ORDER
The opinion and concurrence filed on June 27, 2017, are amended as follows:
1. At page 17 of the slip opinion, delete the sentence: < Perhaps, because of self-selection or other factors, the coaching population differs from the population studied. >.
2. At page 20 of the slip opinion, change <predict the racial effect> to < analyze the racial effect>, and delete con the NCAA’s applicant pool in particular >.
The concurring opinion is amended as follows:
1. At page 28 of the slip opinion, change <Co> to <Co.>.
2. At page 33 of the slip opinion, change cpredict the racial effect of individualized assessments on the NCAA’s applicant pool in particular > to < analyze the racial effect of individualized assessments >.
The panel has voted to deny the petition for panel rehearing. Judges Tallman and Friedland have voted to deny the petition for rehearing en banc and Judge Faber so recommends.
The full court has been advised of the petition for rehearing and no judge has requested a vote on whether to rehear the matter en banc. Fed. R. App. P. 35.
The petition for panel rehearing and the petition for rehearing en banc are DENIED. No further petitions for rehearing or rehearing en banc will be entertained.
Concurrence by Judge Faber
TALLMAN, Circuit Judge:
OPINION
Plaintiff Dominic Hardie appeals the district court’s entry of summary judgment in his suit against the National Collegiate Athletic Association (NCAA). Hardie, who is African American, alleges that the NCAA’s policy of excluding anyone with a felony conviction from coaching at NCAA-certified youth athletic tournaments violates Title II of the Civil Rights Act of 1964, 42 U.S.C. § 2000a(a), which prohibits racial discrimination in places of public accommodation. Hardie’s suit rests on a disparate-impact theory of Title II liability. We have never endorsed or rejected disparate-impact liability under Title II, and we need nót' decide this issue today. We hold that even if disparate-impact claims are cognizable under Title II, Hardie has not shown that an equally effective, less discriminatory alternative to the NCAA’s felon-exclusion policy exists, as he must do under the three-step analysis for disparate-impact claims set forth in Wards Cove Packing Co. v. Atonio, 490 U.S. 642, 109 S.Ct. 2115, 104 L.Ed.2d 733 (1989). We affirm summary judgment for the NCAA.
I
The NCAA is a voluntary, unincorporated association of over 1,200 colleges and [316]*316universities. One of the functions of the NCAA is to develop rules that govern intercollegiate athletics, including rules that limit recruitment of student-athletes. As part of their recruitment activities, coaches and other athletics staff from NCAA member schools attend nonscholastic1 youth athletic tournaments to scout potential recruits. Under NCAA rules, coaches and recruiters from Division I schools may attend nonscholastic tournaments only if the tournaments have obtained certification from the NCAA to verify that they are in compliance with NCAA guidelines. Without such attendance, the chances that players might be scouted and later recruited to play for an NCAA school are significantly diminished. This in turn affects the willingness of teams to play in uncertified tournaments and the profitability of private sponsors who organize these events.
The NCAA’s guidelines impose a number of requirements on tournament operators to ensure the safety of participants and preserve the integrity of college athletics recruiting. The guidelines restrict the number of games athletes may play in, for example, and they mandate that tournament operators obtain insurance and hire medical personnel. Importantly here, the guidelines require that tournament operators abide by the NCAA Participant Approval Policy. The Participant Approval Policy provides that anyone seeking to coach at an NCAA-certified nonscholastic tournament must submit to a criminal background check. Under the current version of the policy, anyone who has been convicted of a felony is automatically denied approval to coach in an NCAA-certified tournament. If a tournament operator fails to comply with NCAA guidelines, in-eluding the Participant Approval Policy, the tournament will not receive NCAA certification, and NCAA Division I coaches and recruiters may not attend the uncerti-fied tournament to scout for new talent.
The NCAA did not always ban anyone with a felony conviction from coaching at certified tournaments. The first Participant Approval Policy governing women’s basketball, adopted in 2006, disqualified only prospective coaches who had been convicted of a violent felony,2 a sex offense, a crime involving children, or a nonviolent felony if the nonviolent felony conviction was less than seven years old. The NCAA asserts, however, that the 2006 policy caused safety concerns and administrative difficulties. Certain crimes classified as nonviolent, including financial crimes, sports bribery, and possession of controlled substances, nonetheless raised significant safety and ethical concerns about coaches interacting with student-athletes. Additionally, differences between states’ classification of the same crimes led to inconsistent outcomes with respect to who was approved as a tournament coach and who was not.
