WESLEY, Circuit Judge:

Plaintiffs primary argument on appeal presents a question of first impression in our circuit: whether recourse to the pattern-or-practice evidentiary framework is appropriate in a suit against individual state officials brought pursuant to 42 U.S.C. § 1983 for intentional discrimination.

I. BACKGROUND

This case has as a backdrop prior litigation involving claims of racial discrimination at Elmira Correctional Facility ("Elmira"), a state maximum-security prison in Elmira, New York. See Santiago v. Miles, 774 F.Supp. 775, 782-88 (W.D.N.Y.1991). In 1986, black and Hispanic (jointly, "minority") inmates at Elmira commenced a class action for injunctive relief, alleging widespread racial discrimination at the facility in housing, job assignment, and the imposition of discipline. Id. at 777. After a bench trial, Judge Larimer found that the plaintiffs had proven a "pattern of racism" at Elmira. Id. On April 13, 1993, Judge Larimer issued a decision requiring, among other things, that the percentage of black and Hispanic inmates in certain "preferred" jobs, including jobs in the Elmira print shop, correspond to the percentage of black and Hispanic inmates in the general prison population.

At the time the suits here were filed, inmates employed in the Elmira print shop were paid an hourly wage, which ranged from sixteen cents to sixty-five cents per hour depending on the inmate's experience and expertise. In addition, inmates were eligible to receive an "incentive bonus" as a reward for good work. Civilian supervisors determined, in their discretion, whether a particular inmate merited promotion and higher pay. Similarly, these supervisors could recommend to the Elmira Program Committee — the entity tasked with assigning and removing inmates from various prison programs — that inmates be terminated from employment in the print shop. As a general matter, an inmate would be removed upon two requests.

In the print shop, inmates were directly supervised by civilian "Industrial Training Supervisors." The Industrial Training Supervisors reported to a general foreman, who in turn reported to an Assistant Industrial Superintendent and the Industrial Superintendent. The Industrial Superintendent answered to Elmira's Superintendent, among other officials.

In 1999, plaintiffs-appellants Jerry Reynolds and Khalib Gould (jointly, "plaintiffs"), inmates formerly employed in the Elmira print shop, filed pro se complaints alleging racial discrimination by civilian supervisors and prison administrators. Two other Elmira inmates, Anthony Mack and Joseph Ponder, commenced similar pro se actions in 2000.

Reynolds's pro se complaint asserted claims pursuant to 42 U.S.C. §§ 1981, 1983, 1985, and 1986 against Floyd Bennett, Elmira's Superintendent; David Barrett, Elmira's Industrial Superintendent; Dana Smith, Elmira's First Deputy Superintendent; Larry Pocobello, the Assistant Industrial Superintendent; Jack Rathbun, the print shop's general foreman; Terry Chamberlain, George Sarno, and Janice Kent, at the time all Industrial Training Supervisors; James Thompson, the chair of Elmira's Program Committee; and John Conroy, Director of Correctional Industry (jointly, "defendants").

Reynolds alleged that Barrett, Pocobello, Rathbun, Chamberlain, Sarno, and Kent demoted minority inmates more often than white inmates, confined minority inmates to low-paying positions, and unfairly docked the pay of minority inmates. Reynolds specifically complained about an incident in which Rathbun docked fifty-seven dollars from Reynolds's pay to reimburse the print shop for a poorly-run print job. Reynolds further alleged that minority inmates employed in the print shop had their pay docked at a much higher rate than white inmate-employees.

Gould's pro se complaint stated, among other things, claims pursuant to 42 U.S.C. §§ 1981, 1983, 1985, and 1986 against Pocobello, Barrett, Rathbun, Chamberlain, Kent, and Bennett. He alleged that they took adverse employment actions against him because of his race and retaliated against him for filing grievances.

