(dissenting).
Because I am not persuaded that the majority’s view represents the state of the law today, I respectfully dissent.
I. THE LAW OF THE CASE
Under the doctrine of the “law of the case,” a decision on an issue of law made by the court at one stage of a case becomes a binding precedent to be followed in successive stages of the same litigation except in unusual circumstances. See Abbadessa v. Moore Business Forms, Inc., 987 F.2d 18, 22 (1st Cir.1993); EEOC v. Trabucco, 791 F.2d 1, 2 (1st Cir.1986). It is well established, however, that a decision of the Supreme Court, that is rendered between two appeals and is irreconcilable with the decision on the first appeal, must be followed on the second appeal. See Linkletter v. Walker, 381 U.S. 618, 627, 86 S.Ct. 1731, 1736-37, 14 L.Ed.2d 601 (1965); Metcalf & Eddy, Inc. v. Puerto Rico Aqueduct and Sewer Auth., 945 F.2d 10, 12 (1st Cir.1991), rev’d on other grounds, 506 U.S. 139, 113 S.Ct. 684, 121 L.Ed.2d 605 (1993); Young v. Herring, 917 F.2d 858 (5th Cir.1990); Fogel v. Chestnutt, 668 F.2d 100, 109 (2d Cir.1981), cert. denied, 459 U.S. 828, 103 S.Ct. 65, 74 L.Ed.2d 66 (1982). I believe that we face such a situation in the instant case.
A. Adarand and Metro Broadcasting
At the time of Cohen v. Brown University, 991 F.2d 888 (1st Cir.1993) (Cohen II), the standard intermediate scrutiny test for discriminatory classifications based on sex required that “a statutory classification must be substantially related to an important government objective.” Clark v. Jeter, 486 U.S. 456, 461, 108 S.Ct. 1910, 1914, 100 L.Ed.2d 465 (1988); see also Mississippi Univ. for *189Women v. Hogan, 458 U.S. 718, 723-24, and n. 9, 102 S.Ct. 3331, 3335-36 and n. 9, 73 L.Ed.2d 1090 (1982); Mills v. Habluetzel, 456 U.S. 91, 99, 102 S.Ct. 1549, 1554-55, 71 L.Ed.2d 770 (1982); Craig v. Boren, 429 U.S. 190, 197, 97 S.Ct. 451, 456-57, 50 L.Ed.2d 397 (1976); Mathews v. Lucas, 427 U.S. 495, 505-06, 96 S.Ct. 2755, 2762-63, 49 L.Ed.2d 651 (1976). As was also the casé under strict scrutiny review prior to Adarand Constructors, Inc. v. Pena, — U.S. -, 115 S.Ct. 2097, 132 L.Ed.2d 158 (1995), however, courts applying intermediate scrutiny sometimes allowed “benign” gender classifications on the grounds that they were a “reasonable means of compensating women as a class for past ... discrimination.” Ronald D. Rotunda & John E. Novack, 3 Treatise on Constitutional Law § 18.23, at 277; see Califano v. Webster, 430 U.S. 313, 317, 97 S.Ct. 1192, 1194-95, 51 L.Ed.2d 360 (1977) (allowing women to compute certain social security benefits with a more favorable formula than could be used by men); Lewis v. Cowen, 435 U.S. 948, 98 S.Ct. 1572, 55 L.Ed.2d 797 (1978) (summary affirmance of a district court decision upholding a provision of the Railroad Retirement Act that allowed women to retire at age 60 while men could not retire until age 65).
In Cohen II, we applied precisely this type of benign-classification analysis to what we viewed to be benign gender discrimination by the federal government. Although Cohen II, in its brief discussion of the equal protection issue, does not specify the precise standard it used, the court stated that “even if we were to assume ... that the regulation creates a gender classification slanted somewhat in favor of women, we would find no constitutional infirmity.” Cohen II, 991 F.2d at 901. Note that the focus is on the government’s ability to favor women in this context, rather than on an “important government objective,” suggesting that the court considered the issue to be one of benign discrimination. Indeed, no governmental interest is even identified in Cohen II. Furthermore, both of the eases cited by the court in Cohen II are cases in which a suspect classification was allowed because it was judged benign, see id. at 901 (citing Metro Broadcasting Inc. v. FCC, 497 U.S. 547, 110 S.Ct. 2997, 111 L.Ed.2d 445 (1990) (race); Califano v. Webster, 430 U.S. 313, 97 S.Ct. 1192, 51 L.Ed.2d 360 (1977) (sex)).
Cohen II’s assumption that a regulation slanted in favor of women would be permissible, Cohen II 991 F.2d at 901, and by implication that the same regulation would be impermissible if it favored men, was based on Metro Broadcasting, which held that benign race-based action by the federal government was subject to a lower standard than non-remedial race-based action. See Metro Broadcasting, 497 U.S. at 564, 110 S.Ct. at 3008. Specifically, the Supreme Court announced that
benign race-conscious measures mandated by Congress are constitutionally permissible to the extent that they serve important governmental objectives within the power of Congress and are substantially related to achievement of those objectives.
