Appeal from the United States District Court for the Western District of New York No. 18-cv-6591, Charles J. Siragusa, Judge. Plaintiffs are African-American men who were hired at a technology services provider before their offers of employment were revoked because of past criminal convictions. Citing national statistics showing that African Americans are arrested and incarcerated at higher rates than whites relative to their share of the national population, Plaintiffs brought a Title VII disparate impact class action against their would-be employer. The district court (Siragusa, J.) dismissed the complaint for failure to state a claim. We agree with that decision. While national statistics may be used to advance a disparate impact claim if there is reason to believe that the general population is representative of the qualified applicant pool subject to the challenged policy, Plaintiffs’ complaint suggests that the jobs they applied for required substantial educational and technical credentials, and Plaintiffs have provided no basis on which to presume that their proffered statistics are representative of the applicant pool in question. Since Plaintiffs have provided no other allegations to demonstrate that the challenged hiring policy has a disparate impact on African Americans, we AFFIRM the district court’s judgment. AFFIRMED. Judge Chin dissents in a separate opinion. RICHARD SULLIVAN, C.J. Facts are stubborn things, but statistics are pliable. As Mark Twain’s saying suggests, though statistics are often a helpful tool, they must be consulted cautiously. This lawsuit provides a case study as to why that is. Plaintiffs George Mandala and Charles Barnett have brought a Title VII disparate impact class action against Defendant NTT Data, Inc., arguing that the company’s alleged policy not to hire persons with certain criminal convictions has a disproportionately large effect on African-American applicants. To support that assertion, Plaintiffs rely on national statistics showing that, on average, African Americans are more likely to be arrested and incarcerated than whites. But the fact that such a disparity exists among the general population does not automatically mean that it exists among the pool of applicants qualified for the jobs in question — what is true of the whole is not necessarily true of its parts. In fact, because the complaint indicates that the positions that Plaintiffs applied for require certain educational and technical credentials, there is good reason to think that these national statistics are not representative of the qualified applicant pool. Consequently, Plaintiffs have set forth no allegations plausibly suggesting that the company’s hiring policy has a disparate impact on African Americans within the relevant hiring pool. We therefore AFFIRM the judgment of the district court (Siragusa, J.) dismissing the complaint. I. Background In early 2017, George Mandala applied for a position as a Salesforce Developer at NTT Data, Inc., a global information technology services provider.1 Impressed by his work experience and his answers to various “technical questions” during the interview process, Compl. 24, NTT offered Mandala a job as an “Application Software Development Senior Principal Consultant,” id. 27. But upon conducting a routine background check, the company discovered that Mandala had been convicted of a felony and quickly withdrew its offer of employment. When a member of NTT’s recruitment team broke the news to Mandala, she indicated that “NTT had a policy not to hire persons with felonies on their records.” Id. 33. Charles Barnett had a similar experience. NTT reached out to him in July 2017 about a “web developer” position on a project for the Kentucky Department of Education. Id. 38. On paper, Barnett appeared to be a strong candidate: he had relevant work experience, a “Masters of Science in Computer Science Technology[,] and an Associate degree in Applied Science/Computer Programming.” Id. 50. And after a few rounds of interviews, NTT offered him the job. But the company pulled that offer once it learned that Barnett had been convicted of several felonies. Though Barnett asked NTT to consider hiring him for other positions, he was informed that he was ineligible “because of his felony convictions.” Id. 48. So, in August 2018, Mandala and Barnett filed a putative class action complaint against NTT, alleging that the company’s hiring practices violate Title VII of the Civil Rights Act of 1964, as well as several New York State anti-discrimination laws. Specifically, they assert that NTT has a policy not to hire “individuals with certain criminal convictions including felonies (or similar criminal classifications),” id. 4, which Plaintiffs say is unlawful because it invariably disqualifies a disproportionate number of African-American applicants. To support this assertion, Plaintiffs point to numerous studies showing that “African Americans are arrested and incarcerated for crimes at higher rates than [w]hites, relative to their share of the national population.” Id. 52. This disparity is compounded, they say, by evidence suggesting that employers place additional weight on criminal history when an applicant is African American as opposed to white. Notably, however, the complaint contains no allegations about racial disparities in NTT’s existing workforce or the demographics of qualified applicants that NTT has rejected as a result of its hiring policy. It also fails to identify the precise contours of the policy itself — Plaintiffs equivocate as to whether the policy covers any prior criminal conviction or only felony convictions. A little less than a year after it was filed, the district court dismissed the complaint for failure to state a claim. See Mandala v. NTT