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

P. DAVID LOPEZ

General Counsel

CAROLYN L. WHEELER

Acting Associate General Counsel

JENNIFER S. GOLDSTEIN

Acting Assistant General Counsel

ANNE NOEL OCCHIALINO

Attorney

EQUAL EMPLOYMENT

OPPORTUNITY COMMISSION

Office of General Counsel

131 M St. NE, 5th Fl.

Washington, D.C. 20507

(202) 663-4724 (phone)

(202) 663-7090 (fax)

Annenoel.Occhialino@eeoc.gov

TABLE OF AUTHORITIES iv

ARGUMENT 1

1. EEOC identified a "particular employment practice." 1

2. The court abused its discretion by holding the amended reports were not Rule 26(e) supplements. 6

3. The reports' exclusion under Rule 37(c) constituted an abuse of discretion. 11

4. The district court abused its discretion in excluding Murphy's reports as unreliable. 15

a. Freeman's criticisms concerned credibility, not admissibility. 15

b. Freeman does not address Murphy's 4/16/13 report. 19

c. Murphy's 1/17/13 and 7/26/12 reports satisfied Daubert. 20

i. Omission of relevant outcomes. 20

ii. Pre-limitations data. 22

iii. 2006 credit check outcomes. 25

iv. Representativeness 25

v. Coding, double-counting, and "cherry-picking" errors. 26

5. External statistics were probative. 27

6. Freeman failed to rebut EEOC's prima facie case. 30

7. The court erred in its timeliness rulings. 31

CONCLUSION 35

CERTIFICATE OF COMPLIANCE C-1

CERTIFICATE OF SERVICE C-2

Freeman suggests this Court need not address the district court’s holding that EEOC failed to identify a “particular employment practice” under 42 U.S.C. §2000e-2(k)(1)(“§703”). Resp.16,47. But this is a threshold issue, and the court’s erroneous ruling on this critical legal issue warrants this Court’s attention.

At the outset, Freeman mischaracterizes EEOC’s argument as “any element more specific than the employer’s decision-making process as a whole constitutes a specific employment practice,” meaning challenges to bottom-line statistics or a generalized policy satisfy §703. Resp.49. That is not what EEOC argued. Br.28-33. Rather, EEOC acknowledged that §703(k)(1) does not allow broad challenges but argued that §703(k)(1) does not require plaintiffs do more than isolate a single employment practice within an overall decisionmaking process having a disparate impact. Br.29-33. Here, EEOC satisfied §703(k)(1) by identifying a “particular employment practice” (credit and criminal checks) within Freeman’s “decisionmaking process” (hiring). EEOC did not even challenge Freeman’s background check policy as a whole; rather, it identified the credit and criminal checks as each causing a disparate impact. While Freeman insists that the statute requires drilling down to the smallest possible sub-factor of a policy under challenge, it does not.

Freeman’s view of the particularity requirement cannot be reconciled with the reason for its existence: to prevent liability based on bottom-line workforce disparities. Nor can Freeman’s view be reconciled with Dothard v. Rawlinson, 433 U.S. 321 (1977), which held that the combined height/weight requirements for prison guards had a disparate impact. Freeman contends Dothard is distinct because the height and weight requirements were “functionally integrated practices.” Resp.50-51. But, of course, the height and weight requirements were perfectly capable of separate analysis, as the Court itself noted that the height requirement “would operate to exclude 33.29% of the women” while the weight requirement “would exclude 22.29% of the women.” 433 U.S. at 329.

According to Freeman, Smith v. City of Jackson, 544 U.S. 228 (2005), mandates breaking down a “particular employment practice” into its smallest elements. Resp.49. Smith, however, merely reiterated that plaintiffs must do more than identify workforce disparities or generalized policies. 544 U.S. at 241. Smith’s holding that the plaintiffs’ challenge to the overall, multi-element pay plan failed to satisfy the particularity requirement does not apply here because EEOC is not challenging Freeman’s overall hiring process (or even its overall screening policy). Moreover, Smith suggests that identification of a “specific test”—like identification of Freeman’s credit and criminal checks—would meet the particularity requirement. Id.

Freeman argues that the test analogy is “wholly inapt,” Resp.49, but Freeman is simply trying to evade the logical consequence of its argument. In test cases, employers make a single hiring decision based on an applicant’s cumulative test score. Although it would be possible to challenge the impact of each test question (or each increment of a cutoff score), the Supreme Court never suggested in Smith—or in Watson v. Ft. Worth Bank & Trust, 487 U.S. 977, 994 (1988), that this level of particularity is required. Similarly, EEOC was not required to separately challenge the impact of each sub-factor of each policy, as Freeman made a single hiring decision based on an applicant’s credit or criminal check. See also Tabor v. Hilti, 703 F.3d 1206, 1212, 1221 (10th Cir. 2013) (particularity requirement satisfied where plaintiff challenged “GDCP” promotion system, which “included multiple components” assessing promotion-readiness and allowed managers to waive requirements).

Although Freeman relies on Davis v. Cintas, 717 F.3d 476 (6th Cir. 2013), Davis supports EEOC. The Davis court rejected a disparate impact challenge to an employer’s hiring process but suggested that both credit and criminal checks—which were part of the employer’s sixteen-step hiring process—were themselves specific employment practices. Id. at 480-81, 495-97.

