Affirmed in part and reversed in part by published opinions. A per curiam opinion announced the judgment of the court. Judge TRAXLER delivered the opinion of the court with respect to Parts I, II, IV, and V, in which Chief Judge WILKINSON and Judges WIDENER, WILKINS, NIEMEYER, and WILLIAMS joined, and an opinion with respect to Parts III and VI, in which Judges WILKINS and WILLIAMS joined. Chief Judge WILKINSON wrote an opinion concurring in part in which Judge NIEMEYER joined. Judge WIDENER wrote an opinion concurring in part and dissenting in part. Judge LUTTIG wrote an opinion concurring in the judgment in part and dissenting from the judgment in part. Judges MOTZ and KING wrote a separate opinion in which Judges MICHAEL and GREGORY joined.
OPINION
PER CURIAM:This case was argued before the en banc Court on February 27, 2001. The parties presented a number of issues for our consideration, including whether the district court erred in (1) finding that unitary status had been achieved and awarding attorneys’ fees to plaintiff-intervenors based on this finding; (2) holding that the establishment of a magnet schools program was an ultra vires, unconstitutional act justifying an award of nominal damages and attorneys’ fees; (3) enjoining the Charlotte-Mecklenburg School Board from considering race in the future assignment of students or allocation of educational resources; and (4) sanctioning the Board for failing to comply with the district court’s discovery order.
Having considered the briefs and arguments of the parties, a majority of the Court holds: (1) by a 7-4 vote (Chief Judge Wilkinson and Judges Widener, Wilkins, Niemeyer, Luttig, Williams and Traxler in the affirmative), the school system has achieved unitary status, but by a 6-5 vote (Chief Judge Wilkinson and Judges Niemeyer, Michael, Motz, King and Gregory in the affirmative) attorneys’ fees for work done on the unitary status issue are denied; (2) by a 6-5 vote (Chief Judge Wilkinson and Judges Niemeyer, Michael, Motz, King, and Gregory in the affirmative), the Board did not forfeit its immunity for the establishment of the magnet schools program, and nominal damages and attorneys’ fees in that regard are denied; (3) by a unanimous vote, the injunction is vacated; and (4) by a unanimous vote, the imposition of sanctions is affirmed.
The judgment of the district court is therefore affirmed on the finding of unitary status and the imposition of sanctions, reversed as to the finding of liability for nominal damages for the establishment of the magnet schools program, reversed as to the imposition of attorneys’ fees for any reason, and reversed on the issuance of the injunction.
Unitary status having been achieved, the judgment of the district court vacating and dissolving all prior injunctive orders and decrees is affirmed. The Board is to operate the school system without the strictures of these decrees no later than the 2002-2003 school year.
AFFIRMED IN PART AND REVERSED IN PART.
TRAXLER, Circuit Judge:This case is hopefully the final chapter in the saga of federal court control over the Charlotte-Mecklenburg Schools (“CMS”). Since 1971 CMS has operated under a federally supervised desegregation plan that included limited use of racial ratios, pairing and grouping of school zones, and extensive busing. So successful was the plan that the district court removed the case from the active docket in 1975, expressing its belief that the once reluctant school board was committed to achieving desegregation and was already *312well on the way toward a unitary school system. Since then, two generations of students have passed through CMS and, until the present case, not one person has returned to court alleging that segregative practices have been continued or revived.
Now, nearly three decades later and prompted by a lawsuit filed by a white student challenging the magnet schools admissions policy, the question of whether CMS has achieved unitary status has been placed before our courts. In 1999, the district court, after a lengthy hearing and searching inquiry, concluded that CMS had indeed achieved unitary status by eliminating the vestiges of past discrimination to the extent practicable. This conclusion was not reached in haste; it was the result of a two-month hearing and an examination of extensive testimony and evidence relating to every aspect of CMS’s educational system.
A majority of this court now affirms the district court’s holding on this issue, satisfied that CMS has dismantled the dual school system. In sharp contrast to the situation in the late 1960s, when black students were segregated in black schools and taught by a predominantly black staff, CMS students today are educated in an integrated environment by an integrated faculty. Nor do we turn over control to an indecisive and uncommitted school board. CMS currently operates under the firm guidance of an integrated school board which has clearly demonstrated its commitment to a desegregated school system.
In sum, the “end purpose” of federal intervention to remedy segregation has been served, and it is time to complete the task with which we were charged — to show confidence in those who have achieved this success and to restore to state and local authorities the control of their school system. Consequently, a majority of this court affirms the district court’s unitary status determination.
However, while a majority of my colleagues agree that CMS has achieved unitary status, and have graciously joined me on this point, I respectfully depart from a separate majority’s decision to reverse the district court’s holding that CMS’s magnet schools program, which was implemented in 1992, was an ultra vires, unconstitutional act justifying an award of nominal damages and attorney fees. By denying children, on account of their race, an equal opportunity to compete for open, unclaimed slots in CMS’s extraordinary magnet program, I believe the school board pushed too far and did more than either was required or permitted. Just as the educational process of the 1960s unconstitutionally deprived black children of educational opportunities solely on account of their race, the magnet schools admissions policy deprives white children of educational opportunities solely on account of their race. Consequently, I depart from the separate majority in that I would affirm the district court’s conclusion that the magnet schools program violated the Equal Protection Clause of the Fourteenth Amendment and the liability of the school board for the violation.
I.
In 1896, the Supreme Court upheld a Louisiana statute “providing for separate railway carriages for the white and colored races.” Plessy v. Ferguson, 163 U.S. 537, 540, 16 S.Ct. 1138, 41 L.Ed. 256 (1896). The Plessy majority characterized the statute as “not necessarily implying] the inferiority of either race,” id. at 544, 16 S.Ct. 1138, but the first Justice Harlan, in dissent, aptly described the true aim of the law: “Everyone knows that the statute in question had its origin in the purpose, not so much to exclude white persons from railroad cars occupied by blacks, as to exclude colored people from coaches occupied by or assigned to white persons,” id. at 557, 16 S.Ct. 1138 (Harlan, J., dissent*313ing). Justice Harlan further “den[ied] that any legislative body or judicial tribunal may have regard to the race of citizens when the civil rights of those citizens are involved.” Id. at 554-55, 16 S.Ct. 1138 (Harlan, J., dissenting). Unfortunately, the principle of “separate but equal” reached much farther than Louisiana railways, and was applied to other public services, including education. The march of progress eventually proved the correctness of Justice Harlan’s principled stand. Segregation, in all of its manifestations, was “arbitrary” and “wholly inconsistent with the civil freedom and the equality before the law established by the Constitution.” Id. at 561-62, 16 S.Ct. 1138 (Harlan, J., dissenting).
Early efforts aimed at combating the injustice wrought by Plessy in educational settings often centered on state-funded graduate and professional schools. See, e.g., Missouri ex rel. Gaines v. Canada, 305 U.S. 337, 59 S.Ct. 232, 83 L.Ed. 208 (1938); see generally Mark V. Tushnet, The NAACP’s Legal Strategy Against Segregated Education 1925-1950 (1987). In Gaines, an African-American student was denied admission to the University of Missouri School of Law on account of his race. Missouri had no “separate but equal” law school for its African-American citizens and instead offered to pay Gaines’ tuition and expenses for a legal education in another state. The Supreme Court held that Missouri’s offer denied Gaines equal protection of the laws. The Court observed that “[t]he admissibility of laws separating the races in the enjoyment of privileges afforded by the State rests wholly upon the equality of the privileges which the laws give to the separated groups within the State.” Gaines, 305 U.S. at 349, 59 S.Ct. 232. Though providing only small victories, cases like Gaines exposed “separate but equal” for the untenable proposition that it was.