In November 2000, the district court appointed counsel for the plaintiffs in all four actions. Counsel moved to consolidate the actions and file an amended complaint. Finding the proposed amended complaint deficient because it lacked detail as to the nature of each plaintiff's claims against each defendant, a magistrate judge directed plaintiffs to file a more detailed amended complaint by December 17, 2001. Instead, the parties agreed to consolidate the actions for the purpose of conducting discovery. They further agreed that no party would suffer prejudice if plaintiffs filed an amended complaint after discovery was completed. The magistrate judge approved the arrangement.

After conducting four years of discovery, plaintiffs sought leave to file an amended class action complaint on October 3, 2005. The proposed complaint defined the class as "all non-Caucasian inmates at [Elmira Correctional Facility] who were employed in the Print Shop from 1994 to the present, as well as all non-Caucasian inmates at [Elmira Correctional Facility] who were deterred from working within the Print Shop because of the discriminatory policies and/or practices set forth in this complaint." JA 64. In addition to claims pursuant to 42 U.S.C. §§ 1981, 1983, 1985, and 1986, the complaint claimed violations of Judge Larimer's order in Santiago v. Miles, 774 F.Supp. 775 (W.D.N.Y.1991), the New York State Human Rights Law, the New York State Constitution, and New York Civil Practice Law and Rule § 8601.

The proposed amended class action complaint asserted that racial discrimination was the "standard operating procedure in the Print Shop," that "incredible statistical disparities" existed between minority and non-minority inmates, and that minority inmates were evaluated more harshly, fired and demoted more often, and paid less than non-minority inmates. The complaint also claimed that the facially neutral subjective evaluation process used by the defendants, which gave them unfettered discretion when making employment decisions, had a disparate impact on minority inmates.

The proposed complaint provided several examples of purportedly discriminatory acts taken against plaintiffs. It stated that Reynolds had his bonus docked while white inmates did not, and that he "was issued several reprimands by defendants Chamberlain, Kent and Sarno in accordance with the discriminatory policies and practices in effect." JA 95. Similarly, the complaint stated that Gould was denied a promotion, demoted, and ultimately removed from the print shop on account of his race. The plaintiffs sought both injunctive relief and monetary damages.

In support of their motion to amend, plaintiffs appended the expert report of statistician Michael J. Guilfoyle, which purported to show, for the period between April 1994 and December 1999, that white inmates had longer average periods of employment in the print shop, were paid more than minority inmates, and were demoted less frequently than minority inmates. In Guilfoyle's view, the results of his study suggested that "there [was] a strong bias against non-white inmates working [in] the Elmira prison print shop when tenure, rate of pay[,] and demotions are examined." JA 157.

On July 1, 2008, with the motion to amend still pending, Judge Larimer ordered the parties to file summary judgment motions no later than August 25, 2008. After an extension of time was granted, defendants filed a summary judgment motion directed at plaintiffs' original pro se complaints on October 29, 2008. Plaintiffs opposed the motion and moved to certify the class action.

Plaintiffs argued that in the event leave to file an amended class action complaint was granted and a class certified, the motion for summary judgment against their individual complaints would be "irrelevant." They contended that the pattern-or-practice method of proof used in Title VII class actions could be employed in this § 1983 suit against individual defendants. Despite the fact that this Court has never applied the pattern-or-practice framework to hold individual state actors liable for intentional discrimination, plaintiffs did not give the district court the benefit of their reasoning as to why the framework was well-suited to that task.

On October 4, 2010, the district court granted summary judgment to defendants on Reynolds's and Gould's individual claims, denied the motion for class certification, and denied the motion for leave to amend the complaint. Reynolds v. Barrett, 741 F.Supp.2d 416 (W.D.N.Y.2010).1 The district court recognized that "[d]espite the variety of claims asserted, the § 1983 claims lie at the heart of these cases. And though § 1983 provides a vehicle by which to seek redress against state actors for a wide range of constitutional violations, it is plaintiffs' equal protection claims that form the core of their § 1983 claims." Id. at 425.