Contrary to Freeman’s argument, Stout v. Potter, 276 F.3d 1118, 1124 (9th Cir. 2002), does not bolster its view of the statue. The Ninth Circuit stated in dicta that it “doubt[ed]” the employer’s “overall screening process”—which measured supervisor evaluations and the “substantive contents of the application,” which itself “tested a candidate’s proficiency on eleven validated competencies”—could be treated as a single employment practice. But EEOC did not challenge Freeman’s overall hiring process, and Stout did not even suggest the plaintiffs would have had to isolate out each individual “competency.”

Tellingly, Freeman fails to respond to EEOC’s argument that Freeman’s interpretation of §703(k)(1) would effectively foreclose disparate impact challenges to background checks. Br.37-38. Freeman seems to concede this point, as it must. Under Freeman’s view of the statute, plaintiffs could almost never show a statistically significant disparate impact, as the number of individuals denied hire due to any particular sub-factor of a policy would be so small as to lack any probative value. See, e.g., Watson, 487 U.S. at 996-97 (statistical evidence based on “small or incomplete data set” may not be probative). But Freeman’s particularity argument does not stop at policy sub-factors. Rather, Freeman argued below (although not on appeal) that after EEOC identified each sub-factor causing the disparate impact, EEOC had to “do this separately for each level of jobs from general employees to executive.” R.114-1,p.19. Truly, this level of disaggregation would preclude any disparate impact challenges to background checks.

Furthermore, Murphy opined that breaking down the policies by specific conviction or credit issue, or by “any other aspects of the selection procedures” (such as conviction timeframe), was “not supported by scientific principles” and contravened “scientific standards in [his] profession.” JA803(¶22). Significantly, Freeman’s expert never said this level of disaggregation was necessary for reliability; she said merely that Freeman’s “descriptions of documents” provided to Murphy led her to believe he could have broken his analysis down by sub-factor. JA500-01;JA803,n.4. Murphy also stated that separately analyzing adverse impact for each of Freeman’s “background levels” and/or job was “contrary to the facts and scientific standards” of his profession. JA802(¶21). As for criminal mitigation, Freeman’s written policies included it only as of 2011; Murphy said “no reliable data” allowed him to isolate mitigation, which was not a separate selection procedure under relevant scientific standards, JA805(¶25); and EEOC disputed Freeman actually considered mitigation, R.121,pp.32-33;R.126,pp.4-8.

The invalidity of Freeman’s argument is best demonstrated by pinning down how many “particular employment practices” Freeman thinks are at stake. According to Freeman, the credit policy has twelve “particular employment practices” within it: (1) eleven disqualifying criteria; and (2) mitigation. The criminal policy has twenty-nine practices: (1) twenty-five disqualifying convictions; (2) warrants; (3) misrepresentations; (4) unspecified mitigation; and (5) a seven-year timeframe. But that is not all, as Freeman asserted these specific employment practices must be further broken down by “job level,” R.108-1,p.24, and the district court suggested they must be further defined by “specific job,” JA1071, and Freeman had 150+ job titles. JA534-38. Title VII’s particularity requirement was intended to avoid liability for bottom-line workforce disparities, not to require this sort of anatomical analysis.

2. The court abused its discretion by holding the amended reports were not Rule 26(e) supplements.

EEOC argued that both Murphy’s and Huebner’s reports constituted Rule 26(e) supplements and that the district court therefore abused its discretion in excluding them.¹ Br.39-40. As EEOC stated, Rule 26(e) contemplates an opportunity for experts to make changes to their reports up until a pre-trial or court deadline, as the rule imposes a duty of supplementation to make “additions or changes” to an expert report upon learning a disclosure was “incomplete or incorrect.” EEOC contended the district court’s most recent scheduling order of 5/11/12 contemplated such changes because it gave the parties a month after EEOC’s rebuttal deadline to complete Rule 26(e) supplementation. Br.39. EEOC explained the supplemental reports merely confirmed the original reports’ core conclusions of disparate impact. EEOC also argued that courts have permitted supplementation even in cases–unlike this one–where an expert offered new conclusions. Br.39-40. Freeman does not dispute these points.

Freeman is left to argue that Murphy did not refine his data but instead “performed a new analysis on an entirely new basis–a purportedly broader data set.” Resp.39. This contradictory assertion makes EEOC’s point. Murphy’s 1/17/13 and 4/16/13 analyses were not based on an “entirely new basis” but instead on a refined version of his original data set. See JA786-809(1/17/13 report); JA1307-1322(4/16/13 report). Rather than throwing out his entire data set and starting over with an entirely new set of applicants, Murphy made corrections, omissions, and additions to his original data set—modifications Freeman insisted were needed. Similarly, Huebner’s 3/11/13 report supplemented her original report by providing “additional demographic and case processing data on individuals convicted in state and federal courts.” JA1328. Freeman cannot explain why Rule 26(e) does not allow an expert to make additions and corrections to his data, run the same test he already ran, and make the same conclusions he already found.

Freeman also fails to rebut EEOC’s explanation for why Campbell v. United States, 470 F.App’x 153 (4th Cir. 2012), is inapposite. Br.40-41. Campbell held Rule 26(e) did not apply because, unlike here, the plaintiff’s initial expert report was untimely and failed to comply with Rule 26(a). Br.40. Freeman calls this a “distinction without a difference” because Murphy’s initial report failed to comply with Rule 702. Resp.40. But Campbell did not hold that an expert report that is subsequently excluded under Rule 702 cannot be supplemented under Rule 26(e). Rather, Campbell, 470 F. App’x at 158, stands for the straightforward proposition that Rule 26(e) cannot be used to make up for a party’s failure to make a timely and compliant Rule 26(a) disclosure.