In 1954, the Supreme Court recognized the futility of measuring equality in segregated facilities. See Brown v. Board of Educ., 347 U.S. 483, 74 S.Ct. 686, 98 L.Ed. 873 (1954) (Brown I). Presented with a direct attack on Plessy in a secondary education case, the Court held that “segregation of children in public'schools solely on the basis of race” violated the Equal Protection Clause of the Fourteenth Amendment. Id. at 493, 74 S.Ct. 686. The Court emphasized that an educational “opportunity, where a state has undertaken to provide it, is a right which must be made available to all on equal terms.” Id. Recognizing that segregation differed from locality to locality, the Supreme Court subsequently declined to craft a broad, one-size-fits-all remedy, and instead instructed the federal district courts to oversee the implementation of appropriate relief based on the dictates of local circumstances. See Brown v. Board of Educ., 349 U.S. 294, 299, 75 S.Ct. 753, 99 L.Ed. 1083 (1955) (Brown II) (“Because of their proximity to local conditions and the possible need for further hearings, the courts which originally heard these cases can best perform th[e] judicial appraisal.”). The district courts were directed to make use of the “traditional attributes of equity power,” id. at 300, 75 S.Ct. 753, to ensure that students were “admit[ted] to public schools on a racially nondiscriminatory basis,” id. at 301, 75 S.Ct. 753. However, under the Brown opinions it was unclear whether a school district was required to take affirmative steps to remedy the constitutional violation, see, e.g., Briggs v. Elliott, 132 F.Supp. 776, 777 (E.D.S.C.1955) (holding that Brown merely prohibited school districts from using, the force of law to separate the races), and very little progress resulted.
Before the Supreme Court provided further guidance to the lower federal courts, in 1965 the Swann plaintiffs, who were the original class action plaintiffs representing the interests of African-American children *314in the district, challenged as constitutionally inadequate the efforts of CMS in complying with Brown. The school district’s desegregation plan was based on freedom of choice whereby “any child, without regard to race, and without regard to minority or majority of race in any particular school, might freely transfer to another school of his ’choice.” Swann v. Charlotte-Mecklenburg Bd. of Educ., 243 F.Supp. 667, 668 (W.D.N.C.1965). The district court approved the plan, observing that more could be done “to increase mixing of the races,” but that the law imposed “no such duty upon ... the School Board.” Id. at 670.
Concerned at the slow pace of school desegregation throughout the nation, the Supreme Court held in 1968 that school boards had an “affirmative duty” to end the state-imposed dual system of education. Green v. County Sch. Bd., 391 U.S. 430, 437, 88 S.Ct. 1689, 20 L.Ed.2d 716 (1968). The Justices underscored that “in desegregating a dual system a plan utilizing ‘freedom of choice’ is not an end in itself.” Id. at 440, 88 S.Ct. 1689. The Swann plaintiffs then filed in the district court a motion for further relief “seeking] greater speed in desegregation of the Charlotte-Mecklenburg schools, and requesting] elimination of certain other alleged racial inequalities.” Swann v. Charlotte-Mecklenburg Bd. of Educ., 300 F.Supp. 1358, 1360 (W.D.N.C.1969). The district court, guided by the mandate of Green, see Swann, 300 F.Supp. at 1362, made a number of factual findings and concluded that the school district remained highly segregated.
The district court noted that over half of CMS’s 24,000 African-American students “attend schools that are all black, or very nearly all black, and most of the 24,000 have no white teachers.” Id. at 1360. However, the court found no violations “in the use of federal funds; the use of mobile classrooms; quality of school buildings and facilities; athletics; PTA activities; school fees; free lunches; books; elective courses; nor in individual evaluation of students.” Id. at 1372.
The district court directed CMS to submit “a positive plan for faculty desegregation effective in the fall of 1969, and a plan for effective desegregation of pupil population, to be predominantly effective in the fall of 1969 and to be completed by the fall of 1970.” Id. at 1360. The board procrastinated, but eventually submitted an enervated desegregation plan that the district court approved “with great reluctance” on a temporary basis. Swann v. Charlotte-Mecklenburg Bd. of Educ., 306 F.Supp. 1291, 1298 (W.D.N.C.1969). CMS officials, however, continued to drag their feet, and the district court was forced to appoint its own expert, Dr. John A. Finger, to craft an efficacious desegregation plan. See Swann v. Charlotte-Mecklenburg Bd. of Educ., 311 F.Supp. 265 (W.D.N.C.1970). Dr. Finger’s plan, adopted by the district court, included limited use of mathematical ratios, pairing and grouping of school zones, and busing. See id. We affirmed a portion of the plan, but vacated provisions dealing with the busing of elementary school students because of the perceived burdens on small children and the cost of purchasing new buses. See Swann v. Charlotte-Mecklenburg Bd. of Educ., 431 F.2d 138, 147 (4th. Cir.1970) (en banc). We remanded “for reconsideration of the assignment of pupils in the elementary schools.” Id. The Supreme Court granted certiorari and reinstated the district court’s plan pending further proceedings. See Swann v. Charlotte-Mecklenburg Bd. of Educ., 399 U.S. 926, 90 S.Ct. 2247, 26 L.Ed.2d 791 (1970). The district court conducted eight days of hearings and examined five different desegregation plans. The district court concluded the Finger plan to be the best of the five, encompassing “a reasonable ... collection of methods for solving the problem” of the dual system. Swann v. Char*315lotte-Mecklenburg Bd. of Educ., 318 F.Supp. 786, 800 (W.D.N.C.1970). As for busing and the cost of new buses, the district court found that the Finger plan took “proper advantage of traffic movement” and that new buses would cost only $660,000, a far cry from the millions of dollars that CMS had originally estimated. See id. at 797-98. Two months later, the Supreme Court granted certiorari and undertook an in-depth review of the power of the federal district courts to craft such sweeping desegregation remedies. See Swann v. Charlotte-Mecklenburg Bd. of Educ., 402 U.S. 1, 91 S.Ct. 1267, 28 L.Ed.2d 554 (1971).
The Supreme Court affirmed the desegregation plan adopted by the district court, and in the course of its opinion identified and offered guidance in “four problem areas.” Id. at 22, 91 S.Ct. 1267. First, the Court addressed the issue of the district court’s use of racial ratios. While the Supreme Court approved of a limited use of mathematical ratios in a plan crafted by a district court, it emphasized that such ratios were “a starting point ... rather than an inflexible requirement.” Id. at 25, 91 S.Ct. 1267. The Court reminded district courts that “[t]he constitutional command to desegregate schools does not mean that every school in every community must always reflect the racial composition of the school system as a whole.” Id. at 24, 91 S.Ct. 1267. Second, the Court dealt with single-race schools. Though the Court concluded that schools consisting of predominantly one race were not per se unconstitutional, the Court instructed the district courts to utilize “close scrutiny to determine that school assignments are not part of state-enforced segregation.” Id. at 26, 91 S.Ct. 1267. Third, the Court considered alterations of attendance zones. The Court held “that the pairing and grouping of noncontiguous school zones is a permissible tool,” id. at 28, 91 S.Ct. 1267, but declined to craft “rigid rules” in light of differing local circumstances, id. at 29, 91 S.Ct. 1267. Finally, the Court tackled the busing issue. The Court confirmed that a district court could order “bus transportation as one tool of school desegregation,” but within reasonable time and distance restrictions. Id. at 30, 91 S.Ct. 1267.