In light of these challenges, the NCAA amended the Participant Approval Policy in 2011 to eliminate the violent-nonviolent felony distinction. Now, anyone with a felony conviction, no matter how old, is denied entry approval to coach. Any prior sex offense conviction, regardless of the charge level, and “active criminal cases” are also disqualifying. Coaches approved under the Participant Approval Policy may coach at NCAA-certified tournaments for two years, and then must reapply.
[317]*317In conformance with the amended Participant Approval Policy, Dominic Hardie was denied approval to coach at the 2013 MidSummer Night’s Madness Western Tournament, an annual NCAA-certified girls’ basketball tournament in San Diego. In 2001, Hardie had pled guilty and was convicted for possession of a controlled substance (cocaine), a felony in Texas. Hardie’s felony conviction had not affected his ability to coach in NCAA-certified tournaments before the Participant Approval Policy was amended. Under the pre-2011 version of the policy, Hardie had been able to coach because his only conviction was over seven years old and was for a nonviolent felony. But in 2012, when Hardie’s coaching approval expired and he reapplied, he was barred from coaching under the amended policy banning all felons. Hardie was allowed to attend the 2013 MidSummer Night’s Madness tournament as a spectator, but he could not participate from the coaches’ bench. Hardie alleges this prevented him from having personal contact with the student-athletes he coaches during the tournament games, which negatively affected his team members’ performance and opportunities to earn college athletics scholarships to NCAA schools.
After exhausting his administrative remedies without obtaining approval to coach, Hardie sued the NCAA in federal district court to enjoin enforcement of the Participant Approval Policy.3 Hardie alleges that the Participant Approval Policy violates Title II of the Civil Rights Act of 1964, 42 U.S.C. § 2000a(a), by denying him the full and equal enjoyment of a place of public accommodation.4 Originally, Hardie advanced both disparate-treatment and disparate-impact theories of Title II liability; on appeal, he now pursues only a disparate-impact theory. To prevail on his claim, Hardie must prove that the Participant Approval Policy has a “‘disproportionately adverse effect on minorities’ and [is] otherwise unjustified by a legitimate rationale.” See Tex. Dep’t of Hous. & Cmty. Affairs v. Inclusive Cmtys. Project, Inc., — U.S. —, 135 S.Ct. 2507, 2513, 192 L.Ed.2d 514 (2015) (quoting Ricci v. DeStefano, 557 U.S. 557, 577, 129 S.Ct. 2658, 174 L.Ed.2d 490 (2009)). He need not show that the NCAA acted with a “discriminatory intent or motive.” See id. Har-die alleges that “[t]he NCAA’s categorical bar” on coaches with felony convictions “falls disproportionately on African Americans like Hardie, who are more than three times as likely as white Americans to have suffered a felony conviction.”
To prove the Participant Approval Policy’s disparate impact, Hardie offers a report prepared by economist Marc Bendick. [318]*318Bendick surveyed 541 applicants who sought participant approval to coach at NCAA-certified nonscholastic youth basketball tournaments between 2011 and 2013. Among applicants surveyed, 46.5% of those approved under the Participant Approval Policy were African American, while 80.1% of those denied because of a felony conviction were African American. Bendick’s results thus show that African American applicants “were represented among felony denied applicants at 1.72 times their representation among approved applicants.”' A supplemental report Bendick prepared using “geocoding”5 produced similar results. In the supplemental report, African American applicants represented 40.3% of applicants denied because of a felony conviction, compared to 26.5% of approved applicants, meaning that African Americans were represented among felony-denied applicants at a rate 1.52 times higher than among approved applicants. Bendick states that the survey and geocoding results are statistically significant.6
The NCAA moved for summary judgment on Hardie’s Title II claim. The district court granted summary judgment for the NCAA, concluding that disparate-impact claims arfe not cognizable under Title II, Hardie timely appealed. We have jurisdiction under 28 U.S.C. § 1291.