Data, Inc., No. 18-cv-6591 (CJS), 2019 WL 3237361 (W.D.N.Y. July 18, 2019). The court concluded that the national statistics on which Plaintiffs rely are “inadequate to show a relationship between the pool of [NTT] applicants who are Caucasian versus African Americans and their respective rates of felony convictions.” Id. at *4. And without any remaining federal claims, the district court refused to exercise supplemental jurisdiction over Plaintiffs’ state law claims and dismissed their complaint in its entirety. Id. Plaintiffs now appeal that decision, arguing that the district court imposed an improperly high pleading standard, and that national arrest and conviction statistics are more than sufficient to state a plausible claim for relief under Title VII. II. Standard of Review We review de novo a district court’s decision to dismiss a complaint under Federal Rule of Civil Procedure 12(b)(6). Littlejohn, 795 F.3d at 306. A complaint will survive a motion to dismiss so long as it “contain[s] sufficient factual matter…to state a claim to relief that is plausible on its face.” Ashcroft v. Iqbal, 556 U.S. 662, 678 (2009) (internal quotation marks omitted). In making that assessment, we accept the plaintiff’s factual allegations as true and draw all reasonable inferences in her favor. Menaker v. Hofstra Univ., 935 F.3d 20, 30 (2d Cir. 2019). But while this plausibility pleading standard is forgiving, it is not toothless. It does not require us to credit “legal conclusion[s] couched as…factual allegation[s]” or “ naked assertions devoid of further factual enhancement.” Iqbal, 556 U.S. at 678 (internal quotation marks and brackets omitted). Nor are allegations that are “merely consistent with” liability enough to defeat a motion to dismiss. Id. (internal quotation marks omitted). Lastly, it bears mentioning that “we are free to affirm a decision [dismissing a complaint] on any grounds supported in the record, even if it is not one on which the trial court relied.” Thyroff v. Nationwide Mut. Ins. Co., 460 F.3d 400, 405 (2d Cir. 2006). III. Discussion A. The Elements of a Title VII Disparate Impact Claim Title VII of the Civil Rights Act of 1964, 42 U.S.C. §2000e et seq., as amended, prevents employers from discriminating against employees or job applicants based on race, color, religion, sex, or national origin. As originally enacted, “Title VII’s principal nondiscrimination provision held employers liable only for” intentional discrimination (known as “disparate treatment”). Ricci v. DeStefano, 557 U.S. 557, 577 (2009). But in Griggs v. Duke Power Co., the Supreme Court construed the statute to prohibit “not only overt discrimination but also practices that are fair in form, but discriminatory in operation” — that is, practices that have a “disparate impact.” 401 U.S. 424, 431 (1971). Griggs thus read Title VII to focus on “the consequences of employment practices, not simply the motivation” behind them.2 Id. at 432; see also M.O.C.H.A. Soc’y, Inc. v. City of Buffalo, 689 F.3d 263, 273 (2d Cir. 2012); Gulino v. N.Y. State Educ. Dep’t, 460 F.3d 361, 382 (2d Cir. 2006). Pursuing a disparate impact claim is often a complicated endeavor. Such claims “follow a three-part analysis involving shifting evidentiary burdens.” Gulino, 460 F.3d at 382 (citing 42 U.S.C. §2000e-2(k)(1)). The plaintiff “bears the initial burden of [making] a prima facie showing of disparate impact.” Id. This requires the plaintiff to “(1) identify a specific employment practice or policy; (2) demonstrate that a disparity exists; and (3) establish a causal relationship between the two.” Chin v. Port Auth. of N.Y. & N.J., 685 F.3d 135, 151 (2d Cir. 2012) (internal quotation marks and citations omitted). Unlike a disparate treatment claim, however, a disparate impact claim does not require the plaintiff to show that the defendant intended to discriminate against a particular group. See Ricci, 557 U.S. at 577-78; M.O.C.H.A., 689 F.3d at 273; see also Chaidez v. Ford Motor Co., 937 F.3d 998, 1006-07 (7th Cir. 2019). Once that prima facie showing is made, “the defendant has two avenues of rebuttal.” Gulino, 460 F.3d at 382. One approach is to undermine the plaintiff’s disparate impact or causal analysis. Id.; see also Watson v. Fort Worth Bank & Tr., 487 U.S. 977, 996 (1988). If the defendant is successful in doing so, that ends the matter. Alternatively, the defendant can concede that the identified policy has a disparate impact, but nevertheless defend it as “job related for the position in question and consistent with business necessity.” Ricci, 557 U.S. at 578 (quoting 42 U.S.C. §2000e-2(k)(1)(A)(i)); see also Gulino, 460 F.3d at 382. Should the defendant succeed in demonstrating the business necessity of the challenged policy, the burden then shifts back to the plaintiff, who has one last chance to prove her case. Namely, she must show that other methods exist to further the defendant’s legitimate business interest “without a similarly undesirable racial effect.” M.O.C.H.A., 689 F.3d at 274 (quoting Watson, 487 U.S. at 998); see Gulino, 460 F.3d at 382. B. The Applicable Pleading Standard Over the years, this three-step analysis has caused confusion about the pleading standard applicable to disparate impact claims. Are plaintiffs required to plead a prima facie case, or is the standard something lower? And if something lower, what is that lower threshold? The Supreme Court appeared to put this issue to rest in Swierkiewicz v. Sorema N.A., when it clarified that prima facie sufficiency is “an evidentiary standard, not a pleading requirement.” 