Although Freeman chides EEOC for relying on out-of-circuit cases, Freeman’s primary authority for its Rule 26(e) argument is Keener v. United States, 181 F.R.D. 639 (D. Mont. 1998). Keener, which this Court has never cited, is inapposite. In Keener, the court found the expert’s initial opinion so cursory it was “tantamount to a non-opinion,” whereas the expert’s second opinion went to “the heart of the case.” Id. at 641. The court held that such a “dramatic, pointed variation of [the] expert’s [initial] disclosure” fell outside Rule 26(e)’s parameters. Id. at 641. Unlike the initial expert report in Keener, Murphy’s and Huebner’s original reports contained clear opinions that went to the heart of this case: that Freeman’s use of credit and criminal checks had a statistically significant disparate impact on Blacks and men. Far from offering a “dramatic, pointed variation,” Murphy’s and Huebner’s subsequent reports merely affirmed their core findings.

Freeman also suggests that Murphy’s 1/17/13 report (and, presumably, his 4/16/13 report) did not fall under Rule 26(e) because it was “based on information the EEOC had prior to its initial disclosure.” Resp.39. But new information is not a prerequisite to Rule 26(e) supplementation. Rather, Rule 26(e)(1)(A) imposes a duty of supplementation upon learning a disclosure is “incomplete or incorrect.” Thus, an expert may supplement under Rule 26(e) based on material available at the time of his initial report. See Talbert v. City of Chicago, 236 F.R.D. 415, 419 (N.D. Ill. 2006) (rejecting argument that expert “should have gotten ‘it right the first time’”).

Freeman additionally argues that Murphy’s 1/17/13 report was not proper supplementation because EEOC never conceded his earlier report was inaccurate or incomplete. This argument fails. As EEOC explained, Murphy disputed the validity of many of the criticisms but went ahead and made the corrections Freeman insisted upon to show they made no difference. Rule 26(e) permits this. See Talbert, 236 F.R.D. at 418, 422 (expert report constituted Rule 26(e) supplement where “the plaintiff was made aware that in the City’s view the initial report was incomplete” and the expert subsequently produced a lengthier supplement using material “the City requested” the report contain).

Finally, it is entirely consistent with the spirit and purpose of Rule 26 to find Murphy’s and Huebner’s reports constitute permissible Rule 26(e) supplementation. “The purpose of Rule 26(a)(2) is to prevent unfair surprise at trial and to permit the opposing party to prepare rebuttal reports, to depose the expert in advance of trial, and to prepare for depositions and cross-examinations at trial.” Richardson v. Korson, 905 F.Supp.2d 193, 198 (D.D.C. 2012) (quotation marks and citation omitted). Here, no trial was set, discovery was stayed (including Freeman’s expert disclosure deadline), and no expert depositions were set. Permitting EEOC to supplement its expert reports therefore did not undermine Rule 26, and the court abused its discretion in holding to the contrary.

3. The reports’ exclusion under Rule 37(c) constituted an abuse of discretion.

EEOC argued that even if Murphy’s and Huebner’s reports were not Rule 26(e) supplements, the district court abused its discretion in excluding them under Rule 37(c) because the untimely disclosures were substantially justified and/or harmless. Br.41-45. Freeman insists that the district court acted within its discretion, but precedent from this Court is to the contrary.

Freeman creates a straw man argument by contending that the district court’s failure to apply Southern States Rack & Fixture v. Sherwin-Williams, 318 F.3d 592 (4th Cir. 2003), did not itself constitute an abuse of discretion. Resp.41. EEOC never argued that it did. Br.42. But EEOC can rely on the Southern States factors as a framework to show why the exclusion of the reports constituted an abuse of discretion. See Southern States, 318 F.3d at 596 (Factors “are helpful in determining whether a party’s nondisclosure of evidence was substantially justified or harmless.”); see also EEOC v. Thompson Contracting, 499 F.App’x 275, 280 (4th Cir. 2012) (Southern States spells out “[t]he test for evaluating . . . substantial justification and harmlessness.”).

EEOC argued that application of the Southern States factors shows the belated disclosure was harmless. Br.41-44. Freeman does not even dispute the first factor–“surprise” to the opposing party–conceding that Freeman could not have been surprised by Murphy’s or Huebner’s supplemental reports, as they merely affirmed their core findings of disparate impact after making data corrections Freeman insisted upon. See DAG Enter. v. Exxonmobil, 2007 WL 4294317, at *1 (D.D.C. 2007) (applying Southern States and finding no “surprise” from two supplemental reports because “defendants had access to the data and knowledge of the [expert’s] methodology . . . prior to his submissions”).

Nor does Freeman dispute that the second Southern States factor—ability to cure any surprise—favors admission of the reports. Any surprise was easily curable because EEOC served the reports while discovery was stayed—before Freeman’s expert disclosure deadline, before expert depositions, months before the trial court hearing, and before any trial date had been set. Br.43.

Instead of addressing directly the third Southern States factor—disruption of trial—Freeman sidesteps it by asserting that Rule 37(c) is not just about trial disruption. Rather, Freeman contends, the rule looks at disruption of motions practice. This argument is unavailing. The clear focus of Rule 37(c) is on trial disruption. See Southern States, 318 F.3d at 597 (articulating third factor as “the extent to which allowing the evidence would disrupt the trial”) (emphasis added); Thompson Contracting, 499 F.App’x at 280 n.4 (same). Here, Freeman offered no explanation for how the belated disclosure of the supplemental reports could have disrupted the trial. The reason for that is self-evident: no trial date had ever even been set.