Id. at 565, 110 S.Ct. at 3026 (emphasis added). Although Metro Broadcasting explicitly discussed race-conscious rather than gender-conscious classifications, we applied its standard in Cohen II. See Cohen II, 991 F.2d at 901.
Since Cohen II, however, Metro Broadcasting has been overruled, at least in part. See Adarand Constr. Inc. v. Pena, — U.S. -, -, 115 S.Ct. 2097, 2111-12, 132 L.Ed.2d 158 (1995). In Adarand, the Supreme Court held that “all racial classifications ... must be analyzed under strict scrutiny.” Adarand, — U.S. at -, 115 S.Ct. at 2113. The Court in Adarand singled out Metro Broadcasting as a “significant departure” from much of the Equal Protection jurisprudence that had come before it, in part because it suggested that “benign” government race-conscious classifications should be treated less skeptically than others. See Adarand, — U.S. at -, 115 S.Ct. at 2112.
In Adarand, the Supreme Court reasoned that “ ‘it may not always be clear that a so-called preference is in fact benign.’” Id. (quoting Regents of Univ. of Cal. v. Bakke, 438 U.S. 265, 98 S.Ct. 2733, 57 L.Ed.2d 750 (1978) (opinion of Powell, J.)). Additionally, the Supreme Court endorsed the view that
*190[a]bsent searching judicial inquiry into the justification for such race-based measures, there is simply no way of determining what classifications are ‘benign’ or ‘remedial’ and what classifications are in fact motivated by illegitimate notions of racial inferiority or simple racial politics.
Id. at -, 115 S.Ct. at 2112; see also Richmond v. J.A. Croson Co., 488 U.S. 469, 493, 109 S.Ct. 706, 721-22, 102 L.Ed.2d 854 (1989).
It is not necessary to equate race and gender to see that the logic of Adarand— counseling that we focus on the categories and justifications proffered rather than the labels attached — applies in the context of gender. While cognizant of differences between race-focused and gender-focused Equal Protection precedent, I nevertheless think that. Adarand compels us to view so-called “benign” gender-conscious governmental actions under the same lens as any other gender-conscious governmental actions. See Adarand, — U.S. at -, 115 S.Ct. at 2112; see also United States v. Virginia, — U.S. -, -, -, 116 S.Ct. 2264, 2274, 2277, 135 L.Ed.2d 735 (1996) (viewing Virginia’s benign justification for a gender classification skeptically); Shuford v. Alabama State Bd. of Educ., 897 F.Supp. 1535, 1557 (D.Ala.1995) (stating that courts “must look behind the recitation of a benign purpose to ensure that sex-based classifications redress past discrimination”). Rather than conduct an inquiry into whether Title IX and its resulting interpretations are “benign” or “remedial,” and conscious of the fact that labels can be used to hide illegitimate notions of inferiority or simple politics just as easily in the context of gender as in the context of race, we should now follow Adarand’s lead and subject all gender-conscious government action to the same inquiry.25
B. United States v. Virginia
A second Supreme Court case has. also made it necessary to review our decision in Cohen II. In United States v. Virginia, — U.S. -, 116 S.Ct. 2264, 135 L.Ed.2d 735 (1996), the Court faced an Equal Protection challenge to Virginia’s practice of maintaining the Virginia Military Institute as an all male institution. Rather than simply apply the traditional test requiring that gender classifications be “substantially related to an important government objective,” Clark v. Jeter 486 U.S. 456, 461, 108 S.Ct. 1910, 1914, 100 L.Ed.2d 465 (1988), the Supreme Court applied a more searching “skeptical scrutiny of official action denying rights or opportunities based on sex,” id., at -, 116 S.Ct. at 2274, which requires that “[pjarties who seek to defend gender-based government action must demonstrate an ‘exceedingly persuasive justification’ for that action,” id. In its discussion, the Court stated that, in order to prevail in a gender case, “the State must show at least that the challenged classification serves important governmental objectives and that the discriminatory means employed are substantially related to the achievement of those objectives.” Id. at -, 116 S.Ct. at 2275 (internal quotations omitted) (emphasis added). Being “substantially related to an important government objective,” therefore, is considered a necessary but not sufficient condition. The Court also requires a focus on “whether the proffered justification is ‘exceedingly persuasive.’” Id.
*191Virginia “drastically revise[d] our established standards for reviewing sex-based classifications.” Id. at -, 116 S.Ct. at 2291 (Scalia, J. dissenting). “Although the Court in two places ... asks whether the State has demonstrated that the classification serves important governmental objectives and that the discriminatory means employed are substantially related to the achievement of those objectives ... the Court never answers the question presented in anything resembling that form.” Id. at -, 116 S.Ct. at 2294 (citations omitted). “[T]he Court proceeds to interpret ‘exceedingly persuasive justification’ in a fashion that contradicts the reasoning of Hogan and our other precedents.” Id.