534 U.S. 506, 510 (2002). The Court went on to explain that to survive a motion to dismiss, a Title VII complaint must contain only enough facts to give the defendant fair notice of the claim and the grounds upon which that claim rests. Id. at 512-14. And while Swierkiewicz employed the McDonnell Douglas burden shifting framework developed for disparate treatment claims, id. at 510 (citing McDonnell Douglas Corp. v. Green, 411 U.S. 792 (1973)), its analysis nonetheless helped inform disparate impact cases. After all, both types of discrimination claims use multi-part analyses that require the plaintiff to put forward a prima facie case of discrimination. But only seven years later, the Supreme Court cast doubt on Swierkiewicz’s vitality. In Ashcroft v. Iqbal, it held that mere notice pleading — the pleading standard underlying Swierkiewicz’s analysis — was inadequate, and that “[t]o survive a motion to dismiss, a complaint must contain sufficient factual matter, accepted as true, to state a claim to relief that is plausible on its face.” 556 U.S. at 678 (internal quotation marks omitted). Unsurprisingly, litigants were anxious to know what Iqbal meant for Title VII cases. Over a series of opinions, we clarified that Iqbal does not require a plaintiff to plead a prima facie case. Instead, it simply requires a plaintiff to “assert [enough] nonconclusory factual matter…to nudge [her] claim[] across the line from conceivable to plausible to proceed.” See EEOC v. Port Auth. of N.Y. & N.J., 768 F.3d 247, 254 (2d Cir. 2014) (internal quotation marks and alterations omitted); see also Vega v. Hempstead Union Free Sch. Dist., 801 F.3d 72, 84 (2d Cir. 2015); Littlejohn, 795 F.3d at 311; cf. NAACP v. Merrill, 939 F.3d 470, 477 (2d Cir. 2019) (recognizing that prima facie “is an evidentiary standard, not a pleading requirement” (internal quotation marks omitted)). And while many of those cases were again decided against the backdrop of the McDonnell Douglas framework, their reasoning is not limited to that context. Compare Vega, 801 F.3d at 84 (“[I]f Swierkiewicz survives for Equal Pay Act cases it surely survives for Title VII cases.”), with Littlejohn, 795 F.3d at 309 n.7 (acknowledging that “the Equal Pay Act does not fall under the burden-shifting framework of McDonnell Douglas”). In simple terms, then, this means that although a plaintiff need not plead a prima facie case, she must at least set forth enough factual allegations to plausibly support each of the three basic elements of a disparate impact claim. See Adams v. City of Indianapolis, 742 F.3d 720, 728, 733 (7th Cir. 2014).3 C. The Role of Statistics in Pleading a Disparate Impact Claim To nudge a disparate impact claim across the line from conceivable to plausible — and, indeed, to ultimately prove such a claim — plaintiffs typically rely on statistical evidence to show a disparity in outcome between groups. See Watson, 487 U.S. at 987; M.O.C.H.A., 689 F.3d at 273; Malave v. Potter, 320 F.3d 321, 325 (2d Cir. 2003). But not just any statistical assessment will do. At the prima facie stage, a plaintiff’s statistical analysis “must [demonstrate] that the disparity is substantial or significant, and must be of a kind and degree sufficient to reveal a causal relationship between the challenged practice and the disparity.” Chin, 685 F.3d at 151 (internal quotation marks omitted); see also Malave, 320 F.3d at 325-26; Brown v. Coach Stores, Inc., 163 F.3d 706, 712-13 (2d Cir. 1998). Naturally, that standard is relaxed at the pleading stage. For one thing, we do not require a plaintiff to prove in detail the methodological soundness of her statistical assessment to survive a motion to dismiss. See John v. Whole Foods Mkt. Gr., Inc., 858 F.3d 732, 737 (2d Cir. 2017). For another, we do not expect a complaint to supplement its statistical analysis with corroborating evidence. Cf. Littlejohn, 795 F.3d at 311; Malave, 320 F.3d at 326. But even at this early juncture, the statistics must plausibly suggest that the challenged practice actually has a disparate impact. See Adams, 742 F.3d at 733 (noting that while “basic” “statistical methods and comparisons” can be enough for a Title VII complaint to survive a motion to dismiss, those statistics must still “move the disparate-impact claims over the plausibility threshold”). This means that the statistical analysis must, at the very least, focus on the disparity between appropriate comparator groups. See Wards Cove Packing Co. v. Atonio, 490 U.S. 642, 651 (1989) (“Measuring alleged discrimination in the selection of accountants, managers, boat captains, electricians, doctors, and engineers…by comparing the number of nonwhites occupying these [skilled] jobs to the number of nonwhites filling [unskilled] cannery worker positions is nonsensical.”), superseded by statute on other grounds, 42 U.S.C. §2000e-2(k), as recognized in Tex. Dep’t of Hous. & Cmty. Affs. v. Inclusive Cmtys. Project, Inc., 576 U.S. 519 (2015). In other words, the statistical analysis must reveal disparities between populations that are relevant to the claim the plaintiff seeks to prove. See Smith v. Xerox Corp., 196 F.3d 358, 368 (2d Cir. 1999), overruled on other grounds by Meacham v. Knolls Atomic Power Lab., 461 F.3d 134 (2d Cir. 2006). For instance, it would make little sense to judge a hospital’s physician-hiring policies by looking at the effect those policies have on a population of high school graduates; most members of that group will be ineligible for the job, irrespective of the challenged policy, because they lack a medical degree. In a typical case concerning racially discriminatory hiring policies, the relevant comparison is between “the racial composition of the at-issue jobs and the racial composition of the qualified population in the relevant labor market.”4 Wards Cove Packing, 490 U.S. at 650 (internal alterations and quotation marks omitted); see also Smith, 196 F.3d at 368. Unfortunately, such figures are not always