Even if Rule 37(c) encompasses disruption of motions briefing, the belated disclosures did not prejudice Freeman. The district court could have allowed discovery to play out—as Rule 26 and the most recent scheduling order contemplated—and then ruled on Freeman’s motions based on a full expert record. See Richardson, 905 F.Supp.2d at 200 (refusing to exclude untimely supplemental expert report, although it delayed resolution of pending summary judgment motion, because “no trial date ha[d] yet been set” and “Defendant . . . point[ed] to no prejudice from the Court’s limited reopening of discovery.”). Indeed, this Court has found no prejudice from the untimely disclosure of a witness and his declaration even after discovery, and where the opposing party (EEOC) was given no opportunity to depose the witness before the court ruled on summary judgment. Thompson Contracting, 499 F.App’x at 280.

Although Freeman disavowed the Southern States factors, it next argues that the fourth factor–importance of the evidence–favors exclusion because EEOC would still have Murphy’s 7/26/12 report. Resp.43. Murphy’s two supplemental reports, and Huebner’s report, are very important, however, because they refute Freeman’s criticisms—presented for the first time with its motions—and because the district court credited those criticisms in holding Murphy’s 7/26/12 report unreliable while simultaneously excluding EEOC’s rebuttal of those criticisms as untimely.

Finally, Freeman argues that “EEOC has pointed to no justification for the untimely disclosure.” Resp.41. But EEOC did point to substantial justification: the supplements responded to criticisms Freeman raised for the first time in its briefs and declarations, which Freeman filed before its expert disclosure deadline, expert depositions, or EEOC’s rebuttal report deadline. Br.44. Freeman contends that the “gravamen of EEOC’s argument” is that Freeman cannot move to exclude EEOC’s expert without using its own expert to identify flaws in EEOC’s expert report. Resp.42. Not so. EEOC’s argument is that because Freeman effectively filed its own expert reports (Bragg’s and Baker’s declarations), EEOC was “substantially justified” in filing what amounted to rebuttal expert reports. The federal rules, and the court’s most recent scheduling order, contemplate no less.

4. The district court abused its discretion in excluding Murphy’s reports as unreliable.

EEOC argued that the district court abused its discretion in excluding the reports as unreliable under Daubert v. Merrell Dow Pharmaceuticals, 509 U.S. 579 (1993), because purported flaws in Murphy’s analyses concerned data, not methodology, and therefore concerned weight/credibility issues for trial, not admissibility. Br.46-48; see also East Tennessee Nat’l Gas v. 7.74 Acres, 228 F.App’x 323, 329 (4th Cir. 2007) (challenge to experts’ basis for valuing property—but not their methodology—went to “the proper weight to be given the evidence at trial”). Freeman does not dispute that Murphy used appropriate statistical tests, JA493(¶11), so this case is not about methodology or “junk science.” Cf. EEOC v. Kaplan, -- F.3d -- , 2014 WL 1378197 (6th Cir. 2014) (affirming unreliability of untested “race-rating” methodology). Rather, this case is about purported database errors, a criticism that goes to the weight to be given Murphy’s testimony at trial.

Freeman argues that EEOC’s reliance on TFWS v. Schaefer, 325 F.3d 234, 240 (4th Cir. 2003), for the proposition Freeman did not mount a true Daubert challenge is misplaced because TFWS concerned expert calculations, not methods or data. This argument fails. In TFWS and this case, the expert attacks concerned purported mistakes, not methodology. And while Freeman argues Murphy’s alleged errors went to his testimony’s foundation and therefore to reliability, the same could be said of the expert’s “calculation” errors in TFWS. But in TFWS, this Court held that the criticism that the expert’s calculations failed to support his conclusion went to “the proper weight to be given to Dr. Levy’s evidence, not to its admissibility.” Id.

Freeman also contends Burns v. Anderson, 123 F.App’x 543, 549 (4th Cir. 2004), and McReynolds v. Sodexo Marriott, 349 F.Supp.2d 30 (D.D.C. 2004) are inapposite because Murphy intentionally manipulated data. Resp.28. Putting aside that Murphy did not need to manipulate the data—as each analysis showed a disparate impact—whether Murphy intentionally “cherry-picked” certain data and engaged in “scientific dishonesty,” JA1063, as the district court held, is a credibility determination.² It is axiomatic that credibility determinations should be made at trial, not on summary judgment, and not on the cold paper record of a Daubert motion. See McReynolds, 349 F.Supp.2d at 40 (rejecting defendant’s argument that expert’s testimony “must be excluded because he is untrustworthy, as allegedly evidenced by his ‘willing[ness] to mislead the Court’”).

Like the district court, Freeman contends that Murphy made errors that “skewed” his results. But perfection is not the standard for reliability. See East Tennessee, 228 F. App’x at 328 (expert testimony need not be “irrefutable or certainly correct” to be admissible under Daubert). Although plaintiffs bear the burden of establishing reliability, it is not enough for a Daubert challenge to point to errors without explaining how they undermine reliability. See McReynolds, 349 F.Supp.2d at 39 (holding expert’s report admissible where defendant failed to show errors “had any substantial bearing on the reliability of [the] report” and the expert explained many purported deficiencies). Otherwise, nearly every expert report could be excluded as unreliable.