II
“We review a district court’s grant of summary judgment de novo, and may affirm on any basis supported by the record.” Gordon v. Virtumundo, Inc., 575 F.3d 1040, 1047 (9th Cir. 2009). On review of a grant of summary judgment, we “must determine, viewing the evidence in the light most favorable to the nonmoving party, whether there are any genuine issues of material fact and whether the district court correctly applied the relevant 'sub-stantivé law.” Szajer v. City of Los Angeles, 632 F.3d 607, 610 (9th Cir. 2011) (quoting Universal Health Servs., Inc. v. Thompson, 363 F.3d 1013, 1019 (9th Cir. 2004)).
Ill
On appeal, the NCAA does not challenge Hardie’s argument that Title II encompasses disparate-impact claims. Instead, the NCAA asks us to affirm entry of summary judgment in its favor on either of two other grounds advanced below, assuming arguendo that disparate-impact claims are cognizable under Title II. First, the NCAA contends that it did not deny Har-die a privilege of a place of public accommodation because tournament operators, not the NCAA, enforce the Participant Approval Policy. Second, the NCAA argues that Hardie has failed to meet his burden under Wards Cove Packing Co. v. Atonio7 of showing in support of his disparate-impact claim that an equally effective, less discriminatory alternative to the felony ban under the Participant Approval [319]*319Policy exists. We affirm summary judgment for the NCAA on the latter ground.
A
Neither the Supreme Court nor we have decided whether disparate-impact claims are cognizable under Title II. A few courts have found that Title II authorizes disparate-impact claims, see Olzman v. Lake Hills Swim Club, Inc., 495 F.2d 1333, 1341-42 (2d Cir. 1974); Robinson v. Power Pizza, Inc., 993 P.Supp. 1462, 1464-66 (M.D. Fla. 1998), while others have rejected disparate-impact liability under Title II, see, e.g., Akiyama v. U.S. Judo Inc., 181 F.Supp.2d 1179, 1187 (W.D. Wash. 2002); LaRoche v. Denny’s, Inc., 62 F.Supp.2d 1366, 1370 n.2 (S.D. Fla. 1999). Several courts have declined to decide the issue altogether. See, e.g., Arguello v. Conoco, Inc., 207 F.3d 803, 813 (5th Cir. 2000); Jefferson v. City of Fremont, 73 F.Supp.3d 1133, 1145-46 (N.D. Cal. 2014) (citing cases).
We express no view today on whether Title II encompasses disparate-impact claims. Even if Title II authorized such claims, Hardie has not met his burden under Wards Cove of showing that an equally effective, less discriminatory alternative to the Participant Approval Policy exists.
B
In Wards Cove, the Supreme Court laid out a burden-shifting framework that applies to disparate-impact claims.8 490 U.S. at 658, 109 S.Ct. 2115. Burden shifting serves to limit disparate-impact liability “in key respects that avoid the serious constitutional questions that might arise .,. if such liability were irh-posed based solely on a showing of statistical disparities.” Inclusive Cmtys., 135 S.Ct. at 2522. Disparate-impact liability may only condemn practices or policies that are “artificial, arbitrary, and unnecessary.” Id. at 2524 (internal quotation mark omitted) (quoting Griggs v. Duke Power Co., 401 U.S. 424, 431, 91 S.Ct. 849, 28 L.Ed.2d 158 (1971)). The parties here agree that, assuming Title II encompasses disparate-impact claims, the Wards Cove framework would apply to Hardie’s claim.
Wards Cove burden shifting proceeds in three steps. First, a plaintiff must establish a prima facie case that the defendant’s challenged policy or practice has , a “significantly disparate impact on nop-whites.” Wards Cove, 490 U.S. at 658, 109 S.Ct. 2115. At the prima facie stage, the plaintiff must point to “the application of a specific or particular-... practice that has created the disparate impact under attack.” Id. at 657, 109 S.Ct. 2115. This “robust causality requirement ensures that ‘[r]acial imbalance ... does not, without more, establish a prima facie case of disparate impact’ and thus protects defendants from being held liable for racial disparities they did not create.” Inclusive Cmtys., 135 S.Ct. at 2523 (quoting Wards Cove, 490 U.S. at 653, 109 S.Ct. 2115).