available, particularly before discovery. So we often allow plaintiffs to rely on other statistics that do not “conform to the preferred methodology” so long as they are probative of whether the challenged policy has a disparate impact on the qualified labor pool in question. Malave, 320 F.3d at 326; see also Wards Cove Packing, 490 U.S. at 651 (holding that “in cases where [the precise] labor market statistics [are] difficult if not impossible to ascertain,…[plaintiffs may rely on] certain other statistics…[so long as they] are equally probative”). One possible substitute is “figures for the general population.” Wards Cove Packing, 490 U.S. at 651 n.6 (internal quotation marks omitted); see also Joint Apprenticeship Comm., 186 F.3d at 119. But such national figures will not always be a viable alternative. General population statistics are a reliable surrogate only when there is reason to think that they “accurately reflect the pool of qualified job applicants” for the position in question. Malave, 320 F.3d at 326 (internal quotation marks omitted); see also Dothard v. Rawlinson, 433 U.S. 321, 330 (1977); EEOC v. Freeman, 961 F. Supp. 2d 783, 798 (D. Md. 2013) (“To use general population statistics to create an inference of disparate impact, the general populace must be representative of the relevant applicant pool.”), aff’d, 778 F.3d 463 (4th Cir. 2015). Otherwise, relying on them to show a disparate impact is a bit like relying on apples to study oranges. D. Plaintiffs Have Failed to State a Claim Plaintiffs allege that NTT has a “policy and practice of denying job opportunities to individuals with certain criminal convictions including felonies (or similar criminal classifications).”5 Compl. 4. This policy has a disparate impact on qualified African-American applicants, Plaintiffs say, because “African Americans are arrested and incarcerated for crimes at higher rates than [w]hites, relative to their share of the national population.” Id. 52. But while this reasoning is facially appealing, it ultimately succumbs to a fatal flaw. Plaintiffs provide no allegations to demonstrate that national arrest or incarceration statistics are in any way representative of the pool of potential applicants qualified for a position at NTT. All Plaintiffs offer is the conclusory and unsupported assertion that these figures are so stark that they must hold true for this (or any) segment of the population. But that is not a plausible — or, for that matter, logical — inference. Plaintiffs’ analysis fails to distinguish between group averages and total averages. To put it more plainly, it is error for Plaintiffs to simply presume that population-level statistics will accurately describe subgroups of that population. See Jones v. City of Lubbock, 730 F.2d 233, 235-36 (5th Cir. 1984) (Higginbotham, J., concurring) (criticizing the assumption that a subgroup is necessarily a “microcosm” of the broader population); Daye v. Cmty. Fin. Serv. Ctrs., LLC, 233 F. Supp. 3d 946, 1017 (D.N.M. 2017) (warning that the assumption “that what is true for the whole is true for a subset of the whole…can be dramatically wrong”). The danger behind this presumption becomes even more pronounced when there is reason to think that some characteristic unique to the subgroup is related to the statistic in question — in other words, when a confounding variable exists. A simple example of this pitfall would be to apply national height averages to certain subgroups of the population, say NBA players or horse-racing jockeys. Here, Plaintiffs have offered no allegations to suggest that the general population statistics on which they rely “might accurately reflect [NTT's] pool of qualified job applicants.” Malave, 320 F.3d at 326 (internal quotation marks and alterations omitted). And while that alone is fatal to their claim, the trouble does not end there. Plaintiffs’ claim concerns hiring policies governing what Plaintiffs allege to be skilled positions. Indeed, the positions’ titles alone — Salesforce Developer and web developer — reflect that they require at least some educational or technical experience that is not shared by the general population, and the complaint takes great pains to highlight Plaintiffs’ qualifications, plainly indicating that those qualifications are relevant to the jobs at the heart of this dispute. And it is beyond cavil that “[w]hen special qualifications are required to fill particular jobs, comparisons to the general population (rather than to the smaller group of individuals who possess the necessary qualifications) may have little probative value.” Hazelwood Sch. Dist. v. United States, 433 U.S. 299, 308 n.13 (1977); see also City of Richmond v. J.A. Croson Co., 488 U.S. 469, 501-02 (1989) (same); Burgis v. N.Y.C. Dep’t of Sanitations, 798 F.3d 63, 70 n.7 (2d Cir. 2015) (same). We therefore see no basis for using national statistics as a proxy for the qualified applicant pools for the developer positions at issue here. After all, it is not much of a stretch to imagine that arrest and conviction rates are negatively correlated with education (at least to some degree). So while Plaintiffs’ statistics show that African Americans are on average more likely to have been convicted of a crime than whites, that does not, without more, make it plausible that an African-American web developer with the educational and technical qualifications to work at NTT is more likely to have been convicted of a crime than his Caucasian counterpart.6 What, then, are Plaintiffs to do at the pleading stage before they have access to more granular data? For one, they could provide additional allegations to explain why their chosen national statistics are in fact likely to be representative of NTT’s qualified applicant pool. Cf. Dothard, 433 U.S. at 