Here, Freeman and the district court considerably downplayed the challenge Murphy faced in compiling the database. Murphy called it “a difficult task.” JA1309(¶9). Freeman did not hand over a neat and tidy spreadsheet listing each applicant, race/sex, and credit/criminal check outcome. Rather, Murphy had to painstakingly compile the database from 40 voluminous files; 3,237 PSA background reports (up to twelve pages each); and 12,000 pdf pages. JA1309(¶¶7-8). Given the extreme challenge of compiling the database, some errors inevitably occurred. See McReynolds, 349 F.Supp.2d at 39 n.4 (noting it was “probably inevitable in a case as complicated and based on so much data” that both experts would make errors). But database error attacks go to an expert’s credibility, which should be assessed at trial.

The district court’s exclusion of the expert testimony was particularly inappropriate here because—although Freeman refers to a “jury” trial—disparate impact cases are tried before a judge, not a jury. As courts have recognized, Daubert’s standard is relaxed in bench trial cases because “[t]here is less need for the gatekeeper to keep the gate when the gatekeeper is keeping the gate only for himself.” United States v. Brown, 415 F.3d 1257, 1269 (11th Cir. 2005). See Nassri v. Inland Dredging, 2013 WL 256747, at *1 (M.D. La. Jan. 23, 2013) (denying motion to strike expert’s testimony and stating, “because this is a bench trial, the objectives of Daubert are no longer implicated”).

Freeman fails to address directly EEOC’s arguments about the reliability of Murphy’s 4/16/13 report, insisting it was not part of the summary judgment record. Resp.44-45. However, the district court accepted the report as a proffer at the hearing; it is part of the record below; and EEOC argued it was a timely Rule 26(e) supplement or fell under a Rule 37(c) exception.

As EEOC argued, Br.49, the court abused its discretion in disregarding the 4/16/13 report because Murphy ran the data as Freeman insisted it should have been done (correcting all purported errors/omissions), and Murphy still found a statistically significant disparate impact. Freeman has no response to EEOC’s argument that Murphy’s willingness to revisit his data to make corrections and additions—insisted upon by Freeman—“argues for the reliability of his testimony, not for its exclusion.” McReynolds, 349 F.Supp.2d at 39. Cf. Fisher v. Vassar College, 70 F.3d 1420, 1444 (2d Cir. 1995) (reversing where plaintiff’s statistics contained numerous errors and plaintiff demonstrated an “unwillingness to consult reliable sources of data”). Because the 4/16/13 report remedied any errors that arose from the extremely challenging task of creating the database here, the court erred in deeming the report unreliable due to Murphy’s earlier errors. See McReynolds, 349 F.Supp.2d at 40 (rejecting argument that errors expert “may have made earlier during years of data gathering and refining of analyses” warranted exclusion where end product satisfied Daubert).

Contrary to Freeman’s response, none of the purported errors identified by the district court rendered Murphy’s earlier reports unreliable.

Freeman contends the district court acted within its discretion in holding that Murphy’s failure to review all of the documents at the outset to cull out each knowable outcome from the relevant time period, while relying on pre-limitations data, rendered Murphy’s analysis unreliable. Resp.23-24. This Court has rejected a similar argument. See Burns, 123 F.App’x at 549 (criticism that expert “failed to review certain documents which would purportedly influence the valuation” of the at-issue stock and instead relied on “unreliable data” “address[ed] the proper weight to afford” the expert’s testimony, “not its admissibility”).

The district court’s holding also overlooked disputed evidence about the need for including all outcomes. While Freeman’s expert opined that omission of applicant outcomes rendered Murphy’s analysis invalid, JA489, Murphy’s 1/17/13 report states that the “non-inclusion of applicants after October 14, 2008 does not affect the reliability of my analysis.” JA796 (emphasis added). Whether the omission of applicants from twenty-one Freeman branches—which pertained to criminal applicants only—rendered Murphy’s analysis unreliable is also a criticism that might render Murphy’s analysis less credible but did not make it unreliable.

Furthermore, Murphy’s 1/17/13 report states that his “augmented” analysis added criminal applicants from missing branches. JA800(¶19) (listing added branches). The augmented analysis added 40 credit check outcomes; 498 criminal outcomes by sex; and 382 criminal outcomes by race. Compare JA329-30 with JA800-02. He still found a statistically significant disparate impact. To be sure, Murphy’s 4/16/13 report shows he subsequently learned that with a less conservative approach, he could have captured more outcomes across a broader period. But this shortcoming did not justify excluding Murphy’s testimony altogether. See Payne v. Travenol Labs., 673 F.2d 798, 821 (5th Cir. 1982) (temporal gaps of three and nine months in statistical data not “fatal” to finding of discrimination). Moreover, while the district court found Murphy’s 7/26/12 report unreliable because it omitted post-10/08 applicants, the district court failed to explain how this omission rendered Murphy’s analysis of earlier time periods unreliable. Nor did omission of data from earlier years justify granting summary judgment, as disparate impact plaintiffs need not prove a violation as to all years to show a violation for some years.

Although Freeman and the district court rely on Payne, 673 F.2d at 823, to show omission of applicants rendered Murphy’s analysis unreliable, Payne supports EEOC’s argument that this critique goes to the weight of the evidence. Payne held that the value of the defendant’s rebuttal statistics was undermined by the omission of 25% of relevant applicants, but the court made this finding after trial.