Next, if a plaintiff makes out a prima facie case, “the case will shift to any business justification [defendants] offer for their use of these practices.” Wards Cove, 490 U.S. at 658, 109 S.Ct. 2115. “This phase of the disparate-impact case contains two components: first, a consideration of the justifications [a defendant] of[320]*320fers for his use of these practices; and second, the availability of alternative practices to achieve the same ... ends, with less racial impact.” Id.
At the justification step of Wards Cove burden shifting, the defendant must show that the “challenged practice serves, in a significant way, the legitimate ... goals of the [defendant].” Id. at 659, 109 S.Ct. 2115. The defendant’s practice need not be “essential” or “indispensable” to achieving its stated goal, but the relationship between the practice and its purpose must be more than “insubstantial.” Id. While the defendant must produce evidence that the practice serves legitimate ends, “[t]he ultimate burden of proving that discrimination against a protected group has been caused by a specific ... practice remains with the plaintiff at all times.” Id. (quoting Watson v. Fort Worth Bank & Tr., 487 U.S. 977, 997, 108 S.Ct. 2777, 101 L.Ed.2d 827 (1988)).
Finally, if the defendant provides a legitimate justification for the challenged practice, the plaintiff must demonstrate that an alternative practice (1) would “serve the [defendant’s] legitimate ,.. interests],” and (2) would not have a “similarly undesirable racial effect.” Id. at 660, 109 S.Ct. 2115 (third alteration in the original) (internal quotation mark omitted) (quoting Albermarle Paper Co. v. Moody, 422 U.S. 405, 425, 95 S.Ct. 2362, 45 L.Ed.2d 280 (1975)). The plaintiffs proposed alternative(s) must be “equally effective” as the defendant’s chosen policy at serving the defendant’s interest(s), taking into account “[fjactors such as the cost or other burdens” that alternative policies would impose. Id. at 661, 109 S.Ct. 2115 (alteration in original) (quoting Watson, 487 U.S. at 998, 108 S.Ct. 2777). A proposed alternative lacks a “similarly undesirable racial effect” if it results in “less disparate impact” compared to the challenged policy. Inclusive Cmtys., 135 S.Ct. at 2518; Ricci, 557 U.S. at 578, 129 S.Ct. 2658. This means plaintiffs must show not merely that an alternative policy would exclude fewer nonwhites, but that the alternative would reduce the overall racial disparity between whites and nonwhites. In some circumstances, an alternative policy may exclude fewer nonwhites but, because many more whites than nonwhites benefit under the alternative policy, the alternative actually exacerbates racial disparities. Alternative policies that are less restrictive than the challenged policy but do not result in “less disparate impact” will not withstand scrutiny at step three.9
Applying the Wards Cove framework here, Hardie relies on the Bendick report to establish a prima facie case of disparate impact. The Bendick report reveals that African Americans were significantly overrepresented—by a factor of 1.52 to 1.72— among felony-denied applicants compared to approved applicants. The NCAA does [321]*321not contest that Bendick’s results represent a significant racial disparity, and we agree. Furthermore, because Bendick surveyed only coaches who sought approval after the amended Participant Approval Policy took effect, and because he identified coaches who were denied specifically because of a prior felony conviction, the Bendick report also establishes a “causal connection” between the Participant Approval Policy’s blanket felon ban and the disproportionate effect on African American coaching applicants. See Inclusive Cmtys., 135 S.Ct. at 2523.
At Wards Cove step two, the NCAA contends that the Participant Approval Policy serves the NCAA’s interest in “protecting the safety of the children who participate in the tournaments and the integrity of the NCAA’s recruiting process and college athletics more generally.” For the most part, Hardie accepts that the NCAA’s proffered rationale .for the Participant Approval Policy is legitimate. We also agree.