What is important for our purposes is that the Supreme Court appears to have elevated the test applicable to sex discrimination cases to require an “exceedingly persuasive justification.” This is evident from the language of both the majority opinion and the dissent in Virginia.
This is not just a matter of semantics. Metro Broadcasting, and our application of its intermediate scrutiny standard in Cohen II, omitted the additional “skeptical scrutiny” requirement of an “exceedingly persuasive justification” for gender-based government action. Compare Virginia, — U.S. at -, 116 S.Ct. at 2274 (citing J.E.B. v. Alabama ex rel. T.B., 511 U.S. 127, 136-87, and n. 6, 114 S.Ct. 1419, - and n. 6, 128 L.Ed.2d 89 (1994)), and Mississippi Univ. for Women v. Hogan, 458 U.S. 718, 724, 102 S.Ct. 3331, 3336, 73 L.Ed.2d 1090 (1982), with Metro Broadcasting, 497 U.S. at 564-65, 110 S.Ct. at 3008-09.
I conclude, therefore, that Adarand and Virginia are irreconcilable with the analysis in Cohen II and, accordingly, we must follow the guidance of the Supreme Court in this appeal. Under the new standards established in those cases, Cohen II is flawed both because it applies a lenient version of intermediate scrutiny that is impermissible following Adarand and because it did not apply the “exceedingly persuasive justification” test of Virginia. We must, as Brown urges, reexamine the Equal Protection challenge to the three-prong test as interpreted by the district court.
C. Preliminary Injunction
In addition to the above reasons for considering the merits of this appeal, it is important to note that Cohen II was an appeal from a preliminary injunction. “When an appeal comes to us in that posture, the appellate court’s conclusions as to the merits of the issues presented on preliminary injunction are to be understood as statements of probable outcomes, rather than as comprising the ultimate law of the ease.” A.M. Capen’s Co. v. American Trading and Prod. Co., 74 F.3d 317, 322 (1st Cir.1996) (internal quotations omitted); see also Narragansett Indian Tribe v. Guilbert, 934 F.2d 4, 6 (1st Cir.1991).
The binding authority of Cohen II, therefore, is lessened by the fact that it was an appeal from á preliminary injunction. First, we now have a full record before us and a set of well-defined legal questions presented by the appellant. Trial on the merits has served to focus these questions and to provide background that allows us to consider these questions in the proper context and in detail. In its decision in Cohen II, this court recognized and, indeed, emphasized the fact that its holding was only preliminary. Cohen II, 991 F.2d at 902 (“a party losing the battle on likelihood of success may nonetheless win the war at a succeeding trial”). Rather than turning that ruling into a permanent one, we should review the question in light of the full set of facts now available.
Second, the standard of review has changed. The Cohen II court stated that it was adopting a deferential standard of review, and that “if ... the district court made no clear error of law or fact, we will overturn its calibration ... only for manifest abuse of discretion.” Id. at 902. . The test applied by the court was based on “(1) the- movant’s probability of victory on the merits; (2) the potential for irreparable harm if the injunction is refused; (3) the balance of interests as between the parties ... and (4) the public interest.” Id. The ease is now before us on appeal from the merits and we must review it accordingly. For the purposes of this ap*192peal, we must review findings of fact under a clearly erroneous standard, Reich v. Newspapers of New England, Inc., 44 F.3d 1060, 1069 (1st Cir.1995) and findings of law de novo, Portsmouth v. Schlesinger, 57 F.3d 12, 14 (1st Cir.1995). Because the standard has changed, it is conceivable that the result of the analysis will change, making review appropriate.
II. BROWN’S EQUAL PROTECTION CHALLENGE
Appellees have argued that the three-prong test does not create a gender classification because the classification applies to both women and men. Although I agree that by its words, the test would apply to men at institutions where they are proportionately underrepresented in intercollegiate athletics, I cannot accept the argument that, via this provision, the Government does not classify its citizens by gender. See United States v. Virginia, — U.S. -, -, 116 S.Ct. 2264, 2274-76, 135 L.Ed.2d 735 (1996) (applying Equal Protection review to “gender-based government action” where Commonwealth of Virginia attempted to maintain two purportedly equal single-sex institutions). Cf. Loving v. Virginia, 388 U.S. 1, 8-9, 87 S.Ct. 1817, 1821-22, 18 L.Ed.2d 1010 (1967) (stating that even though the statute at issue applied equally to members of different racial classifications, it still implicated race-related Equal Protection concerns, since the statute itself contained race-conscious classifications). The fact of gender-conscious classification, even with equal enforcement with respect to both genders, requires the application of a higher level of scrutiny than rational basis review. We cannot pretend that an interpretation of a statute that contains explicit categorization according to gender and that has intentional gender-conscious effect does not represent gender-based government action. Equal Protection is implicated where the claim is made that a classification made by the government intentionally subjects an individual to treatment different from similarly situated individuals based on an impermissible characteristic, such as race, national origin, or gender. Ronald D. Rotunda & John E. Nowak, 3 Treatise on Constitutional Law § 18.2, at 7-8 (2d ed. 1992).