EEOC asserted that case law supported Murphy’s reliance on pre-limitations data; Murphy opined that the use of pre-limitations data was appropriate in his field because it involved the same practices (JA789(¶10)); and reliance on pre-limitations data would be a moot criticism if this Court agrees with EEOC’s timeliness arguments. Freeman does not dispute the latter two points but argues that Paige v. California, 291 F.3d 1141, 1149 (9th Cir. 2002), and McReynolds are inapposite because their concern with small sample size would not be at issue had Murphy included all the temporally relevant outcomes. Small sample size is an issue, though, even if Murphy’s report included all knowable outcomes. The district court temporally limited the claims to less than five years, reducing available applicant outcomes. The court additionally held that each sub-criteria of each policy constituted its own employment practice and that the statistics must be broken down by job category/title, making sample size of paramount importance. And the district court, and Freeman, suggest that once all knowable outcomes were culled out, they had to be “representative” of the larger pool of applicants (including those whose outcomes were unknown), which would have further winnowed the applicant pool.

Citing EEOC v. Federal Reserve Bank of Richmond, 698 F.2d 633, 663 (4th Cir. 1983), vacated on other grounds sub nom. Cooper v. Federal Reserve Bank, 467 U.S. 867 (1984), Freeman next argues that the aggregation of data across liability periods is impermissible. Resp.22. But Federal Reserve was about disparate treatment (where the question is intent, not whether a particular practice has a statistically significant disparate impact), and the parties had agreed on the relevant time period. Moreover, this Court has approved using applicant flow statistics from a single year to extrapolate to other years. See United States v. County of Fairfax, 629 F.2d 932, 940 (4th Cir. 1980) (approving government’s extrapolation of applicant flow data from 1978 to show discrimination from 1974-77 where defendant had destroyed earlier records).

Reliance on pre-limitations data is also appropriate here because the record shows Freeman altered its policies in response to EEOC’s lawsuit, hindering EEOC’s ability to show impact. See generally Ellis v. Costco, 28 F.R.D. 492, 529 (N.D. Cal. 2012) (Pre-limitations data may be more relevant because employers “‘may be improving [their] hiring practices to avoid liability or large damages in their pending discrimination case.’”) (quoting EEOC v. Joe’s Stone Crab, 220 F.3d 1263, 1277 n.14 (11th Cir. 2000)). After EEOC filed suit, Freeman dropped its credit check policy, limiting the data EEOC could use to show impact. In 2011, Freeman also modified its written criminal check policy to account for purported mitigation.

Finally, Freeman fails to respond to EEOC’s argument that aggregating data enhances reliability because it makes it less likely chance caused a particular outcome. Br.36; see Eldredge v. Carpenters 46 N. Calif., 833 F.2d 1334, 1339 (9th Cir. 1987) (in impact case, holding that plaintiff could rely on aggregated data from nine-year period and relying on Lilly v. Harris-Teeter Supermarket, 720 F.2d 326, 336 n.17 (4th Cir. 1983)). In fact, where policies “have remained unchanged over a period of time” without “substantial changes,” requiring a plaintiff to break down the data by year “would be unreasonable.” Id. (quotation marks omitted).

Contrary to Freeman’s argument and the district court’s ruling, the mere fact Freeman identified two applicants whose credit outcomes differed under the 2001/2006 policies does not render Murphy’s analysis unreliable. “Pre-liability data” may be used to show disparate impact where employment practices are “similar”—not identical—“over a long period of time.” Paige, 291 F.3d at 1149.

Moreover, the district court’s analysis misses the mark. The court held Murphy’s analysis unreliable because the 2001 credit policy was “stricter” and 30% of the “fails” in his analysis were under that policy. JA1063-64. If the 2001 credit policy was “stricter,” though, then it was stricter for Black and White applicants (and the racial disparity actually grew under the 2006 policy, Br.19). Given that the policy changes were minute and Murphy said they made no difference, JA806, the district court abused its discretion in excluding Murphy’s testimony on this ground.

Freeman argues in its brief and Rule 28(j) letter, citing Kaplan, that the district court appropriately found Murphy’s analysis unreliable because it was not based on a representative sample. Resp.23-24;JA1064. But Murphy did not purport to sample; instead he included all applicants for whom he could match, with scientific certainty, their outcome and race/sex.

The district court offered no reasoning or authority for its cursory holding that Murphy’s omission of half of the branch offices—which pertains only to criminal applicants and Murphy’s 7/26/12 report—rendered his analysis unreliable. Freeman argues Murphy’s 1/17/13 report—which added applicants from missing branches—is still unreliable because criminal applicants from Mohave Valley, for instance, were underrepresented and the White fail rate was higher than the overall fail rate for either race. Resp.23-24. Even assuming the White fail rate was higher,^³ Freeman has not shown the statistical significance of this criticism, as Murphy sought to assess the overall impact of applying the same selection procedure in each office.

Echoing the district court, Freeman contends that Murphy’s purported coding, double-counting, and cherry-picking errors “skewed” the results. Freeman does not, however, contend that the errors caused the statistically significant disparate impact—and Murphy’s 4/16/13 report shows they did not. While plaintiffs bear the burden on reliability, Freeman had to do more than just point out errors: it needed to “explain how these errors had a[] substantial bearing” on the reports’ reliability. McReynolds, 349 F. Supp. 2d at 39 (report reliable where “purported flaws” did not cast doubt on correctness of methodology or materially affect results).

