultimate conclusion of disparate impact would have been different had Murphy excluded pre-limitations period applicants, and Murphy’s 4/16/13 supplement proves it made no difference. Additionally, as discussed, supra, aggregating data yields a more reliable result because it decreases the likelihood the disparity was due to chance.

Murphy also disagreed with Freeman’s expert that including pre-limitation applicants was improper. JA789(¶9). He explained that the “relevant scientific question . . . is whether the selection procedure has a significant adverse impact on a particular demographic group,” not whether a particular person in the database is legally entitled to recover. JA789(¶10). Thus, at best, this issue presents a “classic battle of the experts” requiring resolution by a factfinder. Phillip v. Cohen, 400 F.3d 388, 399 (6th Cir. 2005).

As for the omission from the database of EEOC claimants listed in Tun’s damage report or otherwise identified by EEOC, Murphy explained the common-sense principle that singling out for inclusion applicants who were already identified by EEOC as having failed the credit/criminal check would constitute “selection bias.” JA795(¶15). Murphy also explained that it “would be unreliable and inappropriate statistical methodology” to rely on depositions as primary data sources. Id.

The court also erred in finding Murphy’s analysis unreliable because it was based on a non-representative sample. JA1064. Freeman presented no evidence that omitted applicants differed in any relevant way from the larger applicant pool, and Murphy’s inclusion of all applicants he could capture in the 4/16/13 supplement confirmed a statistically significant disparate impact. Additionally, had Murphy randomly selected applicants from among those for whom he had reliable outcome/race/sex data, he would have had even fewer applicants, thereby undermining the reliability of the data by further shrinking the sample size. Finally, like other database criticisms, attacks on the representativeness of Murphy’s database go to the weight of his testimony—to be weighed at trial—not to its admissibility. See Marvin, 401 F.3d at 916 (criticism that wood “samples were not taken from a representative geographical cross-section” went to credibility, not admissibility).

As for the coding errors and double-counting, Murphy disputed he made all the errors ascribed to him, JA800(¶19), n.3; even Freeman seems to admit he did not make all of them, R.130,pp.5-6;JA960-61(¶11); and Murphy stated that it does not necessarily follow that applicants listed twice in his dataset were counted twice in the analysis, as statistical software programs like his can be programmed to avoid duplicate counting. JA791(n.1) (saying he made sure he avoided duplicate counting in his re-analysis). But, again, Freeman never even argued that the disparate impact was the result of database errors in the 7/26/12 report or 1/17/13 supplement. Freeman’s failure to argue this point is telling, as it is clear that even correcting for any errors, there was still disparate impact.

Finally, as to Murphy’s inclusion of pre-7/20/06 credit checks, JA1063, case law supports its admissibility because the 2011 policy made only slight changes. See, e.g., Paige, 291 F.3d at 1149 (“[I]t is appropriate to admit pre-liability data . . . in a disparate impact case if . . . practices remain similar over a long period of time . . . .”). And, again, Freeman presented no evidence, nor ever even argued, that inclusion of pre-7/20/06 outcomes changed the ultimate conclusion of a statistically significant disparate impact, and Murphy’s 4/16/13 report shows it did not—in fact, the racial disparity increased under the 7/20/06 policy. See, supra, at p.19.

2. External Published Statistics were reliable and buttressed EEOC’s prima facie case.

The district court concluded that the external published statistics in Murphy’s and Huebner’s reports were both unreliable and irrelevant to establishing EEOC’s prima facie case because they failed to compare Freeman’s hiring to the qualified applicants in the relevant labor market, and because they looked at arrests and arraignments. The district court abused its discretion in finding the discussion of external published statistics unreliable, as nothing in those studies could be deemed “junk science” under Daubert.

The district court also erred in disregarding these statistics because the Supreme Court and this Court have repeatedly recognized the probative value of non-applicant flow statistics to establishing a prima facie case. See, e.g., Dothard, 433 U.S. at 330 (generalized national statistics established prima facie case); United States v. Gregory, 871 F.2d 1239, 1243-44 (4th Cir. 1989) (“Applicant flow data is not required to prove discrimination through statistics.”). Especially where, as here, the application process likely deterred applicants—indeed, Freeman intended to deter applicants with negative information from applying—external statistics are particularly probative. See Dothard, 433 U.S. at 330 (recognizing that height/weight requirement might have deterred applicants); United States v. County of Fairfax, 629 F.2d 932, 940 n.9 (4th Cir. 1980) (applicant pool may not adequately measure labor market if individuals are discouraged from applying); Gregory, 871 F.2d at 1244 n.14 (same).