Shortly after the Supreme Court issued its landmark Swann opinion, CMS asked the district court to abandon the Finger plan and permit the substitution of a “feeder plan” whereby schools would draw pupils from designated attendance areas in an effort to keep children together for their entire public school career. See Swann v. Charlotte-Mecklenburg Bd. of Educ., 328 F.Supp. 1346 (W.D.N.C.1971). Citing concerns of resegregation and the placement of additional burdens on African-American children, the district court questioned the feeder plan. See id. at 1350-53. CMS then withdrew its original feeder plan and began work on a modified version. See id. at 1353. The district court eventually approved a revised feeder plan that reopened several former black schools and prevented over- and under-utilization of facilities. See Swann v. Charlotte-Mecklenburg Bd. of Educ., 334 F.Supp. 623 (W.D.N.C.1971).
However, within just two years it became clear that CMS’s revised feeder plan was inadequate “for dealing with foreseeable problems” in the dismantling of the dual system. Swann v. Charlotte-Mecklenburg Bd. of Educ., 362 F.Supp. 1228, 1229 (W.D.N.C.1973). The district court found “that various formerly black schools and other schools will turn black under the feeder plan,” id., and that “[rjacial discrimination through official action has not ended in this school system,” id. at 1230. The district court again instructed CMS to design a new pupil assignment plan “on the *316premise that equal protection of laws is here to stay.” Id. at 1238.
In 1974 CMS adopted and the district court approved new guidelines and policies for pupil assignment. See Swann v. Charlotte-Mecklenburg Bd. of Educ., 379 F.Supp. 1102 (W.D.N.C.1974). The plan was designed by a citizens advisory group working with the board in an effort to reach “an acceptable consensus” on school desegregation in CMS. Id. at 1103. The plan’s most promising features were the avoidance of any majority black schools (with the exception of Hidden Valley, an exempted school), and a more equal distribution of the busing burden. See id. at 1105-1110. Praising the board for making “a clean break with the essentially ‘reluctant’ attitude which dominated Board actions for many years,” the district court predicted that the policies and positive attitude would eventually result in a unitary school system. Id. at 1103.
The district court closed Swann in 1975 and removed the case from the active docket. See Swann v. Charlotte-Mecklenburg Bd. of Educ., 67 F.R.D. 648 (W.D.N.C.1975). In so doing, the district court observed that the board was “actively and intelligently addressing” recurrent problems related to dismantlement of the dual system. Id. at 649. The district court was so satisfied with the progress being made that it questioned whether it would ever be confronted with a motion to reopen the litigation. See id.
For three years there was no action in the case. This changed in 1978 when a group of white parents sought to enjoin CMS from reassigning over 4000 students in an effort to maintain racial balance in certain schools. See Martin v. Charlotte-Mecklenburg Bd. of Educ., 475 F.Supp. 1318 (W.D.N.C.1979). The parents attacking the 1978 student assignment plan “offered no live evidence but offered and relied upon a few written exhibits and admissions from the pleadings.” Id. at 1321. Not surprisingly, the district court rejected the parents’ challenge to the student assignment plan and praised CMS for its zeal in dismantling the dual system.
In 1980, CMS and the Swann plaintiffs again returned to the district court. The parties informed the district court that the African-American student population in CMS’s elementary schools had grown from twenty-nine percent to forty percent, making it difficult to avoid predominantly black student bodies. To provide the board with some flexibility, the district court permitted operation of elementary schools with African-American student bodies of plus fifteen percent above the district-wide average. See Swann v. Charlotte-Mecklenburg Bd. of Educ., No.1974 (W.D.N.C. Apr. 17, 1980).
Since 1980, neither the board nor the Swann plaintiffs have approached the district court regarding alteration of the earlier desegregation orders. And, until the present litigation, the Swann plaintiffs have never attempted to reopen the case in order to address any alleged failure by the board to comply with the district court’s desegregation orders.
The controversy before us today arose in September 1997 when William Capacchione (“Capacchione”) filed suit against CMS on behalf of his daughter, Cristina, alleging that she had been unconstitutionally denied admission to a magnet school program on account of her race. In 1992, without prior court approval, CMS had adopted a desegregation plan focused mainly on the use of magnet schools. In filling magnet schools, CMS had instituted a black and a non-black lottery to achieve racial balance. If a sufficient number of blacks or whites did not apply and fill the seats allotted to their respective races, then CMS would actively recruit children *317of the desired race despite lengthy waiting lists made up of children of the other race. If the recruitment drive faded, CMS usually left the available slots vacant. Cristina, who is white, was placed on a waiting list and eventually denied admission to a program at the Olde Providence magnet school, which CMS marketed as “a school to benefit everyone.” J.A. XXXII-15,670.
The original Swann plaintiffs moved to reactivate Sivann and to consolidate it with Capacchione’s suit. They asserted that the vestiges of the dual school system had not been abolished and that the use of race in the magnet admissions policy was necessary for the school district to comply with the prior desegregation orders. The district court granted the motion and later permitted Capacchione to intervene in the Swann litigation. Seeking a finding that CMS had eradicated the vestiges of past discrimination, another group of parents, led by Michael P. Grant (“Grant”),1 was also permitted to intervene in the litigation.
After a two-month bench trial, the district court determined that CMS had achieved unitary status, that the race-based admissions policy for CMS’s magnet schools fell outside prior orders and was not narrowly tailored to achieve a compelling state interest, and that an injunction was warranted. The district court “en-joinfed] CMS from any further use of race-based lotteries, preferences, and set-asides in student assignment.” Capacchione v. Charlotte-Mecklenburg Schs., 57 F.Supp.2d 228, 292 (W.D.N.C.1999). Citing interests in stability, the district court concluded that the injunction would not affect student assignments for the 1999 2000 school year, but would apply to student assignments for the 2000-2001 school year. See id. at 292 n. 52. The district court awarded Capacchione nominal damages in recognition of the constitutional violation and also awarded the plaintiff-intervenors attorney fees. CMS and the Swann plaintiffs filed notices of appeal, and CMS moved to stay the injunction, except as applied to the magnet schools, until the 2001-02 school year. The Swann plaintiffs moved for a complete stay pending appeal. On November 15, 1999, the district court denied the motions. CMS and the Swann plaintiffs, pursuant to Federal Rule of Appellate Procedure 8(a)(2), moved this court for a stay. On December 30, 1999, we stayed the district court’s injunction pending further order of this court.
After briefing and appellate arguments, a divided panel of this court vacated and remanded the district court’s unitary status determination, holding that the district court’s findings were insufficient in the areas of student assignment, facilities and resources, transportation, and student achievement. As for CMS’s magnet schools admissions policy, the panel held that the policy was specifically permitted by prior court orders and that the policy did not violate the Constitution. The panel also vacated the district court’s injunction, the award of nominal damages, and the award of attorney fees. See Belk v. Charlotte-Mecklenburg Bd. of Educ., 233 F.3d 232 (4th Cir.2000). A majority of the active circuit judges thereafter voted to hear this appeal en banc.