A. The District Court’s Construction of the Three-Prong Test
1. Prong One
A central issue in this case is the manner in which athletic “participation opportunities” are counted. During the 1990-91 academic year, Brown fielded 16 men’s and 15 women’s varsity teams on which 566 men and 328 women participated. By the 1993-94 year, there were 12 university-funded men’s teams and 13 university funded women’s teams. These teams included 479 men and 312 women. Based on an analysis of membership in varsity teams, the district court concluded that there existed a disparity between female participation in intercollegiate athletics and female student enrollment.
Even assuming that membership numbers in varsity sports is a reasonable proxy for participation opportunities — a view with which I do not concur — contact sports should be eliminated from the calculus. The regulation at 34 C.F.R. § 106.41(b)(1995) provides that an academic institution may operate separate teams for members of each sex “where selection of such teams is based upon competitive skill or the activity involved is a contact sport.” 34 C.F.R. § 106.41(b). When a team is sponsored only for one sex, however, and where “athletic opportunities for members of that sex have previously been limited, members of the excluded sex must be allowed to try-out for the team offered unless the sport involved is a contact sport,” id. (emphasis added). The regulation, therefore, allows schools to operate single-sex teams in contact sports. In counting participation opportunities, therefore, it does not make sense to include in the calculus athletes participating in contact sports that include only men’s teams. For example, if a university chooses to sponsor a football team, it is permitted to sponsor only a men’s team. Not all sports are the same and the university should be given the flexibility to determine which activities are most beneficial to its student body. By including in its accounting a contact sport that requires very large numbers of participants, e.g., football, the district court skews the number of athletic partid-*193pants — making it impossible for the university to provide both men’s and women’s teams in other sports.
If'the athletes competing in sports for which the university is permitted to field single-sex teams are excluded from the calculation of participation rates, the proportion of women participants would increase dramatically and prong one might be satisfied. If so, the inquiry ends and Brown should be judged to be in compliance.
2. Prong Two
The district court concluded, and the majority appears to agree, that Brown failed to satisfy prong two because “merely reducing program offerings to the overrepresented gender does not constitute program expansion for the underrepresented gender.” Majority Opinion at 166. This is a curious result because the entire three-prong test is based on relative participation rates. Prong one, for example, requires that participation opportunities be provided proportionately to enrollment, but does not mandate any absolute number of such opportunities. The district court’s conclusion with respect to prong two, however, implies that a school must not only demonstrate that the proportion of women in their program is growing over time, it must also show that the absolute number of women participating is increasing.26
Under the district court’s interpretation, a school facing budgetary constraints must, in order to comply with prong two, increase the opportunities available to the underrepresented gender, even if it cannot afford to do so. Rather than respecting the school’s right to determine the role athletics will play in the future — including reducing the opportunities available to the formerly overrepresented gender to ensure proportionate opportunities — the district court and the majority demand that the absolute number of opportunities provided to the underrepresented gender be increased. I see no possible justification for this interpretation — the regulation is intended to protect against discrimination, not to promote athletics on college campuses. A school is not required to sponsor an athletic program of any particular size. It is not for the courts, or the legislature, for that matter, to mandate programs of a given size. The most that can be demanded is that athletics be provided in a non-discriminatory manner.
Furthermore, the claim that a reduction in the opportunities given to the overrepresented gender is an unacceptable method of coming into compliance with the three prong test is contrary to both Cohen II and comments of the majority opinion. The majority quotes approvingly Irom Cohen v. Brown Univ., 879 F.Supp. 185 (D.R.I.1995) (Cohen III), to demonstrate the many ways in which a university might achieve compliance:
It may eliminate its athletic program altogether, it may elevate or create the requisite number of women’s positions, it may demote or eliminate the requisite number of men’s positions, or it may implement a combination of these remedies.
Majority Opinion at 185 (quoting Cohen III). This conclusion is consistent with Cohen II, which states that a school may achieve compliance by reducing opportunities for the overrepresented gender. See Cohen II, 991 F.2d at 898 n. 15. I fail to see how these statements can be reconciled with the claim that Brown cannot satisfy prong two by reducing the number of participation opportunities for men.
3. Prong Three
Prong three of the three-prong test states that, where an institution does not comply with prongs one or two, compliance will be assessed on the basis of
whether it can be demonstrated that the interests and abilities of the members of th[e] [proportionately underrepresented] sex have been fully and effectively accommodated by the present program.