The district court analyzed plaintiffs' individual complaints under the McDonnell Douglas burden-shifting framework generally employed in assessing individual claims of disparate treatment under Title VII. Id. at 426-35. The court determined that defendants were entitled to summary judgment on both Reynolds's and Gould's individual claims of discrimination. Although the court noted Guilfoyle's statistical analysis, it concluded that Reynolds had not demonstrated that any adverse action was taken against him on account of his race. Id. at 427-29. Similarly, the court found no evidence from which a factfinder could reasonably conclude that race was a motivating factor in the adverse employment actions taken against Gould. Instead, the court determined that there was abundant evidence that Gould was subject to adverse employment actions "for nondiscriminatory reasons relating to his poor performance." Id. at 433.

Having granted summary judgment on plaintiffs' individual claims, the district court denied class certification and leave to amend. In particular, the court noted that "[a]t bottom, these cases present issues arising out of discrete acts of alleged discrimination and retaliation against two particular inmates." Id. at 444. As such, the court held, among other things, that plaintiffs had not met their burden of demonstrating the existence of questions of law or fact common to the proposed class. Id.

The district court then turned to the remaining issues related to plaintiffs' motion to file an amended complaint. As relevant here, it held that the proposed complaint's claims under New York law were barred by New York Corrections Law § 24(1).2 Similarly, it found that the proposed §§ 1981, 1985, and 1986 claims were not viable.3 Finally, the district court determined that defendants had not violated its prior order in Santiago.4 Id. at 445-46.

Reynolds and Gould timely appealed.

II. DISCUSSION

On appeal, plaintiffs principally contend that the district court should have examined the proposed amended class action complaint under the pattern-or-practice evidentiary framework and disparate impact theory of liability generally applicable in class actions brought pursuant to Title VII of the Civil Rights Act of 1964, 42 U.S.C. § 2000e et seq.5 Whether recourse to the pattern-or-practice framework is appropriate in a suit against individual state officials brought pursuant to 42 U.S.C. § 19836 for intentional discrimination is a question of first impression in our Circuit. Indeed, we have not found, nor have the parties cited to us, a case squarely addressing this issue.7

The gravamen of plaintiffs' proposed amended class action complaint is that there was a "pattern or practice" of racial discrimination in Elmira's print shop, as evidenced by "incredible statistical disparities within the [p]rint [s]hop between Caucasian and non-Caucasian employees" regarding promotion, demotion, discipline, and pay. The proposed class action complaint also asserts that Elmira's facially neutral policy of vesting in the print shop's civilian supervisors and other prison administrators "unfettered discretion" to make employment decisions resulted in a disparate impact on the print shop's minority inmate-employees.

As an initial matter, plaintiffs' novel attempt to impose disparate impact liability on defendants comes up short. Under certain circumstances, Title VII prohibits employment practices that have a disproportionately adverse effect on minorities. See 42 U.S.C. § 2000e-2(k); Ricci v. DeStefano, 557 U.S. 557, 129 S.Ct. 2658, 2672-73, 174 L.Ed.2d 490 (2009). Disparate impact claims "are concerned with whether employment policies or practices that are neutral on their face and were not intended to discriminate have nevertheless had a disparate effect on [a] protected group." Robinson v. Metro-North Commuter R.R. Co., 267 F.3d 147, 160 (2d Cir.2001).

But equal protection claims under § 1983 cannot be based solely on the disparate impact of a facially neutral policy. It is well established that "`[p]roof of racially discriminatory intent or purpose is required' to show a violation of the Equal Protection Clause." City of Cuyahoga Falls v. Buckeye Cmty. Hope Found., 538 U.S. 188, 194, 123 S.Ct. 1389, 155 L.Ed.2d 349 (2003) (quoting Vill. of Arlington Heights v. Metro. Hous. Dev. Corp., 429 U.S. 252, 265, 97 S.Ct. 555, 50 L.Ed.2d 450 (1977)); see Hayden v. Paterson, 594 F.3d 150, 162 (2d Cir.2010). Therefore, "a plaintiff pursuing a claimed violation of § 1981 or a denial of equal protection under § 1983 must show that the discrimination was intentional." Patterson v. Cnty. of Oneida, 375 F.3d 206, 226 (2d Cir.2004). Similarly, §§ 19858 and 19869 require "some racial, or perhaps otherwise class-based, invidiously discriminatory animus behind the conspirators' action." Griffin v. Breckenridge, 403 U.S. 88, 102, 91 S.Ct. 1790, 29 L.Ed.2d 338 (1971); see Soto-Padro v. Pub. Bldgs. Auth., 675 F.3d 1, 4 (1st Cir.2012). Thus, plaintiffs cannot proceed under a disparate impact theory of liability in their claims brought pursuant to §§ 1981, 1983, 1985, and 1986.