That Huebner and Murphy did not break down the external statistics to compare the surrounding labor market to each Freeman branch office does not destroy the probative value of the statistics. “[T]here was no reason to suppose” that the criminal/credit backgrounds of Freeman’s applicants at particular branch offices “differ[ed] markedly from those of the national population.” Dothard, 433 U.S. at 330 (national height/weight statistics showed disparate impact of Alabama prison policy); see also Bradley v. Pizzaco, 939 F.2d 610, 613 (8th Cir. 1991) (generalized evidence of prevalence of pseudofolliculitis barbae among Blacks established disparate impact of no-beard policy). Moreover, courts have held criminal justice statistics relevant to establishing disparate impact. Green, 523 F.2d at 1294 (considering statistics showing higher conviction rate for Blacks in certain areas); Gregory v. Litton Sys., 316 F. Supp. 401, 403 (C.D. Cal. 1970) (generalized statistics as to Black arrest rate showed disparate impact), aff’d 472 F.2d 631 (9th Cir. 1972).

Nor could the statistics be disregarded because they did not address the job qualifications of Freeman’s 150+ positions, as the district court seemed to suggest. JA1069. This Court has said that where it is “manifest as a matter of law” that no special qualifications are required, general population statistics are probative of a plaintiff’s prima facie case. EEOC v. Radiator Specialty, 610 F.2d 178, 185 (4th Cir. 1979). Where it is not manifest whether special qualifications exist, the burden is on “defendant to establish” special qualifications exist and are “not possessed or readily acquired by the general population,” or the statistics will be presumed appropriate. Id. (emphasis added). Only where it is manifest that special qualifications exist might it be inappropriate to rely on general statistics. Id. But even then, “the plaintiff should have an opportunity to adjust its statistical proof to reflect a labor pool base with the special qualifications.” Gregory, 871 F.2d at 1245 n.19 (deputy sheriff did not require special skills).

Here, Freeman did not even identify—much less prove—which positions required special qualifications not generally possessed or readily obtainable. Additionally, Huebner said no data would allow her to determine which convicted offenders might qualify for jobs at Freeman, JA1331, and EEOC cannot analyze what does not exist.

Finally, the court erred in disregarding the external statistics on the grounds that they looked at arrest, arraignment, and incarceration rates. Huebner’s supplement mooted that concern by focusing on federal felony conviction rates (80% male; 20% Black) and state conviction rates (83% male; 38% Black). JA1330;JA1328. But, in any event, arrest, arraignment, and incarceration rates were also probative, as a factfinder could infer that Blacks and males also have a higher conviction rate.

Consequently, both Huebner’s reports and Murphy’s discussion of external statistics (discussing criminal and credit studies showing Blacks and men have higher incarceration rates and lower credit scores) were both reliable and probative. JA824-33.

II. The district court erred in limiting the temporal scope of EEOC’s suit.

The district court also erred by granting the motion to dismiss and for partial summary judgment, limiting EEOC’s credit claim to acts within 300 days of Vaughn’s charge and the criminal claim to acts within 300 days of EEOC’s letter formally notifying Freeman of the expanded investigation.

A. Section 707 has no limitation period.

As the district court noted below, whether the 300-day individual charge-filing limitations period of §706(e) applies to §707 actions is an issue of first impression in the circuits and one that has divided district courts. JA51. Purporting to apply a “plain language” analysis, the court concluded that §706(e)’s limitation applies to government §707 actions. JA52-55. The court erred.

Section 707 authorizes pattern-or-practice actions by EEOC against private entities and by the Department of Justice (DOJ) against state and local entities. Consistent with the essential nature of a “pattern or practice” of discrimination—which extends beyond any one particular application of a company’s standard operating procedure—§707 contains no limitations period. The district court, however, determined that §707(e) engrafts the individual charge-filing limitations period of §706. It does not. Section 707(e) confers on EEOC the “authority to investigate and act on a charge of a pattern or practice of discrimination” against private employers and provides that “[a]ll such actions shall be conducted in accordance with” §706. Read in context, §707(e)’s requirement that “all such actions” be conducted in accordance with §706 refers to EEOC’s “authority to investigate and act on a charge of a pattern or practice of discrimination,” i.e., to EEOC’s administrative process. It does not refer to the limitations period of §706(e), which attaches to lawsuits. Had Congress intended to impose a limitations period on EEOC’s lawsuits, the place for that would have been §707(a), which authorizes the government to sue. But Congress did not include a limitations period in §707(a), and the district court erred in construing §707(e)—which pertains to the processing of a charge—as engrafting the individual charge-filing limitations period of §706.