II. Unitary Status
The district court’s unitary status finding is reviewed for clear error. See Riddick v. School Bd., 784 F.2d 521, 533 (4th Cir.1986); Fed.R.Civ.P. 52(a). “A *318finding is clearly erroneous when, although there is evidence to support it, on the entire evidence the reviewing court is left with the definite and firm conviction that a mistake has been committed.” Faulconer v. Commissioner, 748 F.2d 890, 895 (4th Cir.1984). In clarifying the clearly erroneous standard, the Supreme Court has explained:
If the district court’s account of the evidence is plausible in light of the record viewed in its entirety, the court of appeals may not reverse it even though convinced that had it been sitting as the trier of fact, it would have weighed the evidence differently. Where there are two permissible views of the evidence, the factfinder’s choice between them cannot be clearly erroneous.
Anderson v. City of Bessemer City, 470 U.S. 564, 573-74, 105 S.Ct. 1504, 84 L.Ed.2d 518 (1985). The Supreme Court also stressed that even when appellate review is based primarily on documentary evidence, the clearly erroneous standard of review remains the same. See id. at 574, 105 S.Ct. 1504. So long as the district court’s unitary status determination rests on a permissible view of the evidence, it must be affirmed.
The Supreme Court has declined to define or provide a “fixed meaning” for the term “unitary.” Freeman v. Pitts, 503 U.S. 467, 487, 112 S.Ct. 1430, 118 L.Ed.2d 108 (1992). However, in light of the aim of Brown I, which was “the elimination of state-mandated or deliberately maintained dual school systems,” Milliken v. Bradley, 418 U.S. 717, 737, 94 S.Ct. 3112, 41 L.Ed.2d 1069 (1974) (Milliken I), a school system must be declared unitary when it no longer discriminates between children on the basis of race, see Green, 391 U.S. at 442, 88 S.Ct. 1689. The burden of proof falls on the party seeking an end to court supervision. See Freeman, 503 U.S. at 494, 112 S.Ct. 1430.
In undertaking a unitary status inquiry, a court must ask “whether the Board ha[s] complied in good faith with the desegregation decree since it was entered, and whether the vestiges of past discrimination ha[ve] been eliminated to the extent practicable.” Board of Educ. v. Dowell, 498 U.S. 237, 249-50, 111 S.Ct. 630, 112 L.Ed.2d 715 (1991). Implicit in the Supreme Court’s use of the term “practicable” is “a reasonable limit on the duration of ... federal supervision.” Coalition to Save Our Children v. State Bd. of Educ., 90 F.3d 752, 760 (3d Cir.1996); see also Dowell, 498 U.S. at 247, 111 S.Ct. 630 (“From the very first, federal supervision of local school systems was intended as a temporary measure to remedy past discrimination.”). Hence, the goals of a desegregation order not only encompass a remedy for the violation, but also prompt restoration of local control. See Freeman, 503 U.S. at 490, 112 S.Ct. 1430 (“Returning schools to the control of local authorities at the earliest practicable date is essential to restore their true accountability in our governmental system.... Where control lies, so too does responsibility.”); Milliken I, 418 U.S. at 741-42, 94 S.Ct. 3112 (“No single tradition in public education is more deeply rooted than local control over the operation of schools; local autonomy has long been thought essential both to the maintenance of community concern and support for public schools and to quality of the educational process.”).
Among the most important reference points in determining whether a school board has fulfilled its duties so that local control may be resumed are the factors set out in Green: student assignment, faculty assignment, facilities and resources, transportation, staff assignment, and extracurricular activities. See Green, *319391 U.S. at 435, 88 S.Ct. 1689. In its discretion, a court conducting a unitary status hearing may consider other relevant factors not mentioned in Green. See Freeman, 503 U.S. at 492, 112 S.Ct. 1430. We address the district court’s consideration of each factor in turn, but only to determine whether “the district court’s account of the evidence is plausible in light of the record viewed in its entirety.” Anderson, 470 U.S. at 573-74, 105 S.Ct. 1504.
A. Student Assignment
Student assignment is perhaps the most critical Green factor because state-mandated separation of pupils on the basis of race is the essence of the dual system. See Freeman, 503 U.S. at 474, 112 S.Ct. 1430 (observing that the issue of student assignment is “fundamental” because “under the former de jure regimes racial exclusion was both the means and the end of a policy motivated by disparagement of ... the disfavored race”). To determine whether a school was racially balanced or imbalanced, the district court adopted a plus/minus fifteen percent variance from the district-wide ratio of black to white students. See Capacehione, 57 F.Supp.2d at 246. However, the district court emphasized “that there is no level of compliance with the standard that is determinative.” Id. When schools are outside the variance, a “reasonable and supportable explanation! ]” will suffice. Id.
The district court did not err in adopting a plus/minus fifteen percent variance. Considering that the only variance ever approved by the district court in the course of the Swann litigation was a “ ‘plus 15%’ from the district-wide average,” id. at 245, the addition of a minus fifteen percent is reasonable. Moreover, the Supreme Court has permitted a “limited use ... of mathematical ratios” by district courts, Swann, 402 U.S. at 25, 91 S.Ct. 1267, and much higher variances have been used to define desegregation, see Manning v. Hillsborough County Sch. Bd., 244 F.3d 927, 935 (11th Cir.2001) (using a plus/minus twenty percent variance); see generally, David J. Armor, Forced Justice: School Desegregation and the Law 160 (1995) (observing that in over seventy percent of the school districts with desegregation plans where racial balance is measured by numerical standards, a variance of plus/minus fifteen percent or greater is used).2 In sum, the plus/minus fifteen percent variance is clearly within accepted standards, and provides a reasonable starting point in the unitary status determination.
1. CMS’s Compliance Record
The district court began by observing that since 1970, of the 126 schools in operation, “only twenty schools (16%) have had black student bodies higher than 15% above the district-wide ratio for more than three years, and only seventeen schools *320(13%) have had black student bodies lower than 15% below the district-wide ratio for more than three years.” Capacchione, 57 F.Supp.2d at 248 (footnote omitted). In addition, the district court found that CMS has not operated a single-race school since 1970. See id.
The district court also turned to two desegregation indices: the dissimilarity index and the index of interracial exposure. The former “measures the degree of racial imbalance, and it is derived by comparing the racial composition of each school to the district-wide composition,” J.A. XXXIII-16,172, and the latter measures “the average percent white in schools attended by black students, weighted by the proportion of black students in each school.” J.A. XXXIII-16,172. According to the report of the plaintiff-intervenors’ expert witness, Dr. David J. Armor, a dissimilarity value of twenty or below signifies “a highly balanced school system” and a score under thirty signifies “a substantially desegregated system.” J.A. XXXIII-16,172. CMS’s dissimilarity score was sixteen in 1980 and twenty-six in 1995. From this it is clear that CMS quickly desegregated in the 1970s and continues to maintain a “substantially desegregated system.” The dissimilarity index also indicates that CMS has better racial balance than several comparable districts did when they were declared unitary. See J.A. XXXIII-16,173.
The index of interracial exposure, like the dissimilarity index, shows that CMS has made great leaps of progress. A score of zero on the exposure index signifies total segregation, while a score of fifty or above indicates a “highly desegregated system.” J.A. XXXIII-16,172. Schools in CMS typically score above fifty, whereas before the desegregation order the schools’ scores hovered near twenty or below. See J.A. XXXIII-16,194-96.