The parties’ disagreement thus focuses on step three of the Wards Cove analysis. To satisfy his burden at step three, Hardie proposes two alternatives to the Participant Approval Policy: (1) the NCAA could revert to the pre-2011 version of the policy, which disqualified applicants with violent but not nonviolent felony convictions that were at least seven years old; or (2) the NCAA could conduct individualized assessments of applicants with felony convictions to determine if they would pose an unacceptable risk to the safety of tournament participants. We hold that Hardie has failed to show that either of his proposed alternatives would be both equally effective compared to, and less discriminatory than, the current policy. We address each of these alternatives in turn.
1.
Hardie first proposes that the NCAA revert to the version of the Participant Approval Policy that was in effect before 2011. The pre-2011 policy excluded applicants convicted of a violent felony, a sex offense, or a crime involving children, no matter how old, or a nonviolent felony conviction if the conviction was less than seven years old.
We find Hardie has failed to establish that the pre-2011 policy would be equally effective as the current policy in serving the NCAA’s legitimate interests. Hardie contends that the pre-2011 policy was proven equally effective because no documented safety incidents occurred during the few years when that version of the policy was in force. But the NCAA could have reasonably concluded that the level of risk under the pre-2011 policy was unacceptable, even if no tournament participants had yet been harmed'. Cf. El v. Se. Pa. Transp. Auth., 479 F.3d 232, 244 (3d Cir. 2007) (“In a broad sense, hiring policies ... ultimately concern the management of risk.”). Indeed, Hardie has not rebutted the NCAA’s assertion that under the pre-2011 policy, certain nonviolent felonies, such as financial crimes, possession of controlled substances, and sports bribery, posed unreasonable risks for the safety of student-athletes and the integrity of the recruiting process.
To compare the relative effectiveness of the pre-2011 policy and the current policy, Hardie submits a report by Dr. Kiminori Nakamura, a criminologist and expert on recidivism. The Nakamura report concludes that the probability that someone with a prior conviction will recidivate decreases the longer that person goes without committing another crime. At some point, known as the “redemption time,” the risk that someone with a prior conviction will reoffend becomes equal to the risk of [322]*322arrest for the general population.10 For people convicted of violent crimes, the redemption time is four to seven years; for drug and property crimes, the redemption time .is shortened to about four years. Na-kamura states that, because Hardie has not.-committed another criminal offense since. 2001, and because of his age, employment, and educational achievements, his risk of future arrest is in fact lower than that of the general population.
For several reasons, the Nakamura report does not establish that the pre-2011 Participant Approval Policy overall screens coaching applicants as effectively as the current policy. First, it does not attempt to quantify and compare the risk of future arrest for those permitted to coach under the current Participant Approval Policy with .the risk of those permitted under the pre-2011 version. Second, Nakamura acknowledges that, even after many years have passed, the risk of future arrest for someone who has a criminal record may remain' higher than the risk of future arrest for someone who has never been arrested. Hardie has offered no evidence to suggest that this difference in risk, even if small, is immaterial to achieving the NCAA’s interests. Particularly considering that NCAA-approved coaches work with minors, we cannot conclude on the record before us that this additional risk is insignificant.
We view this case as similar to El v. Southeastern Pennsylvania Transportation Authority, 479 F.3d 232 (3d Cir. 2007), in which the Third Circuit affirmed summary judgment for a local transit authority in a Title VII suit. Id. at .235. The transit authority in that case refused to hire anyone who had been convicted of a violent crime as a paratransit driver for disabled residents. In support of its policy, the transit authority offered unrebutted evidence that, even after many years, people convicted of violent crimes “are at least somewhat more likely than members of the general population to commit a future violent act.” Id. at 246. The transit authority’s expert testified that, even if the additional risk that a former violent felon posed “might be small,” “given the marked sensitivity of the paratransit position at issue, a small but extant difference is sufficient” -to justify the policy. Id. at 246-47. Analogizing to Hardie’s Title II claim here, Hardie has offered no evidence to suggest that the increased risk posed by coaches with prior felony convictions, even if small, is immaterial to protecting the safety of young athletes. Cf. N.Y.C. Transit Auth. v. Beazer, 440 U.S. 568, 587 n.31, 99 S.Ct. 1355, 59 L.Ed.2d 587 (1979) (holding that transit authority was justified in refusing to hire methadone users for “safety-sensitive” positions). Nor does he offer evidence to address the increased risk to preserving the integrity of college athletics from nonviolent crimes like sports bribery.11
We must also note that reverting to the pre-2011 policy would impose some increased administrative burden on the NCAA’s participant approval process. The NCAA submitted evidence that the 2011 amendment to the Participant Approval [323]*323Policy was motivated in part by administrative difficulties in distinguishing between violent and nonviolent crimes given states’ differing definitions of the same offenses. While costs and other administrative burdens are only a “factor” in the equal effectiveness analysis, see Wards Cove, 490 U.S. at 661, 109 S.Ct. 2115, and perhaps only a small one in this case, the increased administrative burden here is an additional factor that weighs slightly in the NCAA’s favor.