44 Fed.Reg. 71,413, 71,418 (December 11, 1979).
*194According to the district court, Brown’s athletics program violates prong three because members of the proportionately underrepresented sex have demonstrated interest sufficient for a university-funded varsity team that is not in fact being funded. The district court asserts that this is not a quota. Brown, on the other hand, argues that prong three is satisfied when (1) the interests and abilities of members of the proportionately underrepresented gender (2) are accommodated to the same degree as the proportionately overrepresented gender.
The district court’s, narrow, literal interpretation should be rejected because prong three cannot be read in isolation. First, as Brown points out, the Regulation that includes prong three provides that, in assessing compliance under the regulation, “the governing principle in this area is that the athletic interests and abilities of male and female students be equally effectively accommodated.” Policy Interpretation, 44 Fed. Reg. 71,413, 71,414. Thus, Brown contends, to meet “fully” — in an absolute sense — the interests and abilities of an underrepresented gender, while unmet interest among the overrepresented gender continues, would contravene the governing principle of “equally effective accommodation]” of the interests and abilities of students of both genders.
It is also worthwhile to note that to “fully” accommodate the interests and abilities of the underrepresented sex is an extraordinarily high — perhaps impossibly so — requirement. How could an academic institution with a large and diverse student body ever “fully” accommodate the athletic interests of its students? Under even the largest athletic program, it would be surprising to find that there is not a single student who would prefer to participate in athletics but does not do so because the school does not offer a program in the particular sport that interests the student. To read fully in an absolute sense would make the third prong virtually impossible to satisfy and, therefore, an irrelevant addition to the test.
This difficulty was recognized in Cohen II, which stated that “the mere fact that there are some female students interested in a sport does not ipso facto require the school to provide a varsity team in-order to comply with the third benchmark.” Cohen II 991 F.2d at 898. The balance that Cohen II advocates would require the institution to ensure “participatory opportunities when, and to the extent that, there is sufficient interest and ability among the members of the excluded sex to sustain a viable team.” Id. (internal citations omitted). This standard may be practical for certain sports that require large teams, but what of individual sports? A “viable” tennis team may require only a single player. The same could be said of any individual sport, including golf, track and field, cycling, fencing, archery, and so on. Therefore, we still have the problem that to “fully accommodate” the interests of the underrepresented sex may be impossible under the district court’s interpretation.
In light of the above, Brown argues that prong three is in fact ambiguous with respect to whether “fully” means (1) an institution must meet 100% of the underrepresented gender’s unmet reasonable interest and ability, or (2) an institution must meet the underrepresented gender’s unmet reasonable interest and ability as fully as it meets those of the .overrepresented gender. I agree with Brown that, in the context of OCR’s Policy Interpretation, prong three is susceptible to at least these two plausible interpretations.
Additionally, section 1681(a), a provision enacted by Congress as part of Title IX itself, casts doubt on the district court’s reading of prong three. 20 U.S.C. § 1681(a) (1988). As Brown points out, Title IX, of which the Policy Interpretation is an administrative interpretation, contains language that prohibits the ordering of preferential treatment on the basis of gender due to a failure of a program to substantially mirror the gender ratio of an institution. Specifically, with respect to Title IX’s guarantee that no person shall be excluded on the basis of sex from “participation in, be denied the benefits of or be subjected to discrimination under any education program or activity receiving Federal financial assistance,” 20 U.S.C. § 1681(a),
[n]othing contained [therein] shall be interpreted to require any educational institution to grant preferential or disparate *195treatment to the members of one sex on account of an imbalance which may exist with respect to the total number or percentage of persons of the sex participating in or receiving the benefits of any federally supported program or activity, in comparison with the total number or percentage of persons of that sex in any community.
Id. § 1681(b). Section 1681(b) provides yet another reason why the district court’s reading of prong three is troublesome and why Brown’s reading is a reasonable alternative.
Since the applicable regulation, 34 C.F.R. § 106.41, and policy interpretation, 44 Fed... Reg. 71,418, are not manifestly contrary to the objectives of Title IX, and Congress has specifically delegated to an agency the responsibility to articulate standards governing a particular area, we must accord the ensuing regulation considerable deference. Chevron, U.S.A. v. Natural Resources Defense Council, Inc., 467 U.S. 887, 844, 104 S.Ct. 2778, 2782-83, 81 L.Ed.2d 694 (1984). That notwithstanding, where — as here — the resulting regulation is susceptible to more than one reasonable interpretation, we owe no such deference .to the interpretation chosen where the choice is made not by the agency but by the district court. Therefore, like other eases of statutory interpretation, we should review the district court’s reading de novo.