CMS and the Swann plaintiffs correctly point out that the data suggest that in recent years racial imbalance has increased in some schools. Aware of this trend, the district court made a number of findings on growth and demographic change in the Charlotte-Mecklenburg area. The most revealing findings are as follows:
• the county population has increased from 354,656 in 1970 to 613,310 in 1997
• in 1970 the school district was the forty-third largest in the nation and is today the twenty-third largest
• among cities with more than 500,000 people, Charlotte ranks second in population growth in the 1990s
• the racial composition of the county has changed from seventy-six percent white and twenty-four percent black in 1970 to sixty-eight percent white, twenty-seven percent black, and five percent other in 1997
• the current racial composition of schoolchildren is fifty percent white, forty-two percent black, and eight percent other
• as the county has become more suburban the inner city and nearby suburbs have lost large numbers of white residents as they spread farther out into the formerly rural sections of the county
• some middle suburban communities that were once all white are now predominately black
• the rural black population in the southern part of the county has remained relatively constant while the white population has tripled because of suburbanization
See Capacchione, 57 F.Supp.2d at 236-39. “These findings are supported by the report of the plaintiff-intervenors’ expert in demographics, Dr. William Clark. See J.A. *321XXXIII-16230-306. Accordingly, the district court concluded that “[t]here can be no doubt that demography and geography have played the largest role in causing imbalance.” Capacchione, 57 F.Supp.2d at 250.
Testimony from Dr. John Murphy, CMS’s superintendent from 1991 to 1995, corroborates the district court’s conclusion. Dr. Murphy testified that when he assumed his duties he “was quite concerned about the increasing difficulty in bringing about racial balance ... because of the demographic shifts that were occurring.” J.A. VI-2712. Population growth translated into more automobiles on the road, making increased busing impracticable because “the travel time to move youngsters from the suburbs into the city with the flow of rush hour traffic was a problem.” J.A. VI-2732. In the fall of 1991, CMS hired Dr. Michael J. Stolee to examine the problem and offer solutions. Dr. Stolee also concluded that CMS’s task “has been complicated by population growth,” J.A. XXXII-15,571, and he recommended the adoption of a magnet schools program, which CMS promptly implemented.
The Supreme Court has dealt with similar population growth and shifting demographics in the context of unitary status. In Freeman, the court unequivocally stated that “racial imbalance ... [is] not tantamount to a showing that the school district [is] in noncompliance with the decree or with its duties under the law.” 503 U.S. at 494, 112 S.Ct. 1430. Brown I, of course, does not mandate that racial balance be pursued in perpetuity. Once the original racial imbalance caused by a constitutional violation has been rectified, “the school district is under no duty to remedy imbalance that is caused by demographic fae-tors.” Freeman, 503 U.S. at 494, 112 S.Ct. 1430.
The Swann plaintiffs contend that consideration of demographics and the rationale of Freeman are misplaced because the growth and shifting demographics of DeKalb County, Georgia, the school district under court order in Freeman, exceeded that of Charlotte-Mecklenburg. While CMS’s growth rates and demographic shifts certainly do not equal those experienced in DeKalb,3 we can find nothing in Freeman limiting its holding to the specific facts of DeKalb County or establishing DeKalb as the standard for measuring imbalance caused by demographic factors. On the contrary, the opinion speaks in general terms. The Supreme Court observed that in the United States “it is inevitable that the demographic makeup of school districts, based as they are on political subdivisions such as counties and municipalities, may undergo rapid change.” Id. at 495, 112 S.Ct. 1430. Mobility, the Court noted, “is a distinct characteristic of our society.” Id. at 494, 112 S.Ct. 1430.
Similarly, the Swann plaintiffs contend that unlike DeKalb County, Mecklenburg County has become more integrated as the black population has increased. This is simply not true. For example, a report prepared in 1992 by the Charlotte-Mecklenburg Planning Staff for Chairman Arthur Griffin concluded that “Charlotte-Mecklenburg continues to be a city of segregated neighborhoods” with “Concentrations of Black households ... generally located in the central city.” J.A. XXI-10,485; see also J.A. XXVIII-13,803 (1992 student assignment plan stating that “housing across the county is not racially integrated. Approximately 50% of all *322black students live within one district, while only 10% of white students reside in that district.”); J-A. XXII-10,575 (CMS report chronicling growth of the black population and decline of the white population in the inner city). Clearly, increased housing integration is not necessarily a corollary of African-American population growth. Hence, despite the Swann plaintiffs’ best efforts, Freeman cannot be distinguished into nothingness, nor does the standard of review permit this court to reweigh the evidence of the changes in CMS.
We also note that when confronted with growing imbalance in certain schools, the district court demanded cogent and supportable explanations from the plaintiff-intervenors, paying special attention to the former de jure schools still in use. See Capacchione, 57 F.Supp.2d at 246. Evidence presented at trial indicated that “[o]f the 16 former black schools that are still open, 13 are currently balanced and have been desegregated for periods ranging from 22 to 28 years. Of the 3 that currently exceed the 4-15% black variance, each has been balanced for at least 22 years.” J.A. XXXIII-16,176. Interestingly, of the seventy-two former white schools that are still open, fifteen are now majority black and were in balance for periods of twelve to twenty-five years. See J.A. XXXIII-16,176.
In addition, Dr. Armor examined the seventeen schools in CMS that exceeded the plus fifteen percent variance for three or more years during the last decade. See J.A. XXXIII-16,174 76.4 Sixteen of the seventeen were balanced for periods ranging from nineteen to twenty-six years, with one school experiencing balance for sixteen years. To the extent that CMS’s pupil reassignments could be assessed, Dr. Armor concluded that changes instituted by CMS were “attempts to maintain or restore racial balance in the face of overwhelming demographic growth and mobility.” J.A. XXXIII-16,176. Indeed, Dr. Armor concluded that imbalance had been reduced in several of the schools because CMS’s magnet program attracted white students from the outer reaches of the county.
Long periods of almost perfect compliance with the court’s racial balance guidelines,5 coupled with some imbalance in the wake of massive demographic shifts, strongly supports the district court’s finding that the present levels of imbalance are in no way connected with the de jure segregation once practiced in CMS. See Freeman, 503 U.S. at 495, 112 S.Ct. 1430 (“Where resegregation is a product not of state action but of private choices, it does *323not have constitutional implications.”); Manning, 244 F.3d at 944 (“Where a [party seeking a finding of unitary status] shows that demographic shifts are a substantial cause of the racial imbalances, [the party] has overcome the presumption of de jure segregation.”); United States v. Meriwether County, 171 F.3d 1333, 1339 (11th Cir.1999) (observing that a “school district need not wage a battle against demographics to achieve perfect racial balance”). The evidence presented at trial adequately explained why a few schools have become imbalanced, and we can discern no evidence or omissions that indicate clear error has been committed in this regard.