330. And if they are unable to do so, they could attempt to identify other publicly available information that could plausibly support a Title VII claim here. Of course, we are sensitive to the fact that Plaintiffs are undoubtedly working from an informational disadvantage at this early point in the proceedings. But that does not mean that they are free to rely on conclusory statistical inferences to force their way into discovery. In short, if a Title VII plaintiff intends to rely on national statistics to plead a disparate impact claim, she must explain why those statistics can plausibly be expected to hold true for the qualified applicant pool in question. And while we cannot provide a bright line rule for how to make such a showing, the statistics must at least be able to support the plaintiff’s claim without resorting to a statistical fallacy.7 IV. Conclusion For the reasons set forth above, we AFFIRM the district court’s judgment. CHIN, Circuit Judge, dissenting: As alleged in their complaint, plaintiffs-appellants George Mandala and Charles Barnett were offered jobs by defendant-appellee NTT Data, Inc. (“NTT”) — Mandala a position as a software development consultant in New York and Massachusetts, and Barnett a position as a website designer in Kentucky. NTT withdrew the offers, however, after learning that Mandala and Barnett had previously been convicted of crimes. NTT did so pursuant to its blanket policy of denying employment to job applicants based solely on the fact of a prior conviction, without considering individual circumstances such as the nature and circumstances of the offense, its age, whether the crime had any bearing on the applicant’s ability to perform the job sought, or evidence of rehabilitation and post-conviction good conduct. Mandala and Barnett, who are African American, brought this action below, alleging, inter alia, that NTT’s policy of rejecting applicants for employment because of prior convictions without individualized consideration has a disparate impact on African Americans, in violation of Title VII of the Civil Rights Act of 1964, as amended, 42 U.S.C. §2000e et seq. (“Title VII”). On a disparate impact claim, a plaintiff need not prove intent to discriminate; he is required only to prove that “a facially neutral employment policy or practice has a significant disparate impact” on members of a protected group. Brown v. Coach Stores, Inc., 163 F.3d 706, 712 (2d Cir. 1998); see Ricci v. DeStefano, 557 U.S. 557, 577-78 (2009); Griggs v. Duke Power Co., 401 U.S. 424, 431 (1971) (Title VII prohibits “not only overt discrimination but also practices that are fair in form, but discriminatory in operation”); 42 U.S.C. §2000e-2(k)(1)(A)(i)).8 Plaintiffs supported the allegations in their class action complaint with statistics showing that, on a national basis, “African Americans are arrested and incarcerated for crimes at higher rates than Whites, relative to their share of the national population.” Joint App’x at 15. The complaint cites statistics from the Federal Bureau of Investigation (the “FBI”), the U.S. Department of Justice, the U.S. Census Bureau, and the U.S. Equal Employment Opportunity Commission (the “EEOC”), as well as audit studies and other scholarly works. The district court granted NTT’s motion to dismiss the complaint pursuant to Federal Rule of Civil Procedure 12(b)(6). The district court rejected plaintiffs’ reliance on national statistics, holding that the statistics “do not indicate whether the individuals in the general population cited shared qualifications that would make them viable candidates for either of the positions offered to Plaintiffs.” Joint App’x at 70. The majority affirms, largely adopting that reasoning, holding that plaintiffs “provide no allegations to demonstrate that national arrest or incarceration statistics are in any way representative of the pool of potential applicants qualified for a position at NTT.” Maj. Op. at 17. I would vacate the dismissal of the complaint and remand for further proceedings. First, in my view, the district court did not properly apply the standards governing Rule 12(b)(6) motions to dismiss, and instead held plaintiffs to a higher standard. Second, I do not believe plaintiffs’ reliance on national statistics was improper, particularly at the pleadings stage and in light of the other detailed facts alleged. Third, based on the national statistics and other facts alleged in the complaint, I conclude that plaintiffs plausibly alleged that NTT’s policy had a disparate impact on African American job applicants in violation of Title VII. Accordingly, I dissent. I. In my view, the district court did not properly apply the standards applicable to Rule 12(b)(6) motions to dismiss. Although it acknowledged the “plausibility standard” in discussing the “general legal principles,” Joint App’x at 67-68, in determining whether the complaint stated a plausible claim, the district court relied on a decision setting forth the standards for proving — not pleading — a disparate impact case, Joint App’x at 69 (citing Robinson v. Metro-North Commuter R.R., 267 F.3d 147 (2d Cir. 2001)). Indeed, the district court wrote as follows: To plead a plausible claim of disparate impact, a plaintiff must allege ‘that a facially neutral employment policy or practice has a significant disparate impact.’ Brown v. Coach Stores, 163 F.3d 706, 712 (2d Cir. 1998). ‘To make this showing, a plaintiff must (1) identify a policy or practice, (2) demonstrate that a disparity exists, and (3) establish a causal relationship between the two.’ Robinson v. Metro-North Commuter R.R., 267 F.3d 147, 160 (2d Cir. 2001). Joint App’x at 69. The “showing” referred to in the language from Robinson, however, is “the prima facie showing of disparate impact” that a plaintiff must make as the first of three steps to prove a disparate impact claim. Robinson, 267 F.3d at 160. To plead a disparate impact claim, however, a plaintiff is not required to show, demonstrate, or identify anything — he need only allege a plausible claim. As we have held, for plaintiffs to “‘nudge[] their claims across the line from conceivable to plausible,’ they must ‘raise a reasonable expectation that discovery will reveal evidence’ of the wrongdoing alleged, ‘even if it strikes a savvy judge that actual proof of those facts is improbable.’” Citizens United v. Schneiderman, 882 F.3d 374, 380 (2d Cir. 2018) (quoting Bell Atlantic Corp. v. Twombly, 550 U.S. 544, 570, 556(2007); accord Arista Records, LLC v. Doe 3, 604 F.3d 110, 120 (2d Cir. 2010) (plausibility “simply calls for enough fact to raise a reasonable expectation that discovery will reveal evidence of illegal[ity]“) (quoting Twombly, 550 U.S. at 556). By concluding that plaintiffs were required to make the “showing” discussed in Robinson — “the prima facie showing of disparate impact” required to prove discrimination or survive a summary judgment motion — the district court erred, for in employment discrimination cases it is now settled that at the pleadings stage a plaintiff is not required to plead (much less “show” or “prove”) a prima facie case. Robinson, 267 F.3d at 160. Rather, as we have held, “a plaintiff has a ‘minimal burden’ of alleging facts ‘suggesting an inference of discriminatory motivation.’” Vega v. Hempstead Union Free Sch. Dist., 801 F.3d 72, 85 (2d Cir. 2015) (quoting Littlejohn v. City of N.Y., 795 F.3d 297, 310 (2d Cir. 2015)). While Vega and Littlejohn were disparate treatment cases, their reasoning applies here: Mandala and Barnett had the “minimal burden” of alleging facts “suggesting” an “inference” of a disparate impact based on race — facts that plausibly give rise to an inference that an employment practice “has the effect of denying members of a protected class equal access to employment opportunities.” M.O.C.H.A. Soc’y, Inc. v. City of Buffalo, 680 F.3d 263, 273 (2d Cir. 2012); see also Vega, 801 F.3d at 87 (“On a motion to dismiss, the question is not whether a plaintiff is likely to prevail, but whether the well-pleaded factual allegations plausibly give rise to an inference of unlawful discrimination, i.e., whether plaintiffs allege enough to ‘nudge[ ] their claims across the line from conceivable to plausible.’”) (citations omitted); see also Swierkiewicz v. Sorema N.A., 534 U.S. 506, 511-12 (2002) (“under a notice pleading system, it is not appropriate to require a plaintiff to plead facts establishing a prima facie case”).9 As discussed further below, the national statistics and other facts alleged by plaintiffs were sufficient, in my view, to meet this minimal burden, as the statistics and other factual assertions suggest an inference of a disparate impact. Instead of measuring plaintiffs’ allegations by this “minimal” standard, the district court engaged in the kind of analysis appropriate only at trial: weighing and rejecting the evidence and finding that the national statistics were “inadequate to show a relationship between the pool of applicants who are Caucasian versus African Americans and their respective rates of felony convictions.” Joint App’x at 70-71. Instead of drawing the reasonable inferences in favor of plaintiffs, the district court rejected the national statistics, requiring plaintiffs to present more: the precise statistical evidence they would use to prove their claim. This was error. As the Seventh Circuit has observed, “‘[d]isparate impact plaintiffs are permitted to rely on a variety of statistical methods and comparisons to support their claims,’” and “‘[a]t the pleading stage, some basic allegations of this sort will suffice.’” Chaidez v. Ford Motor Co., 937 F.3d 998, 1007 (7th Cir. 2019) (citation omitted); see also, e.g., John v. Whole Foods Mkt. Group, Inc., 858 F.3d 732, 737 (2d Cir. 2017) (“At the pleading stage, [plaintiff] need not prove the accuracy of [a statistical study's] findings or the rigor of its methodology; he need only generally allege facts that, accepted as true, make his alleged injury plausible.”); Brown v. City of N.Y., No. 16 Civ. 1106, 2017 WL 1102677, at *6 (E.D.N.Y. Mar. 23, 2017) (“[S]tatistics that may ultimately prove insufficient can nevertheless support a plausible inference of disparate impact on a motion to dismiss.”). II. Next, plaintiffs’ reliance on national statistics was appropriate, and surely so at the pleadings stage of the litigation. As the majority acknowledges, in some circumstances national statistics can be probative of whether a challenged policy has a disparate impact. The cases have so held. In Dothard v. Rawlinson, for example, the Supreme Court observed that “[t]here is no requirement…that a statistical showing of disproportionate impact must always be based on analysis of the characteristics of actual applicants.” 