B. The District Court’s Interpretation and the Resulting Equal Protection Problem
The district court’s interpretation of prongs one and three creates an Equal Protection problem, which I analyze in two steps. First, the district court’s interpretation creates a quota scheme. Second, even assuming such a quota scheme is otherwise constitutional, appellees have not pointed to an “exceedingly persuasive justification,” see Virginia, — U.S. at -, 116 S.Ct. at 2274, for this particular quota scheme.
1. The Quota
I believe that the three prong test, as the district court interprets it, is a quota. I am in square disagreement with the majority, who believe that “[n]o aspect of the Title IX regime at issue in this case ... mandates gender-based preferences or quotas.” Majority Opinion at 1,70. Put another way, I agree that “Title IX is not an affirmative action statute,” id., but I'believe that is exactly what the district court has made of it. As interpreted by the district court, the test constitutes an affirmative action, quota-based scheme.
I am less interested in the actual term “quota” than the legally cognizable characteristics that render a quota scheme impermissible. And those characteristics are present here in spades. I am not persuaded by the majority’s argument that the three-part test does not constitute a quota because it does not permit an agency or court to find a violation solely on the basis of prong one of the test; instead, an institution must also fail prongs two and three. As Brown rightly argues, the district court’s application of the three-prong test requires Brown to allocate its athletic resources to meet the as-yet-unmet interest of a member of the underrepresented sex, women in this case, while simultaneously neglecting any unmet interest among individuals of the overrepresented sex. To the extent that the rate of interest in athletics diverges between men and women at any institution, the district court’s interpretation would require that such an institution treat an individual male student’s athletic interest and an individual female student’s athletic interest completely differently: one student’s reasonable interest would have to be met, by law, while meeting the other student’s interest would only aggravate the lack of proportionality giving rise to the legal duty. “The injury in eases of this kind is that a ‘discriminatory classification prevent[s] ... competition on an equal footing.’” Adarand, — U.S. at -, 115 S.Ct. at 2104 (quoting Northeastern Fla. Chapter, Assoc’d Gen’l Contractors of America v. Jacksonville, 508 U.S. 656, 666, 113 S.Ct. 2297, 2303, 124 L.Ed.2d 586 (1993)). As a result, individual male and female students would be precluded from competing against each other for scarce resources; they would instead compete only against members of their own gender. Cf. Hopwood v. Texas, 78 F.3d 932, 943-46 (5th *196Cir.) (concluding that not only would government action precluding competition between individuals of different races for law school admissions be unconstitutional, but in fact even partial consideration of race among other factors would be unconstitutional), cert. denied, — U.S. -, 116 S.Ct. 2581, 135 L.Ed.2d 1095 (1996).27
The majority claims that “neither the Policy Interpretation nor the district court’s interpretation of it, mandates statistical balancing.” Majority Opinion at 175. The logic of this position escapes me. A school can satisfy' the test in three ways. The first prong is met if the school provides participation opportunities for male and female students in numbers substantially proportionate to their enrollments. This prong surely requires statistical balancing. The second prong is satisfied if an institution that cannot meet prong one can show a “continuing practice of program, expansion which is demonstrably responsive to the developing interest and abilities of the members of the underrepresented sex.” 44 Fed.Reg. at 71,-418. It can hardly be denied that this prong requires statistical balancing as it is essentially a test that requires the school to show that it is moving in the direction of satisfying the first prong. Establishing that a school is moving inexorably closer to satisfying a requirement that demands statistical balancing can only be done by demonstrating an improvement in the statistical balance. In other words, the second prong also requires balancing. Finally, the third prong, interpreted as the majority advocates, dispenses with statistical balancing only because it choose to accord zero weight to one side of the balance. Even a single person with a reasonable unmet interest defeats compliance. This standard, in fact, goes farther than the straightforward quota test of prong one. According to the district court, the unmet interests of the underrepresented sex must be completely accommodated before any of the interest of the overrepresented gender can be accommodated.28
A pragmatic overview of the effect of the three-prong test leads me to reject the majority’s claim that the three-prong test does not amount to a quota because it involves multiple prongs. In my view it is the result of the test, and not the number of steps involved, that should determine if a quota system exists. ' Regardless of how many steps are involved, the fact remains that the test requires proportionate participation opportunities for both sexes (prong one) unless one sex is simply not interested in participating (prong three). It seems to me that a quota with an exception for situations in which there are insufficient interested students to allow the school to meet it remains a quota. All of the negative effects of a quota remain,29 and the school can escape the quota *197under prong three only by offering preferential treatment to the group that has demonstrated less interest in athletics.
2. “Extremely Persuasive Justification” Test
In view of the quota scheme adopted by the district court, and Congress’ specific disavowal of any intent to require quotas as part of Title IX, appellees have not met their burden of showing an “exceedingly persuasive justification” for this gender-conscious exercise of government authority. As recently set forth in Virginia,, “[pjarties who seek to defend gender-based government action must demonstrate an ‘exceedingly persuasive justification’ for that action.” Virginia, — U.S. at -, 116 S.Ct. at 2274. While the Supreme Court in Virginia acknowledged that “[pjhysieal differences between men and women ... are enduring,” id. at -, 116 S.Ct. at 2276, it went on to state that such “ ‘[^Inherent differences’ between men and women, we have come to appreciate, remain cause for celebration, but not for ... artificial constraints on an individual’s opportunity.” Id.