2. Martin and Unitary Status
The Swann plaintiffs also point to school sitings, transportation burdens, and school transfers as evidence that the growing imbalance is caused by state action rather than private choices, and that CMS has not complied with the district court’s orders in good faith. In advancing their argument, the Swann plaintiffs rely chiefly on Martin v. Charlotte-Mecklenburg Board of Education, 475 F.Supp. 1318 (W.D.N.C.1979), in which a group of parents sought to enjoin CMS from reassigning over 4000 students in order to maintain racial balance in certain schools. The plaintiffs in Martin based their position on Pasadena City Board of Education v. Spangler, 427 U.S. 424, 96 S.Ct. 2697, 49 L.Ed.2d 599 (1976), and Regents of the University of California v. Bakke, 438 U.S. 265, 98 S.Ct. 2733, 57 L.Ed.2d 750 (1978). In the former case, the Supreme Court reaffirmed that district courts could not order a school district “to rearrange its attendance zones each year so as to ensure that the racial mix desired by the court was maintained in perpetuity,” Spangler, 427 U.S. at 436, 96 S.Ct. 2697, and in the latter the Court struck down a medical school admissions policy that reserved sixteen of one hundred seats in the entering class for applicants who were “ ‘economically and/or educationally disadvantaged’ ” and who were members of certain minority groups, Bakke, 438 U.S. at 274, 98 S.Ct. 2733. The district court in Martin distinguished Spangler by observing that it was but a restatement of the Swann Court’s admonition about the use of racial quotas and that, unlike Pasadena City, CMS had not achieved racially neutral attendance patterns. See Martin, 475 F.Supp. at 1340. As for the Bakke decision, the district court pointed out that no student in CMS was denied “an equal educational opportunity” and that the admissions policy in Bakke was implemented “against a backdrop devoid of specific judicial findings or administrative acknowledgments of the prior segregated status of the school system.” Id. at 1345. Accordingly, the Martin court concluded that CMS’s reassignment of students was “within constitutional limits and should be upheld.” Id. at 1321. The district court took pains to ensure that its opinion would not be interpreted too broadly: “This order simply upholds the actions of the 1978 Board against the attacks by the plaintiffs.” Id. at 1347. In the course of the Martin opinion, the district court observed that CMS had fallen short in four areas: construction and location of facilities in parts of the county likely to enhance desegregation, placement of elementary and kindergarten grades in schools throughout the county, monitoring of student transfers so as to prevent re-segregation, and allocation of the burdens of busing. See id. at 1328-29. However, the district court also noted that CMS had made great progress and that a return to the old system of segregation “has not tempted the present School Board, who are standing fast in their endeavor to run the schools according to law while providing quality education.” Id. at 1347.
*324In Capacchione, the district court correctly observed that “Martin was not a unitary status hearing,” Capacchione, 57 F.Supp.2d at 250, and that because “the desegregation plan was still in its fledgling stages, the Court was inclined to keep the pressure on CMS,” id. at 251. The Capacchione court further observed that post-Martin changes in Charlotte-Mecklenburg counseled looking at the “concerns [of Martin ] in a new light.” Id. The district court’s interpretation of Martin is reasonable and in accord with the rule in this circuit that a district court, as a continuous institution, is “best able to interpret its own orders.” Vaughns v. Board of Educ., 758 F.2d 983, 989 (4th Cir.1985) (school desegregation case). Moreover, the Martin order was issued thirteen years before the Supreme Court made clear in Freeman that the affirmative measures mandated by Green are not meant to remedy “private choices” that lead to resegregation. Freeman, 503 U.S. at 495, 112 S.Ct. 1430. The state of the law and the understanding of duties upon school districts were far different when Martin was handed down. Hence, a number of assertions in Martin cannot be squared with the present state of the law. See, e.g., Martin 475 F.Supp. at 1346 (stating that segregated housing patterns must necessarily lead to the unconstitutional segregation of schools). Ignoring the changes in Charlotte-Mecklenburg and in the law by erecting Martin as the framework for unitary status, as the Swann plaintiffs urged below, would defy common sense and run afoul of developments in the Supreme Court’s school desegregation jurisprudence. See United States Gypsum Co. v. Schiavo Bros., 668 F.2d 172, 176 (3d Cir.1981) (concluding that a successor judge “is empowered to reconsider [the legal conclusions of an unavailable predecessor] to the same extent that his or her predecessor could have”); see also Meriwether County, 171 F.3d at 1339 (“The law does not make a school district a prisoner based on factors, such as demographic tendencies, that are beyond its control.”). We will examine the district court’s Martin findings in turn.
a. School Siting
The district court found that CMS had not shirked its duties under the law with regard to school sitings. See Capacchione, 57 F.Supp.2d at 251-53. The record reveals that CMS has, to the extent practicable, continually endeavored to site schools in order to foster integration, and has adopted a policy of building schools in areas equally accessible to blacks and whites. Testimony of current board members indicated that in efforts to fulfill this policy, CMS has purchased property in low growth areas for school construction even though schools in predominantly white high growth areas were overcrowded. See J.A. V-1986-87. In 1992 CMS reaffirmed its siting policy and resolved that, “whenever possible,” new schools would be built in areas that would “provide black student enrollment of not less than 10 percent from the census tracts serving the new school.” J.A. XXXII-15,686. The impetus behind the resolution was growth in the periphery of the county which the board speculated would continue patterns of housing segregation, thus making it more difficult to maintain racial balance in the schools. Evidence presented at trial indicated that the ten percent rule was destined for failure because it was not possible to implement the rule and still “meet the 60-minute bus ride limit.” J.A. XXII-10,869. Nevertheless, extensive evidence was presented showing that CMS never sited schools in order to foster segregation and that “every effort was made to try to find school sites that would bring people together in balanced numbers.” J.A. VI-*3252752; see Meriwether County, 171 F.3d at 1337 (stating that “the absence of evidence indicating that racial motives played any part in the Board’s decisionmaking process” is relevant in accessing compliance with desegregation orders). For example, CMS’s executive director of planning and student placement testified that in siting schools CMS “looked at both African-American and all populations not only in the vicinity of the site, but in the entire district.” J.A. VII-2920. So dedicated was CMS to siting schools in integrated areas that it contemplated refusing a gift of land for school use because the land was in a predominantly white area. See J.A. V-1985.
Faced with growth in the predominantly white regions of the far south and north, see J.A. XXXIII-16,261, CMS was compelled to serve populations in those areas via school sitings. CMS’s data show that in the late 1990s, student population was “growing at nearly 4,000 students per year,” J.A. XXIX-14,133, and consequently the board was “just trying to keep up” with the population explosion in building schools, J.A. V-2249. Overcrowding was a problem, and in the late 1990s “the average high school expected to operate at 109 percent of its capacity.” J.A. XXIX-14,-133. Even though CMS was forced to build schools at a rapid rate to serve an expanding student population, pupil assignment plans in which CMS described population growth as a “major consideration [ ]” are replete with efforts to improve racial balance. J.A. XXIX-14,133. For example, the 1997-98 assignment plan highlighted the creation and expansion of several magnet programs specially designed to reduce the black ratio in a number of schools. See J.A. XXIX-14,147-51. To the extent practicable, CMS did not sacrifice racial balance concerns to population growth. Though the two often pulled CMS in different directions, the record indicates that the board coordinated racial balance and school sitings as best it could under the circumstances. The evidence does not indicate that the abandonment of the ten percent rule or other decisions regarding school siting were the result of a desire to perpetuate the dual school system or circumvent the district court’s orders.
CMS and the Swann plaintiffs, citing to prior orders, counter that the board has not done all that it could do in the area of school siting. Erection of such a standard, however, would effectively replace practicability with possibility. See Manning, 244 F.3d at 945 (observing that “the law does not require a defendant school board to take every conceivable step in attempting to desegregate”). The former implies measures that can be reasonably implemented under the circumstances, while the latter omits the reasonableness requirement. For instance, it was possible for CMS to adhere to the ten percent rule while ignoring growth in the far north and south of the county. Youngsters would have been compelled to ride buses for long periods while traveling with the flow of rush hour traffic, but it was nonetheless possible to adhere to the ten percent rule. Of course, the practicability of a refusal to respond to growth in Charlotte-Mecklen-burg is another matter.