What remains, then, is plaintiffs' assertion that the Title VII pattern-or-practice framework10 may be applied to analyze discrimination claims brought pursuant to 42 U.S.C. § 1983 against individual state officials. We have never employed the framework in such a manner, and we decline to do so here.

It is true that we have previously observed that "[m]ost of the core substantive standards that apply to claims of discriminatory conduct in violation of Title VII are also applicable to claims of discrimination in employment in violation of ... the Equal Protection Clause." Patterson, 375 F.3d at 225; see also Annis v. Cnty. of Westchester, 136 F.3d 239, 245 (2d Cir. 1998); Jemmott v. Coughlin, 85 F.3d 61, 67 (2d Cir.1996). But each of those occasions involved individual claims of discrimination, and in each we applied either the McDonnell Douglas framework or a hostile work environment analysis. By urging this Court to find that the pattern-or-practice framework is applicable to § 1983 claims against individual state officials, plaintiffs seek a significant extension of our case law.

Employers, not individuals, are liable under Title VII. See Patterson, 375 F.3d at 226; Wrighten v. Glowski, 232 F.3d 119, 120 (2d Cir.2000) (per curiam). Title VII disparate treatment claims are of two types: (1) individual claims, which follow the familiar McDonnell Douglas burden-shifting framework, and (2) pattern-or-practice claims, which focus on allegations of widespread discrimination and generally follow the Teamsters burden-shifting framework. Robinson, 267 F.3d at 157 n. 3.

Under the McDonnell Douglas framework, a plaintiff establishes a prima facie case of intentional discrimination by showing that "(1) he is a member of a protected class; (2) he was qualified for the position he held; (3) he suffered an adverse employment action; and (4) the adverse action took place under circumstances giving rise to [an] inference of discrimination." Ruiz v. Cnty. of Rockland, 609 F.3d 486, 491-92 (2d Cir.2010). If the plaintiff establishes a prima facie case of discrimination, the burden shifts to the employer to come forward with a legitimate, nondiscriminatory reason for the adverse employment action. Id. at 492. If the employer does so, the burden then returns to the plaintiff to demonstrate that race was the real reason for the employer's adverse action. Id. Importantly, "[t]he ultimate burden of persuading the trier of fact that the defendant intentionally discriminated against the plaintiff remains at all times with the plaintiff." Tex. Dep't of Cmty. Affairs v. Burdine, 450 U.S. 248, 253, 101 S.Ct. 1089, 67 L.Ed.2d 207 (1981). Statistics alone do not suffice to establish an individual disparate treatment claim for a very good reason: the particular plaintiff must establish he was the victim of racial discrimination. See Hudson v. Int'l Bus. Mach. Corp., 620 F.2d 351, 355 (2d Cir.1980).11

In contrast to individual disparate treatment claims, "[p]attern-or-practice disparate treatment claims focus on allegations of widespread acts of intentional discrimination against individuals." Robinson, 267 F.3d at 158.12 To prevail on a pattern-or-practice claim, the plaintiffs must demonstrate that "intentional discrimination was the defendant's `standard operating procedure.'" Id. (quoting Teamsters, 431 U.S. at 336, 97 S.Ct. 1843).

A pattern-or-practice lawsuit proceeds in two phases. First, during the "liability phase," the plaintiffs are required to establish "a prima facie case of a policy, pattern, or practice of intentional discrimination against [a] protected group." Id. Unlike in individual disparate treatment claims, "[s]tatistics alone can make out a prima facie case of discrimination [in a pattern-or-practice suit] if the statistics reveal a gross disparity in the treatment of workers based on race." Id. (alterations and internal quotation marks omitted). Anecdotal evidence of discrimination may be highlighted to bring "the cold numbers convincingly to life." Teamsters, 431 U.S. at 339, 97 S.Ct. 1843.