The court’s reading of §707(e) as incorporating the individual charge-filing limitations period of §706 also cannot be reconciled with the two-year backpay limitation of §706(g). If, as the district court found, §707(e) incorporates the limitations period of §706(e), then it must also incorporate all of §706 -- including the two-year backpay limitation of §706(g). See 42 U.S.C. §2000e-5(g)(1) (back pay accrues no more than two years before charge). But that makes no sense. If Congress wanted to limit §706 recovery to acts within 300 days of a charge, why would it allow backpay recovery for a two-year period before a charge? See Nat’l R.R. Passenger Corp. v. Morgan, 536 U.S. 101, 119 (2002) (“If Congress intended to limit liability [for a hostile work environment] to conduct occurring in the period within which the party must file the charge, it seems unlikely that Congress would have allowed recovery for two years of backpay.”).

The court’s analysis also conflicts with Title VII’s application to DOJ cases. It is undisputed that DOJ’s §707 actions have no limitations period. See Gregory, 871 F.2d at 1240 n.1, 1247 (in DOJ pattern-or-practice case where a charge was filed, holding that “[m]ake-whole relief is justified when the Government has shown a practice or pattern of discrimination and the complainant applied for the job during the period when the discriminatory policy operated”) (emphasis added). The district court’s conclusion that §706 renders EEOC’s authority more narrow than “that possessed by DOJ,” JA54, would therefore lead to an anomalous result: where a charge was filed,[8] EEOC’s §707 suits against private employers would be subject to §706’s limitation period while DOJ’s §707 suits against governmental employers would not. Such a distinction makes no sense and is contrary to legislative intent to give EEOC authority coextensive with DOJ’s. See generally Gen. Tel. Co. v. EEOC, 446 U.S. 318, 328 (1980) (in §706 case, discussing legislative history).

While no circuit has explicitly addressed the applicability of §706’s 300-day individual charge-filing limitation to §707(e) actions, the district court’s holding conflicts with circuit decisions interpreting §707(e). The Ninth Circuit has observed that “[s]ome of the [§706] procedural requirements seem to apply only to individual unlawful employment practices and not to pattern or practice suits.” United States v. Fresno Unified Sch. Dist., 592 F.2d 1088, 1096 n.5 (9th Cir. 1979) (emphasis added). The Ninth Circuit suggested that the 300-day limitation period of §706 would not apply, as “it takes more than one unlawful practice to constitute a ‘pattern or practice’ of employment discrimination,” meaning there was no “certain date” from which the time period would run. Id. Similarly, the Fifth Circuit has rejected the notion that §707(e) incorporates the right-of-intervention of §706(f)(1), stating that Congress intended EEOC’s “investigative and conciliatory” authority in pattern-or-practice be comparable to the agency’s §706 authority and that “there is no indication that Congress intended the duplication of procedures to extend beyond the administrative level.” United States v. Allegheny-Ludlum Indus., 517 F.2d 826, 843-44 (5th Cir. 1975).

While district courts are divided on the issue, several district courts have held explicitly that the individual charge-filing limitations period of §706 does not apply to §707 actions. See EEOC v. Sterling Jewelers, 2010 WL 86376, at *2-5 (W.D.N.Y. Jan. 6, 2010) (300-day limit inapplicable to EEOC’s §706 and §707 actions); EEOC v. LA Weight Loss, 509 F. Supp. 2d 527, 534-36 (D. Md. 2007) (300-day limit inapplicable to §707 claims); EEOC v. Scolari Warehouse Mkts., 488 F. Supp. 2d 1117, 1136 (D. Nev. 2007) (“Unlike private litigants, a statute of limitations does not apply when the EEOC brings a pattern-or-practice suit” under §707.); EEOC v. Dial, 156 F. Supp. 2d 926, 966-69 (N.D. Ill. 2001) (no limitation period in §707 pattern-or-practice harassment action); EEOC v. Mitsubishi, 990 F. Supp. 1059, 1084 (C.D. Ill. 1998) (no limit in §707 case brought on Commissioner’s charge). But see, e.g., EEOC v. Princeton Healthcare, 2012 WL 5185030, at *3 (D.N.J. Oct. 18, 2012) (rejecting §707 argument).

Thus, EEOC is entitled to relief for applicants denied hire during any period during which EEOC proves Freeman’s hiring had an unlawful disparate impact, although back pay would be limited to within two years of the charge.

B. The continuing violation doctrine applies.

The court also erred in rejecting EEOC’s argument—independent of the §707 argument—that the continuing violation doctrine applies to pattern-or-practice discrimination under §706, rendering the 300-day limit inapplicable. JA55-58. In Morgan, 536 U.S. at 114-15, involving a single plaintiff, the Supreme Court held that the continuing violation doctrine does not apply to “serial violations”; rather, discrete acts of discrimination must occur within the charge-filing period to be actionable. But the Supreme Court explicitly stated it was not considering “the timely question with respect to ‘pattern-or-practice’ claims brought by private litigations,” Id. at 115 n.9. Because private litigants can allege pattern-or-practice discrimination only under §706, the Supreme Court had to be excluding §706 pattern-or-practice discrimination from its holding.