In the same vein, the Swann plaintiffs contend that school siting decisions were a response to white flight, which is an impermissible reason for failing to comply with a desegregation order. Growth, of course, is far different from flight. And experts offered evidence of “the economic boom in the Charlotte Metropolitan area in the last decade.” J.A. XXXIII-16,233. Charlotte-Mecklenburg is one of the most dynamic areas in the South; it is far different from the Charlotte-Mecklenburg of Swann, and much changed from that of Martin. In *326light of the growth in the county and a plethora of evidence demonstrating that the board used its best efforts to site schools in order to foster integration, the district court did not commit error when it concluded that there is no “continuing constitutional violation[ ] in the area of school siting.” Capacchione, 57 F.Supp.2d at 253.
b. Burdens of Busing
As for the burdens of busing, the district court found that in the most recent school year, 15,533 black students and 11,184 non-black students were bused for balancing purposes. Id. As stated earlier, traffic patterns make busing suburban students into the inner city far more difficult than busing inner-city children into the suburbs. See J.A. VI-2732; J.A. V-2228. Though a disproportionate number of African-American students are bused, the growth, housing patterns, and traffic patterns support the district court’s conclusion that the realities of the current situation should not block a unitary status determination. See Meriwether County, 171 F.3d at 1341 (finding no constitutional violation when white students are “somewhat less burdened by the transportation scheme” because of demographic factors).
c. Student Transfers
Finally, Martin’s concern with student transfers appears to have been based on the assumption that CMS would experience average growth. Courts are not omniscient, and the district court in 1979 could not have foreseen the changing demographics that would make student transfers the least of CMS’s worries. In the present litigation, the district court observed “that CMS ‘kept an eye on [magnet transfers] so that there wouldn’t be a run on the bank so to speak from any one school.’” Capacchione, 57 F.Supp.2d at 250 n. 10 (alteration in original). This finding is not clearly erroneous, nor can we discern the need for more findings on this issue in light of post-Martin changes.
3. Conclusion
In sum, the district court’s findings on student assignment are “plausible in light of the record viewed in its entirety.” Anderson, 470 U.S. at 573, 105 S.Ct. 1504. The dual system of student assignment in CMS has been eradicated “to the extent practicable.” Dowell, 498 U.S. at 250, 111 S.Ct. 630. The imbalance existing in some schools is not traceable to the former dual system or to renewed discriminatory actions, but rather is a result of growth and shifting demographics. Consequently, we hold that the district court’s findings on student assignment are not clearly erroneous.
B. Faculty Assignment
In examining faculty assignment, the district court again used a plus/minus fifteen percent variance. Of the 126 schools operating in CMS, the district court found that in 1997-98 only ten schools were out of balance. The Swann plaintiffs point out that this number grew to sixteen in 1998-99, but this means that a mere twelve percent of the schools were out of balance. This is a far cry from the dual system in which “most of the 24,000 [black students] ha[d] no white teachers.” Swann, 300 F.Supp. at 1360. There is simply no evidence that CMS assigns black teachers to predominantly black schools and white teachers to predominantly white schools. Thus, the district court’s conclusion that this Green factor has been satisfied is not clearly erroneous.
C. Facilities and Resources
The Swann plaintiffs and CMS contend that the district court impermissibly shift*327ed the burden of proof on this factor. As a result of the alleged error of law, CMS and the Swann plaintiffs contend that this issue must be remanded to the district court.
This court has previously made clear that “once a court has found an unlawful dual school system, [those alleging the existence of racial disparities] are entitled to the presumption that current disparities are causally related to prior segregation, and the burden of proving otherwise rests on the defendants.” School Bd. of the City of Richmond v. Baliles, 829 F.2d 1308, 1311 (4th Cir.1987). In this case, however, the district court noted that none of the prior orders entered in the long history of the Swann litigation had ever found racial disparities to exist with regard to school facilities and concluded that CMS and the Swann plaintiffs bore the burden of establishing discrimination with regard to facilities. See Capacchione, 57 F.Supp.2d at 263 (“[I]t would defy logic to place now the burden of proof on the Plaintiff Interve-nors, requiring them to prove that vestiges of discrimination in facilities have been remedied, when the Court originally found no vestiges to exist.”). In our view, this erroneous assignment of the burden of proof, which did not affect the manner in which the parties tried the case or otherwise prejudice their rights, is harmless and does not undermine the district court’s factual conclusions regarding the facilities factor.6
Immediately after assigning the burden to CMS and the Swann plaintiffs, the district court’s order nonetheless summarized and weighed the facilities evidence presented by the parties. The district court carefully analyzed the testimony and report of Dr. Dwayne Gardner, an expert witness for CMS. Dr. Gardner analyzed seventy-three schools — every identifiably black school in CMS and a sampling of balanced schools and predominantly white schools. Dr. Gardner measured the adequacy, safety, healthfulness, accessibility, flexibility, efficiency, expansibility, and appearance of the schools. Based on the inspection he grouped schools as follows: “0-44 (suggests replacement), 45-59 (needs major improvement), 60-74 (needs minor improvement), 75-89 (serves program needs), and 90-100 (exceptional quality).” Id. at 264. The survey revealed that of the four schools that warranted replacement, two were majority white, and two were imbalanced black. See J.A. XXV-12,182-86. Thirty-four schools fell into the “needs major improvement” category, of which sixteen were imbalanced black and eighteen identifiably white.
The district court determined that Dr. Gardner’s testimony established that any current disparities were functions of the age of the facilities at issue, because
different building standards apply when a new facility is constructed as compared to when an older facility is renovated or upgraded. In other words, the renovation of an older facility usually complies with the code under which the facility was built. Because most facilities in the predominately black inner city are older while facilities in the predominately white suburbs are newer, the inference is that differences in building standards tend to affect black students disproportionately. This does not amount to racial discrimination. In*328deed, this practice applies regardless of the racial composition of the school. Thus, older schools that are predominately white — several of which were built in the 1920s — are likewise affected by this practice.
Capacchione, 57 F.Supp.2d at 265 (footnote and transcript references omitted). Thus, the district court concluded from Dr. Gardner’s testimony and report “that CMS’s facilities needs are spread across the system without regard to the racial composition of its schools.” Id.
The district court also considered the testimony of CMS’s assistant superintendent of building services, who testified that out of 108 schools in need of renovations, eighty-one percent were racially balanced or identifiably white. See J.A. VIII-3810 & 3818. The district court concluded that this witness’s testimony likewise demonstrated that the deficiencies in CMS’s facilities were unrelated to the former de jure system.
Finally, the court considered CMS’s track record in renovating old facilities, praising its practice of allocating funds on a per-pupil basis and noting that “CMS has spent a large portion of[its] bond money on improving schools in predominantly black areas.” Capacchione, 57 F.Supp.2d at 266.
After an extensive discussion of this evidence, the court made the following finding of fact with regard to facilities:
Just as Judge McMillan found thirty years ago, the Court finds today that inequities in facilities exist throughout the system regardless of the racial makeup of the school. These disparities are generally the result of the relative ages of the facilities, combined with an ongoing lack of funding and the need to accommodate unprecedented growth.
Id.