Once the plaintiffs make out a prima facie case of discrimination in a pattern-or-practice case, the burden of production shifts to the employer to show that the statistical evidence proffered by the plaintiffs is insignificant or inaccurate. See id. at 360, 97 S.Ct. 1843. Typically, this is accomplished by challenging the "source, accuracy, or probative force" of the plaintiffs' statistics. Robinson, 267 F.3d at 159 (internal quotation marks omitted). If the defendant satisfies its burden of production, the trier of fact must then determine, by a preponderance of the evidence, whether the employer engaged in a pattern or practice of intentional discrimination. Id. If the plaintiffs succeed in proving a pattern or practice of discrimination, the court "may proceed to fashion class-wide injunctive relief." Id. Importantly, the plaintiffs are "not required to offer evidence that each person [who] will ultimately seek [individualized] relief was a victim of the employer's discriminatory policy" in order to prevail in the liability phase. Teamsters, 431 U.S. at 360, 97 S.Ct. 1843.

When plaintiffs seek individualized relief — i.e., back pay, front pay, or compensatory recovery — the case proceeds to the "remedial phase." Robinson, 267 F.3d at 159. During this phase, a particular plaintiff "need only show that he ... suffered an adverse employment decision and therefore was a potential victim of the proved class-wide discrimination." Id. (internal quotation marks and alteration omitted); see Teamsters, 431 U.S. at 361, 97 S.Ct. 1843. The employer then bears the burden of persuasion of demonstrating that the employee was subjected to an adverse employment action for legitimate, nondiscriminatory reasons. Robinson, 267 F.3d at 159-60; see Teamsters, 431 U.S. at 361, 97 S.Ct. 1843.

It bears noting that "[t]he heavy reliance on statistical evidence in a pattern-or-practice disparate treatment claim distinguishes such a claim from an individual disparate treatment claim proceeding under the McDonnell Douglas framework." Robinson, 267 F.3d at 158 n. 5. As this Court has recognized, the pattern-or-practice framework "substantially lessen[s] each class member's evidentiary burden relative to that which would be required if the employee were proceeding separately with an individual disparate treatment claim under the McDonnell Douglas framework." Id. at 159.

The McDonnell Douglas and Teamsters frameworks differ in important respects. However, both recognize that direct proof of intentional discrimination by an employer is hard to come by, and thus provide carefully calibrated burden-shifting structures designed to determine whether the employer intentionally discriminated against the plaintiffs. See Patterson v. McLean Credit Union, 491 U.S. 164, 186, 109 S.Ct. 2363, 105 L.Ed.2d 132 (1989).

As previously noted, proof of discriminatory intent is required to show a violation of the Equal Protection Clause. City of Cuyahoga Falls, 538 U.S. at 194, 123 S.Ct. 1389. Because neither a state nor a state official in his official capacity is a "person" within the meaning of § 1983, see Will v. Mich. Dep't of State Police, 491 U.S. 58, 71, 109 S.Ct. 2304, 105 L.Ed.2d 45 (1989), the requisite discriminatory intent must be held by the state official in his individual capacity. Thus, liability for an Equal Protection Clause violation under § 1983 requires personal involvement by a defendant, who must act with discriminatory purpose. See Ashcroft v. Iqbal, 556 U.S. 662, 676, 129 S.Ct. 1937, 173 L.Ed.2d 868 (2009). "[P]urposeful discrimination requires more than `intent as volition or intent as awareness of consequences.... It instead involves a decisionmaker's undertaking a course of action `because of, not merely in spite of, the action's adverse effects upon an identifiable group.'" Id. (quoting Personnel Adm'r of Mass. v. Feeney, 442 U.S. 256, 279, 99 S.Ct. 2282, 60 L.Ed.2d 870 (1979)).