Neither appellees nor the district court have demonstrated an “exceedingly persuasive justification” for the government action that the district court has directed in this case. In fact, appellees have failed to point to any congressional statement or indication of intent regarding a proportional representation scheme as applied by the district court. While they point to Congress’ decision to delegate authority to the relevant agencies, this does not amount to a genuine — that is, not hypothesized or invented in view of litigation, id. at -, 116 S.Ct. at 2275 — exceedingly persuasive justification in light of section 1681(b)’s “no quota” provision. We are left with the explanations discussed in Cohen II to the effect that Congress conducted hearings on the subject of discrimination against women in education. There is little more than that, because Congress adopted Title IX as a floor amendment without committee hearings or reports. See Cohen II, 991 F.2d at 898.
I believe that the district court’s interpretation of the Policy Interpretation’s three-prong test poses serious constitutional difficulties. “[Wjhere an otherwise acceptable construction of a statute would raise serious constitutional problems, [we] construe the statute to avoid such problems unless such construction is plainly contrary to the intent of Congress.” Edward J. DeBartolo Corp. v. Florida Gulf Coast Bldg. & Constr. Trades Council, 485 U.S. 568, 108 S.Ct. 1392, 99 L.Ed.2d 645 (1988); see NLRB v. Catholic Bishop of Chicago, 440 U.S. 490, 507, 99 S.Ct. 1313, 1322, 59 L.Ed.2d 533 (1979). To the extent that Congress expressed a specific intent germane to the district court’s interpretation, Congress, if anything, expressed an aversion to quotas as a method to enforce Title IX. As a result, I opt for Brown’s construction of prong three, which, as we have discussed, infra, is also a reasonable reading.
Accordingly, I would reverse and remand for further proceedings.
III. EVIDENTIARY ISSUES
In disputes over the representation of women in athletic programs, it is inevitable that statistical evidence will be relevant. There is simply no other way to assess participation rates, interest levels, and abilities. The majority opinion, however, offers inconsistent guidance with respect to the role of statistics in Title IX claims. Early in the opinion, the majority approvingly cites to the statistical evaluations conducted in Cohen I, Cohen II, and Cohen III. Majority Opinion at 163. The figures in question demonstrate that women’s participation in athletics is less than proportional to their enrollment. Later in the opinion, however, when the level of interest among women at Brown is at issue, the court adopts a much more critical attitude towards statistical evidence: “[T]here exists the danger that, rather than providing a true measure of women’s interest in sports, statistical evidence purporting to reflect women’s interest instead provides only a measure of the very discrimination that is and has been the basis for women’s lack of opportunity.” Majority Opinion at 179. In other words, evidence of differential levels of interest is not to be credited because it may *198simply reflect the result of past discrimination.
The refusal to accept surveys of interest levels as evidence of interest raises the question of what indicators might be used. The majority offers no guidance to a school seeking to assess the levels of interest of its students. Although the three-prong test, even as interpreted by the district court, appears to allow the school the opportunity to show a lack of interest, the majority rejects the best — and perhaps the only — mechanism for making such a showing.
Brown claims that the district court erred in excluding evidence pertaining to the relative athletic interests of men and women at the university. Brown sought to introduce the NCAA Gender Equity Study and the results of an undergraduate poll on student interest in athletics, but was not permitted to do so. The majority is unsympathetic to Brown’s claim that the disparity between athletic opportunities for men and women reflect a gender-based difference in interest levels. Indeed, despite Brown’s attempt to present evidence in support of its claim, the majority characterizes Brown’s argument as an “unproven assertion.” Majority Opinion at 178.30 '
Furthermore, the majority recognizes that institutions are entitled to use any nondis-eriminatory method of their choosing to determine athletic interests. Majority Opinion at 179 n. 15. If statistical evidence of interest levels is not to be considered by courts, however, there is no way for schools to determine whether they are in compliance. Any studies or surveys they might conduct in order to assess their own compliance would, in the event of litigation, be deemed irrelevant. Regardless of the efforts made by the academic institution, the specter of a lawsuit would be ever-present.
In addition, the majority has put the power to control athletics and the provision of athletic resources in the hands of the underrepresented gender. Virtually every other aspect of college life is entrusted to the institution, but athletics has now been carved out as an exception and the university is no longer in full control of its program. Unless the two genders participate equally in athletics, members of the underrepresented sex would have the ability to demand a varsity level team at any time if they can show sufficient interest. Apparently no weight is given to the sustainability of the interest, the cost of the sport, the university’s view on the desirability of the sport, and so on.