In sum, Hardie has not carried his burden at step three of the Wards Cove framework with respect to this proposed alternative. He has not adduced sufficient evidence from which a reasonable jury could conclude that excluding violent felons, but not nonviolent felons, from coaching would be equally effective at achieving the NCAA’s goals as the current policy. While we recognize that the plaintiffs burden at the alternatives stage is a demanding one, courts must take caution before displacing reasonable business judgments. Wards Cove, 490 U.S. at 661, 109 S.Ct. 2115 (“Courts are generally less competent than employers to restructure business practices.” (quoting Furnco Constr. Corp. v. Waters, 438 U.S. 567, 578, 98 S.Ct. 2943, 57 L.Ed.2d 957 (1978))). Because we hold that Hardie has failed to prove that the pre-2011 Participant Approval Policy would serve the NCAA’s legitimate interests as effectively as the current policy, we need not decide whether reverting to the pre-2011 policy would result in a lesser adverse racial effect.
As a second alternative, Hardie proposes that the NCAA conduct individualized assessments of applicants with a felony conviction to calculate the actual risks posed by an applicant. Hardie’s expert on human resources practices, Lester S. Rosen, states that individualized assessments may take into account factors such as “any mitigating circumstance about the offense, the age of the offense, ... past employment, educational achievements ;since the offense, and other signs of rehabilitation.”
Hardie’s individualized assessments alternative fails at step three of the Wards Cove analysis, because hé has put forward no evidence from which a reasonable jury could conclude that individualized assessments would have less disparate impact than the current Participant Approval Policy. None of Hardie’s experts analyze the expected racial impact of individualized assessments using the criteria Hardie proposes. The Rosen report only remarks generally that “[t]here is widespread recognition that a policy of automatically rejecting an applicant oh the basis of a felony conviction, without any consideration of the offense, the position in' question, the age of the offense, and evidence of rehabilitation, is both unfair and potentially violates Civil Rights laws.” Hardie also points to EEOC Guidelines that recommend employers adopt individualized assessments as a tool to avoid Title VII liability in the employment context. U.S. Equal Emp. Opportunity Comm’n, No. 915.002, EEOC Enforcement Guidance: Consideration of Arrest and Conviction Records in Employment Decisions Under Title VII of the Civil Rights Act of 1964 18 (2012). Neither the Rosen report nor the EEOC Guidelines analyze the racial effect of individualized assessments. Without more, we cannot say that Hardie has met his burden to show that individualized assessments would be a less discriminatory alternative to the current Participant Approval Policy.
Because neither of Hardie’s proposed alternatives to the Participant Approval Policy pass muster .at the final stage of [324]*324Wards Cove burden shifting, the NCAA is entitled to summary judgment.
C
Alternatively, the NCAA urges us to affirm summary judgment on the ground that it has not actually denied Hardie a “privilege! ] • • • of [a] place of public accommodation,” 42 U.S.C. § 2000a(a), because tournament operators, rather than the NCAA, actually enforce the Participant Approval Policy. Because we find that Hardie failed to produce sufficient proof at step three of the Wards Cove analysis, we need not reach this alternative basis for upholding summary judgment.
IV
We do not decide today whether Title II of the Civil Rights Act of 1964 encompasses disparate-impact claims. Even assuming arguendo that disparate-impact claims are cognizable under Title II, Hardie has not created a genuine issue of material fact that one of his proposed alternatives to the NCAA’s Participant Approval Policy would be both equally effective and less discriminatory. The NCAA is therefore entitled to summary judgment.
AFFIRMED.