While Freeman characterizes the district court’s opinion as holding that external statistics (in Huebner’s and Murphy’s reports) had only “negligible probative value,” Resp.34, the court actually held they had no probative value. JA1069. The court erred. As EEOC argued, the Supreme Court and this Court have recognized the probative value of non-applicant flow data. Br.58-59. Freeman does not dispute that courts have relied on external statistics. Br.59. Nor does Freeman dispute that it intended its credit and criminal policy to deter applicants, enhancing the probative value of the external statistics in this case. Br.58.

External statistics as to Blacks and males were not rendered irrelevant merely because the applicant flow data failed to show a disparate impact on Hispanics. Resp.37. EEOC did not argue that external statistics alone established a prima facie case. Rather, EEOC argued the external statistics added to EEOC’s prima facie case.

Freeman also argues the external statistics lack probative value because they were not tailored to the qualified applicant pool for each Freeman branch or job grouping. Resp.34. As EEOC discussed, this Court held in EEOC v. Radiator Specialty, 610 F.2d 178, 185 (4th Cir. 1979), that defendants bear the burden of establishing special qualifications exist and are not possessed by the general population, where it is not manifest that such special qualifications exist. Br.59-60. Freeman responds that the policy identified special qualifications for “some jobs” but does not clarify which ones. Freeman suggests jobs requiring verified educational credentials involve special qualifications, Resp.35, but only a sub-set of Freeman’s jobs required education verification, and education requirements do not automatically constitute “special qualifications.” See United States v. Gregory, 871 F.2d 1239, 1244 & n.18 (4th Cir. 1989) (holding that the “relevant labor pool did not require any special skills” although deputy sheriff’s position required a high school degree, plus driver’s license, physical, and background investigation). Even if Freeman meets its burden of showing special qualifications are required, EEOC “should have an opportunity to adjust [its] statistical proof to reflect a labor pool base” with the requisite qualifications. Radiator, 610 F.2d at 185-86 (professional positions manifestly required special qualifications; clerical, managerial, and sales positions did not).

Next, Freeman argues EEOC’s reliance on Dothard, 433 U.S. 321, is misplaced because unlike height/weight requirements, “there are manifest reasons to suppose that individuals with criminal justice involvement” would have other disqualifying characteristics. Resp.35. Citing New York City Transit Auth. v. Beazer, 440 U.S. 568, 586 (1979), Freeman suggests courts should assume those with criminal histories disproportionately abuse drugs and alcohol. Freeman does not extend this argument to applicants with credit check issues, suggesting courts need not assume they also disproportionately abuse drugs and alcohol. But this criticism does not work for criminal applicants anyway because in Beazer the Court cited evidence that methadone users tended to suffer from drug/alcohol issues, and the record here contains no such evidence. Id. at 586 n.28, 575-76.

Freeman argues the external statistics failed to “satisfy[y] the EEOC’s prima facie case” because they encompassed arrests, arraignments, incarceration, and credit scores. Resp.36. EEOC is not arguing the statistics satisfied the prima facie case but that they are probative. As EEOC explained, Br.60-61, Huebner’s 3/11/13 report addressed convictions, and a factfinder could infer from the other statistics that Blacks/men had higher conviction rates.

EEOC met its burden of producing “sufficient, reliable evidence of disparate impact” to overcome summary judgment. See Bazemore v. Friday, 478 U.S. 385, 400 (1986) (plaintiff “need not prove discrimination with scientific certainty” but “by a preponderance of the evidence”). Murphy’s statistical analyses showed that Blacks failed the credit and criminal checks at statistically significant higher rates than Whites, and males failed the criminal checks at a statistically significant higher rate than women. JA1307(4/16/13 report); JA786(1/17/13 report); JA317(7/26/12 report). Huebner’s reports confirm the disparate impact. Thus, EEOC showed the checks “operate[d] as ‘built-in headwinds’” for Blacks and males seeking work. Griggs v. Duke Power, 401 U.S. 424, 432 (1971).

Freeman argues it is not required to “introduce a rebuttal expert report to negate disparate impact.” Resp.30. EEOC agrees. But EEOC’s argument is that the flaws identified by Freeman did not suffice to defeat EEOC’s prima facie case and that Freeman failed otherwise to rebut EEOC’s prima facie case on summary judgment. Br.50-52; see Gregory, 871 F.2d at 1244 n.21 (employer can rebut prima facie case “by offering his own statistics, including applicant flow data”).

EEOC argued that the court erred in applying the 300-day limitation period of §706 to this §707 action; refusing to apply the continuing violation doctrine; and holding the 300-day limit for the criminal claim ran from the formal letter notifying Freeman of the expanded investigation.

Freeman argues that §707(e)’s mandate that “all such actions” be conducted “in accordance with the procedures” of §706 refers to EEOC’s civil actions. Resp.54. But “all such actions” refers back to EEOC’s newly-granted authority to “investigate and act on a charge” alleging a pattern or practice. EEOC “acts” on a charge when, as under §706, it processes the charge; notifies the employer; investigates; issues a determination; and conciliates. These are all administrative steps. Freeman’s argument that “actions” naturally refers to EEOC’s civil actions is further undermined by the fact that EEOC’s litigation authority comes from §707(a)—which has no limitation period—not §707(e).

Freeman argues that, in any event, §706’s charge-filing period is really part of the administrative process, so §707 incorporates the limit. Resp.53. The issue on appeal, however, is not whether EEOC had a timely charge (it did), but whether the 300-day charge-filing limit of §706(e)(1), which applies to individuals, restricts the temporal scope of remedies EEOC can obtain under §707(a) when EEOC sues to address a pattern or practice of discrimination.