Certainly, this case—involving a written policy, applied consistently over a sustained period of time to exclude applicants—is the very kind of pattern-or-practice of discrimination that warrants application of the continuing violation doctrine rather than a rigid limitations period. See Sterling, 2010 WL 86376, at *6-7 (in nationwide pattern-or-practice case alleging managers denied females promotions and equal pay, finding that continuing violation doctrine precluded dismissal of otherwise time-barred claims); Anderson v. Boeing, 222 F.R.D. 521, 546-48 (N.D. Okla. 2004) (applying continuing violations doctrine in §706 pattern-or-practice action). And §706(g)’s two-year backpay accrual limit implicitly recognizes individuals can recover for violations preceding 300 days. Schlei & Grossman, Employment Discrimination Law (5^th ed.) §41.II.B.2.a (“Where a continuing violation exists . . . a court may . . . “take[] into account the effects” occurring within the two-year period, even though they may have stemmed from acts . . . occurring more than two years before”) (quotations and citation omitted).

Williams v. Giant Food, 370 F.3d 423, 429 (4th Cir. 2004), does not compel a contrary finding. Williams rejected the plaintiff’s argument that her promotion denials were part of a broader pattern-or-practice of discrimination that made the continuing violation doctrine applicable. But, unlike here, the plaintiff could not have pursued a pattern-or-practice theory as an individual, which this Court made clear. See id. at 429-30.

C. If the 300-day limit applies, it runs for both claims from the charge.

Even if the 300-day limitation applies, the district court erred in ruling that the criminal claim runs from EEOC’s September 25, 2008, letter formally notifying Freeman of its expanded investigation. JA127-37. EEOC may pursue violations ascertained in the course of a reasonable investigation, so long as they were included in the cause determination and were conciliated. See EEOC v. Gen. Elec., 532 F.2d 359, 365-66 (4th Cir. 1976). The absence of language in Title VII setting out expanded investigation procedures does not mean courts are free to come up with their own determination for when the 300-day limit begins to run. Rather, courts should look to Title VII’s text. That text simply states that “a charge . . . shall be filed within” 180/300 days “after the alleged unlawful employment practice occurred.” 42 U.S.C. §2000e-5(e)(1).

To be sure, General Electric, 532 F.2d 359, held that for purposes of computing the two-year back pay period of §706(g)—which runs from “the filing of a charge”—the operative date was EEOC’s notice to defendant of commencement of a disparate treatment pattern-or-practice case. General Electric, 532 F.2d at 372 (adding that “provided there were no countervailing equities,” it would be an abuse of discretion not to use the date of notice for computing the back pay period). But General Electric concerned back pay—an equitable remedy—and stated that whether the equities weigh against using the date of notice is “a matter better resolved after trial.” Id.

Finally, it defies the realities of the administrative process to think Freeman did not realize EEOC was investigating its criminal policy. Vaughn’s charge specifically mentioned Freeman’s criminal policy, and the 9/25/08 letter references earlier requests for information about it. JA73.

CONCLUSION

EEOC requests reversal of the district court’s judgment.

REQUEST FOR ORAL ARGUMENT

Given the significance of the issues on appeal to EEOC’s enforcement efforts, the Commission respectfully requests oral argument.

[1] “JA” refers to the Unsealed Joint Appendix; “R*” to the district court docket.

[2] “Charge-offs” refer to a creditor’s declaration that a debt is delinquent and unlikely to be collected.

[3] Applicants who lied about any category faced automatic disqualification. JA468.

[4] EEOC also alleged the criminal policy discriminated against Hispanics but later dismissed this claim. JA259.

[5] Table 3 includes only those applicants from Table 2 for whom Murphy had discerned their race/sex, making the total number of applicants in Table 3 lower than in Table 2. JA326-27; see JA325(gender “unknown” for 2,191 applicants; race “unknown” for 5,393 applicants).

[6] In our view, whether the data could have, or should have, been disaggregated by “specific job,” JA1071, is a separate question from whether EEOC identified the “particular employment practice” under challenge. But we discuss that issue here because the court did.

[7] The court also criticized Murphy for relying “almost entirely on the two Excel spreadsheets,” JA1061. Murphy denied this, JA788(¶8), although even if he had, it is hard to see how it rendered his analysis of the data they contained unreliable. Anyway, Murphy added additional applicants in his supplemental reports and still found disparate impact each time.

[8] Another conundrum the court’s view of §707 raises is how the 300-day limitations period would apply when EEOC institutes suit without a charge, as it is entitled to do. See Serrano v. Cintas, 699 F.3d 884, 896 (6th Cir. 2012).