433 U.S. 321, 330 (1977). There, the plaintiffs were permitted to rely on national height and weight data for men and women, even though the defendant argued that the only relevant data was that of the applicant pool for corrections officer positions in Alabama. Id. at 329-30. Likewise, in Malave v. Potter, this Court rejected the district court’s holding, on a summary judgment motion, that plaintiffs were required to provide statistical information as to “the applicant pool or the eligible labor pool.” 320 F.3d 321, 325-26 (2d Cir. 2003). We rejected the district court’s “adoption of a rule that the lack of statistical information as to an applicant pool always renders it impossible to establish a prima facie disparate impact case.” Id. at 327. We remanded for the district court to determine, inter alia, “the most appropriate labor pool,” and reminded the district court of “the Supreme Court’s teaching that ‘statistics come in infinite variety and…their usefulness depends on all of the facts and surrounding circumstances.’” Id. (quoting Watson v. Fort Worth Bank & Trust, 487 U.S. 977, 994 (1988)); see also EEOC v. Joint Apprenticeship Comm. of Joint Indus. Bd. of Elec. Indus., 164 F.3d 89, 97 (2d Cir. 1998) (at summary judgment, plaintiff’s prima facie case can include “studies based on general population data”). And in Wards Cove Packing Co. v. Atonio, the Supreme Court recognized that “where ‘figures for the general population might…accurately reflect the pool of qualified candidates,’…we have even permitted plaintiffs to rest their prima facie cases on such statistics as well.” 490 U.S. 642, 651 n.6 (1989) (citation omitted). The Court noted that “where such labor market statistics [of the racial composition of qualified persons in the labor market and the persons holding at-issue jobs] will be difficult if not impossible to ascertain,…certain other statistics — such as measures indicating the racial composition of ‘otherwise-qualified applicants’ for at-issue jobs — are equally probative for this purpose.” Id. at 651. Hence, general population statistics may be relevant in a disparate impact case, depending on the facts and surrounding circumstances. It was premature for the district court to conclude, on a motion to dismiss for failure to state a claim, that the national statistics were not relevant. The complaint alleges that NTT is a “global” information technology services company with some 18,000 employees in North America and over twenty offices in the United States, and it contends that NTT’s policy has a national disparate impact. The specific allegations support the assertion of a national impact: Mandala applied for a Salesforce Developer position while residing in Rochester, New York, and he was offered a position as an Application Software Development Senior Principal Consultant in Wellesley, Massachusetts (to work remotely). Barnett applied for a web developer position while residing in Frankfort, Kentucky, and was offered employment designing websites for the Kentucky Department of Education. Both were impacted by NTT’s policy even though they lived in different parts of the country and applied for different types of work. After Barnett’s offer was withdrawn, he “sought to apply for other positions with NTT,” but NTT would not consider him for any position because of its policy. Joint App’x at 14. And, the complaint alleges, NTT’s policy “systematically eliminates qualified African American applicants based on their race, color or national origin,” Joint App’x at 8, throughout the United States. In light of these allegations, it was surely premature for the district court to conclude that national statistics had no probative value. Given the alleged impact of the policy in such distant areas as Rochester, New York, Wellesley, Massachusetts, and Frankfort, Kentucky, the alleged impact of the policy on different jobs (salesforce developer, software development consultant, software developer, and “other positions with NTT,” Joint App’x at 14), and the alleged impact of the policy nationally, the national statistics were relevant and should not have been rejected by the district court at the pleadings stage. In rejecting the national statistics, the district court faulted plaintiffs for not “alleg[ing] facts showing that [NTT]‘s facially-neutral policy of not hiring convicted felons is related to the statistical disparity in the numbers of African-Americans arrested and convicted of crimes in proportion to their representative numbers in the pool of qualified applicants for [NTT]‘s positions.” Joint App’x at 70. But the district court did not identify what that pool is, see Malave v. Potter, 320 F.3d at 327 (remanding for district court to consider “the most appropriate labor pool”), and information as to NTT’s “pool of qualified job applicants” — whatever that pool is — is not something that would be readily available without discovery, see Wards Cove Packing Co., 490 U.S. at 651 (noting that where statistics as to racial composition of qualified persons in labor market and persons holding at-issue jobs would be “difficult if not impossible to ascertain,” other statistics, including national statistics, could be considered). Accordingly, it was error, in my view, for the district court to conclude, on a motion to dismiss, before plaintiffs had the opportunity to seek discovery, that the national statistics were of no value.10 III. Finally, plaintiffs plausibly alleged that NTT’s policy had a disparate impact on African American job applicants in violation of Title VII. First, plaintiffs alleged that NTT’s policy had an adverse impact on them personally. Mandala and Barnett both received offers of employment from NTT. The offers were withdrawn, however, once NTT learned that they had prior convictions. NTT never asked Mandala and Barnett for information about their convictions or rehabilitation or post-conviction conduct. Barnett, for example, had obtained two degrees, including a masters in computer science technology — after his conviction. He also worked for four years for the Commonwealth of Kentucky doing information technology and administrative work — after his conviction. But these facts were not considered by NTT because of its blanket policy. Mandala and Barnett were clearly qualified for the NTT positions — they were offered employment — but the policy barred them from employment without any individualized consideration of the circumstances of their convictions, the relationship between their criminal history and their ability to perform the jobs, or their efforts