IV. FIRST AMENDMENT ISSUE
Finally, it is important to remember that Brown University is a private institution with a constitutionally protected First Amendment right to choose its curriculum. Athletics are part of that curriculum. Although the protections of the First Amendment cannot be used to justify discrimination, this court should not forget that it has a duty to protect a private institution’s right to mould its own educational environment.
The majority pays lip service to these concerns in the final pages of its long opinion, stating that “ Ve are a society that cherishes academic freedom and recognizes that universities deserve great leeway in their operations.’ ” Majority Opinion at 185 (quoting Cohen II, 991 F.2d at 906), and “[o]ur respect for academic freedom and reluctance to interject ourselves into the conduct of university affairs counsels that we give universities as much freedom as possible.” Majority Opinion at 185. Despite these statements, *199however, the majority in its opinion today, and the district court before it, have failed to give Brown University freedom to craft its own athletic program and to choose the priorities of that program. Instead, they have established a legal rule that straight-jackets college athletics programs by curtailing their freedom to choose the sports they offer.
. Our discussion in Cohen II also cited Califano v. Webster, 430 U.S. 313, 97 S.Ct. 1192, 51 L.Ed.2d 360 (1977), which has not been explicitly overruled. That case concerned Congress' provision, under the Social Security Act, for a lower retirement age for women than for men, with the result that, as between similarly situated male and female wage-earners, the female wage-earner would be awarded higher monthly social security payments, id. at 314-16, 97 S.Ct. at 1193-94. In that case, Congress specifically found that more frequent and lower age limits were being applied to women than to men in the labor market. Id. at 319, 97 S.Ct. at 1195-96. This led the Supreme Court to characterize the provision at issue as remedial rather than benign, noting that the provision had been repealed in 1972, roughly contemporaneously with “congressional [anti-discrimination] reforms [that] ... have lessened the economic justification for the more favorable benefit computation” for women. Id. at 320, 97 S.Ct. at 1196. The instant case should be distinguished from Califano for two reasons. First, Califano did not necessarily rule on benign classifications, as Metro Broadcasting and Adarand clearly did. Second, Califano, unlike the instant case, contained an "exceedingly persuasive justification” for its gender-conscious state action.
. This requirement presents a dilemma for a school in which women are less interested in athletics, as Brown contends is the case. Under such conditions, a school may be unable to succeed under the second prong because there may not be enough interested female students to achieve a continuing increase in the number of female participants.
. In response, appellees cite Kelley v. Board of Trustees, 35 F.3d 265 271 (1994), for the proposition that the three-prong test does not constitute a quota, because it does not "require any educational institution to grant preferential or disparate treatment” to the gender underrepresented in that institution’s athletic program. Id. However, in Kelley, the Seventh Circuit, unlike the district court, did not use the three-prong test as a definitive test for liability. Rather, the Seventh Circuit endorsed the test as one for compliance, in dismissing the plaintiff s claims. The Seventh Circuit did not consider the question of whether, had the defendant University of Illinois not been in compliance, lack of compliance with the three-prong test alone would trigger automatic liability, nor did the Seventh Circuit spell out what steps would have been required of defendant. At any rate, Kelley pre-dates the Supreme Court's opinions in Adarand arid Virginia, meaning that it suffers from the same defects as Cohen II.
. The problem with the majority’s argument can be illustrated with a hypothetical college admissions policy that would require proportionality between the gender ratio of the local student aged population and that of admitted students. This policy is comparable to prong one of the three prong test and is, without a doubt, a quota. It is no less a quota if an exception exists for schools whose gender ratio differs from that of the local population but which admit every applicant of the underrepresented gender. It remains a quota because the school is forced to admit every female applicant until it reaches the requisite proportion. Similarly, the district court's interpretation requires the school to accommodate the interests of every female student until proportionality is reached.
.Nor does the second prong of the test change the analysis. That prong merely recognizes that a school may not be able to meet the quotas of the first or third prong immediately, and therefore deems it sufficient to show program expansion that is responsive to the interests of the underrepresented sex.
. Among the evidence submitted by Brown are: (i) admissions data showing greater athletic interest among male applicants than female applicants; (ii) college board data showing greater athletic interest and prior participation rates by prospective male applicants than female applicants; (iii) data from the Cooperative Institutional Research Program at UCLA indicating greater . athletic interest among men than women; (iv) an independent telephone survey of 500 randomly selected Brown undergraduates that reveals that Brown offers women participation opportunities in excess of their representation in the pool of interested, qualified students; (v) intramural and club participation rates that demonstrate higher participation rates among men than women; (vi) walk-on and try-out numbers that reflect a greater interest among men than women; (vii) high school participation rates that show a much lower rate of participation among females than among males; (viii) the NCAA Gender Equity Committee data showing that women across the country participate in athletics at a lower rate than men.