The pattern-or-practice framework is ill-suited to the task of identifying which individual defendants engaged in purposeful discrimination in cases such as this one. Statistics proffered during the "liability phase" of a pattern-or-practice suit purport to demonstrate that a pattern of discrimination exists at an entity. In a Title VII case, these statistics can make out a prima facie case that the employer was engaged in a pattern or practice of discrimination. This is because an analysis of the collective acts of those who do the employer's bidding bespeak the employer's motivation.13

But statistics showing entity-level discrimination shed little light on whether a particular individual defendant engaged in purposeful discrimination. Just as statistics alone are insufficient to establish a prima facie case under the McDonnell Douglas framework, see Hudson, 620 F.2d at 355, statistics demonstrating employer-wide discrimination are insufficient to establish which individual defendants engaged in purposeful discrimination. Statistical disparities may be, and often are, attributable to a subset of actors — not to every actor who had an opportunity to discriminate. Cf. Wal-Mart Stores, Inc. v. Dukes, ___ U.S. ___, 131 S.Ct. 2541, 2555, 180 L.Ed.2d 374 (2011).

Thus, to import the pattern-or-practice framework into the Equal Protection context would substantially circumvent the plaintiffs' obligation to raise a prima facie inference of individual discriminatory intent. If "[s]tatistics alone [could] make out a prima facie case of discrimination," Robinson, 267 F.3d at 158, a § 1983 plaintiff could shift the burden to the defendant without any showing of individual discriminatory intent. Such a result would seem to contravene well-established precedent that "[p]roof of racially discriminatory intent or purpose is required to show a violation of the Equal Protection Clause" in a claim brought pursuant to § 1983. City of Cuyahoga Falls, 538 U.S. at 194, 123 S.Ct. 1389 (internal quotation marks omitted).

Plaintiffs in this case offer no authority for the proposition that a statistics-based evidentiary framework used to determine the liability of an entity under Title VII is appropriate to establish the liability of individual state officials under § 1983. They argue only that "individuals can engage in a pattern or practice of discrimination and there is no reason why such discrimination cannot be shown primarily through statistical proof." Reynolds Reply Br. 7. In their view, this is particularly true where the individual defendants "are the only actors whose decisions could have resulted in the statistical disparities." Reynolds Reply Br. 7-8. We disagree. Proffering statistical evidence that purports to show discrimination at an entity and naming as defendants all of the individuals who could possibly be responsible for such discrimination may support an inference that one or more of the named individual defendants committed acts of intentional discrimination. But such evidence provides little or no basis for discerning which individual defendants are responsible for the statistical disparities.

For example, the Guilfoyle report purports to show statistically significant racial disparities in the average employment tenure, rate of pay, and demotions of inmates in the Elmira print shop during the period between April 1994 and December 1999. Defendant Janice Kent began working as an Industrial Training Supervisor in the print shop in the fall of 1998. Even assuming that the Guilfoyle report supports the contention that discrimination was occurring in the print shop during the relevant period, the report says very little about whether Kent herself discriminated against minority inmates on account of their race. In other words, the statistics do not establish that discrimination was Kent's standard operating procedure. Unlike the statistics in a Title VII suit against an employer, the statistics proffered here do not place Kent in the position of "a proved wrongdoer," Teamsters, 431 U.S. at 359 n. 45, 97 S.Ct. 1843, and thus do not justify shifting the burden of persuasion to Kent to establish that every adverse employment action she took against a class member was animated by legitimate, nondiscriminatory reasons.

For the foregoing reasons, the pattern-or-practice framework is ill-suited to establish the liability of the individual defendants named in the proposed amended complaint.14 We therefore conclude that the district court did not err in declining to independently analyze plaintiffs' proposed class action amended complaint under the pattern-or-practice framework. We affirm the district court's denial of leave to amend and denial of class certification for substantially the same reasons stated by the district court.

III. CONCLUSION

The district court's order of October 4, 2010 granting summary judgment to defendants on plaintiffs-appellants' claims of individual discrimination and retaliation, denying leave to amend the complaint, and denying class certification is hereby AFFIRMED.