to rehabilitate and post-conviction conduct.11 Second, the national statistics set forth in the complaint show that “African Americans are arrested and incarcerated for crimes at higher rates than Whites, relative to their share of the national population.” Joint App’x at 15. Those statistics include: as of 2010, 40 percent of prisoners in the United States were African American, while African Americans represented only 13 percent of the overall U.S. population (Prison Policy Initiative study)12; some 26.9 percent of arrests are of African Americans, double their percentage of the general population (FBI and Census statistics); projections based on recent trends in incarceration estimate that one out of every three African American males born today will go to prison, compared to just one out of every seventeen White males (Prison Journal study)13; audit studies conducted by researchers at Harvard and Princeton found that African Americans with criminal records were particularly disadvantaged in the job market compared to Whites with criminal records (scholarly journals); and the Department of Justice found that Blacks are arrested and convicted at higher rates than Whites, leading the EEOC to conclude that “[n]ational data supports a finding that criminal record exclusions have a disparate impact based on race and national origin.” Joint App’x at 15.14 Based on these national statistics, plaintiffs plausibly alleged, at least as a general matter, that a policy that precludes employment opportunities based solely on an individual’s criminal record, without consideration of the circumstances of the conviction or the individual’s efforts at rehabilitation, has a disparate impact on African Americans seeking employment. Finally, the question is whether the national statistics provide support, at the motion to dismiss stage, for plaintiffs’ claim of disparate impact discrimination. They do. They are a logical starting point for a disparate impact analysis, as plaintiffs are alleging a national impact. It may be that statistics as to a specific applicant pool, e.g., salesforce developers in upstate New York, software consultants in Massachusetts, software developers in Kentucky, “other positions” with NTT, or — in the majority’s words — applicants “representative of the pool of potential applicants qualified for a position at NTT,” Maj. Op. at 17, would be more accurate. But the absence of such statistics at the motion to dismiss stage should not be fatal to plaintiffs’ claims, as the appropriate applicant pool likely cannot be defined until after discovery, when more details about NTT’s job requirements and applicant pools would become available. Moreover, some of this information would only be in NTT’s possession and likely is not publicly available.15 And even if specific applicant pool statistics would be more precise, general population statistics would still be relevant and a logical part of the analysis. See Williams, 2016 WL 4149987, at *5 (“Although the Court agrees with Defendant that the [national] statistics are not tailored to the New Jersey counties in which Defendant does business, the Court finds that they lend sufficient support to Plaintiff’s allegations to survive the Motion to Dismiss.”). It may turn out, after the parties have completed discovery and the appropriate applicant pool is determined, that the probative value of the national statistics is low, but it is certainly plausible that the racial disparities in the general statistics will not disappear once the statistics for particular job markets are ascertained. It may turn out, I suppose, that NTT’s applicant pool differs to such an extent from the general population that the national statistics will be entirely irrelevant, but it is certainly plausible, at the pleadings stage, that NTT’s applicant pool will not differ significantly from the general population. It is unlikely, in my view, that the racial disparities in arrest and incarceration rates that exist nationally and in every state somehow disappear at NTT. * * * * * As the statistics show, there are significant racial disparities in arrest, conviction, and incarceration rates in this country. As scholars and the EEOC have recognized, criminal history screens can have a substantial adverse disparate impact based on race,16 and, as discussed above, a number of courts have denied motions to dismiss disparate impact claims using general population statistics to challenge such policies, concluding that plaintiffs had plausibly stated a disparate impact claim under Title VII. Some states, including New York, recognize the harm that blanket criminal history screens can cause, forbidding companies from denying employment solely because a job applicant has a criminal record, and instead requiring employers to engage in an individualized consideration. If the facts alleged in the complaint are true, then Mandala and Barnett are vivid examples of the adverse impact an absolute convictions bar can have on individuals generally — and African Americans in particular — seeking employment. As the Eighth Circuit recognized many years ago: We cannot conceive of any business necessity that would automatically place every individual convicted of any offense, except a minor traffic offense, in the permanent ranks of the unemployed. This is particularly true for blacks who have suffered and still suffer from the burdens of discrimination in our society. To deny job opportunities to these individuals because of some conduct which may be remote in time or does not significantly bear upon the particular job requirements is an unnecessarily harsh and unjust burden. Green v. Missouri Pac. R. Co., 523 F.2d 1290, 1298 (8th Cir. 1975). These observations still hold true today. I respectfully dissent.