EEOC argued it makes no sense to apply a 300-day limitations period to EEOC’s §707 claims but not the Department of Justice’s §707 claims and that Congress did not intend such divergent practices. Freeman responds rhetorically that §707(e) “plainly evidences” Congress’ intent to do so, Resp.54, but Freeman cites no authority for its view other than §707(e). Nor does Freeman offer any reason why Congress would have wanted to hold state and local entities (including schools and fire departments) liable from the beginning of a pattern-or-practice of discrimination but would have limited the liability of private employers to the 300 days preceding a charge.

EEOC argued that a conundrum raised by subjecting its §707 suits to a 300-day limitation is how that limitation would work when EEOC proceeds without a charge. Br.64 n.8. Freeman responds, incorrectly, that EEOC cannot file suit without a charge. While EEOC typically files suit after a charge, §707(a) permits EEOC to sue “whenever” it “has reasonable cause to believe” there has been a pattern or practice of discrimination; no is charge required. 42 U.S.C. §§2000e-6(a),(c).

Freeman concedes some district courts have found that §707 has no limitations period but argues more courts have disagreed. Resp.56. The proper interpretation of the statute should be guided by Title VII’s text and purpose, not whether more district court decisions have piled up on Freeman’s side. Title VII’s text and purpose weigh against the 300-day limit.

Freeman misunderstands EEOC’s §706(g)(1) backpay argument. Resp.55. EEOC did not argue §707 incorporates §706(g)(1)’s two-year backpay limitation. Rather, EEOC asserted that if §707(e) incorporates the 300-day limit of §706—as Freeman contends—then §707 must necessarily incorporate the two-year backpay provision. Br.62. Such incorporation would effectively render the backpay provision a nullity. Because Freeman contends only hostile work environment claimants can recover for acts more than 300 days before a charge, only harassment victims could benefit from the two-year backpay provision. But harassment victims do not normally receive backpay. Freeman’s view of the law, then, would render §706(g)(1)’s backpay provision superfluous because no one could receive two years of back pay. In contrast, EEOC’s view of §707(e) as incorporating only the administrative procedures of §706 avoids the inconsistency between the 300-day limit and the two-year backpay provision. Similarly, EEOC’s view that the continuing violation doctrine applies to §706 pattern-or-practice actions is the only way to give effect to §706(g)’s backpay provision.

Although Freeman insists the continuing violation doctrine is a dead doctrine, Resp.58, Freeman fails to offer any explanation for the Supreme Court’s statement in National Railroad Passenger Corp. v. Morgan, 536 U.S. 101, 115 n.9 (2002), that it was not addressing private “pattern-or-practice” discrimination claims. Freeman also admits cases rejecting the continuing violation doctrine, including Williams v. Giant Food, 370 F.3d 423, 429 (4th Cir. 2004), did not concern an EEOC or class pattern or practice of discrimination. Resp.58. While Freeman contends Lewis v. City of Chicago, 560 U.S. 205 (2010), defeats EEOC’s argument, the Court did not address the continuing violation argument; rather, Lewis addressed “whether the practice” of using test results in a particular round of hiring “can be the basis for a disparate-impact claim at all.” Id. at 211.

Finally, Freeman fails to counter persuasively EEOC’s argument that if the 300-day limit applies, it ran for the criminal claim from the charge.

Freeman acknowledges EEOC can sue for claims that grow out of a reasonable investigation, and Freeman does not dispute that Title VII contains no “formal notice” provision requiring EEOC to notify employers of additional discrimination uncovered during an EEOC investigation. Freeman’s only authority to support its view that the 300-days runs from “formal notice,” rather than the charge, consists of unpublished district court decisions and EEOC v. General Electric, 532 F.2d 359, 368 (4th Cir. 1976), which concerns §706(g)’s backpay provision. Resp.60. The equities in this case also weigh in favor of starting the 300-day period from the charge, which explicitly mentioned Freeman’s criminal check policy.

Respectfully submitted,

P. DAVID LOPEZ

General Counsel

CAROLYN L. WHEELER

Acting Associate General Counsel

JENNIFER S. GOLDSTEIN

Acting Assistant General Counsel

__s/Anne Noel Occhialino ____

ANNE NOEL OCCHIALINO

Attorney

U.S. EQUAL EMPLOYMENT

OPPORTUNITY COMMISSION

Office of General Counsel

131 M St., NE, 5th Fl.

Washington, D.C. 20507

(202) 663-4724 (phone)

(202) 663-7090 (fax)

Annenoel.Occhialino@eeoc.gov

May 5, 2014

I certify that this brief complies with the type-volume limitation of Fed. R. App. P. 32(a)(7)(B) because it contains 6,991 words, excluding the parts of the brief exempted by Fed.R.App.P. 32(a)(7)(B)(iii).

I certify that this brief complies with the typeface requirements of Fed.R.App.P.32(a)(5) and the type style requirements of Fed. R. App. P. 32(a)(6) because it has been prepared in a proportionally spaced typeface using Microsoft Word 2003 in Georgia 14 point.

I certify that on May 5, 2014, I electronically filed this brief via the CM/ECF system, which will send notice to all counsel listed below. I also sent by overnight mail to the Court eight paper copies of the brief.

Counsel for Amicus EEAC, Nat’l Federation in Independent Business Legal Center & Retail Litigation Center