This finding is clearly determinative of the question of unitary status as to facilities, regardless of which party carried the burden of proof. That is, the district court, after carefully considering and weighing all the evidence presented on this factor, concluded that any disparity as to the condition of the facilities that might exist was not caused by any intentional discrimination by CMS, but instead was a function of the age and location of the facilities and the ever-present problem of allocating all too scarce funds. Even if the district court had assigned the burden of proof to the plaintiff-intervenors, this factual finding would have compelled a ruling in their favor. In fact, the district court acknowledged as much, stating “that the Plaintiff-intervenors have proven, to the extent possible, the absence of intent and causation.” Capacchione, 57 F.Supp.2d at 267 n. 38.7
Therefore, because the district court’s findings, which were based on the court’s weighing of all of the relevant evidence presented at trial, would have yielded the same conclusion under a proper assignment of the burden of proof, any error with regard to the burden of proof is harmless. See Washington State Dep’t of *329Transp. v. Washington Natural Gas Co., 59 F.3d 793, 801 (9th Cir.1995) (finding district court’s improper assignment of the burden of proof to be harmless because review of the entire record established “that under the proper assignment of the burden of proof, the district court would have reached the same decision”); Applewood, Landscape & Nursery Co. v. Hollingsworth, 884 F.2d 1502, 1506 (1st Cir.1989) (concluding that, if the district court improperly allocated burden of proof on a particular issue, the error was harmless because the district court’s decision on that issue turned on the weight of the evidence in the record and not on burden of proof rules); cf. Vaughns, 758 F.2d at 992 (recognizing that an error in shifting the burden of proof in a school desegregation ease may be harmless if the record is such that the court can conclude that substantial rights have not been prejudiced).
Because any error associated with the burden of proof is harmless, the only question that remains is whether the district court’s factual findings about the facilities are clearly erroneous. Though the evidence could have been weighed differently on this factor, “[w]here there are two permissible views of the evidence, the fact-finder’s choice between them cannot be clearly erroneous.” Anderson, 470 U.S. at 574, 105 S.Ct. 1504. In 1969, the district court found that there was no constitutional violation in the “quality of school buildings and facilities.” Swann, 300 F.Supp. at 1372. The Capacchione court found that this remains true today, and the evidence as a whole indicates that this finding is not clearly erroneous.
D.Transportation
During the 1998 school year, five out of every six students in CMS rode a school bus. See Capacchione, 57 F.Supp.2d at 267. The parties do not dispute the district court’s finding that “CMS provides free bus transportation to all students who do not live within a mile and a half of their schools.” Id. The focus of the Swann plaintiffs’ argument on this factor deals with the Martin opinion. As previously discussed, Martin does not provide the framework for a unitary status determination and the district court’s interpretation of Martin, along with the finding that the present state of busing “may be about the best CMS can do,” Capacchione, 57 F.Supp.2d at 253, does not constitute error.
E.Staff Assignment
The district court, noting that findings of discrimination in school staffing were never made, concluded that.CMS has complied with its constitutional duties. The parties point this court to no contrary evidence, nor have we discovered such in the record. Therefore, we hold that the district court’s findings regarding the fifth Green factor are not clearly erroneous.
F.Extracurricular Activities
The district court concluded that there was no discrimination or vestiges of discrimination with regard to extracurricular activities. The evidence presented at trial showed that the ratios of blacks and whites participating in extracurricular activities, though varying somewhat from year to year, is approximately equal. See J.A. XXIV-11,634. Areas where there are disparities were not shown to be linked to the former dual system. For example, blacks often outnumber whites in holding elective offices in student government, but whites have a higher level of representation in honors programs. No evidence is found in the record to indicate that CMS somehow pushes African-Americans toward student government and away from honors programs. Consequently, the district court’s *330conclusion that CMS has satisfied this Green factor is not clearly erroneous.
G. Ancillary Factors
1. Teacher Quality
The district court found that there was no discrimination in the quality of teaching. The Swann plaintiffs contend that this finding is clearly erroneous because students in imbalanced African-American schools are more likely to have inexperienced teachers. This “experience gap,” to the extent it exists, is minuscule. The district court found that “teachers in imbalanced-black schools had 0.7 to 1.3 fewer years experience than the district averages and had 1.6 to 2.9 fewer years experience than teachers in imbalanced-white schools.” Capacchione, 57 F.Supp.2d at 271. To use middle school teachers as an example, the statistics reveal that the average middle school teacher in an imbalanced African-American school had 8.2 years experience versus 9.8 years for his counterpart in an imbalanced white school. Id. These numbers clearly support a finding of equality rather than disparity, and cannot undermine the district court’s conclusion on this factor.
The district court also pointed to evidence indicating that experience does not necessarily relate to competency. For example, according to former Superintendent Murphy, it is not uncommon to have “excellent first-year teachers” and “very weak 35th-year teachers.” J.A. VI-2795. Other witnesses observed that the newer teachers had better “knowledge of various teaching strategies” and were more comfortable with diverse classrooms. J.A. VII-3275.
The Swann plaintiffs also assert that imbalanced African-American schools have fewer teachers with advanced degrees. For instance, in imbalanced black high schools only thirty-one percent of the teachers held advanced degrees, while forty-six percent of the teachers in imbalanced white high schools held advanced degrees. See Capacchione, 57 F.Supp.2d at 271. As it was with teacher experience, testimony was offered establishing that the number of degrees a teacher possesses does not necessarily translate into competence or quality instruction. See J.A. VII-3276. According to former Superintendent Murphy, “the degree level was not a significant indicator of getting better performance on the part of the teacher.” J.A. VI-2795. Expert reports submitted by the plaintiff-intervenors also indicated that there is “no significant relationship” between black achievement and teacher education levels. J.A. XXXIII-16,221. In sum, the district court’s conclusion that African-American students receive equal access to quality teachers is not clearly erroneous.
2. Student Achievement
The district court found that the existence of an achievement gap between black and white students was not a vestige of the dual system or evidence of discrimination in the current operation of CMS. This was an area of immense disagreement at trial, and the parties presented a mountain of data on this subject. Though the Fourteenth Amendment guarantees equal protection but not equal outcomes, if low African-American achievement is a result of the former de jure system, it must be eliminated to the extent practicable. See Dowell, 498 U.S. at 249-50, 111 S.Ct. 630. Conversely, to the extent that low achievement is linked to other factors, it is beyond the reach of the court’s authority. Most courts of appeals confronting this issue, including this court, have declined to consider the achievement gap as a vestige of discrimination or as evidence of current discrimination. See Baliles, 829 F.2d at *3311313 (upholding lower court’s findings that low achievement is “primarily attributable to the high incidence of poverty” in the school district); see also United States v. City of Yonkers, 197 F.3d 41, 54 (2d Cir.1999) (observing that “using achievement test scores as a measure, either direct or indirect, of a school system’s movement away from segregation is deeply problematic”), cert. denied, 529 U.S. 1130, 120 S.Ct. 2005, 146 L.Ed.2d 956 (2000); People Who Care v. Rockford Bd. of Educ., 111 F.3d 528, 537 (7th Cir.1997) (explaining that a number of variables, other than discrimination, account for the achievement gap); Coalition to Save Our Children, 90 F.3d at 778 (finding “a causal link between ... socioeconomic factors and student achievement”).
The plaintiff-intervenors’ expert witness, Dr. Armor, presented evidence indicating that there is no correlation between African-American performance and the racial balance of schools. See J.A. XXXIII-16,-178. For example, Dr. Armor’s studies showed that African-American students in the third through fifth grades attending schools sixteen to twenty-five percent African-American scored the same on standardized tests as their counterparts in schools seventy-five percent black or greater. See J.A. at 16,214. Similarly, African-American students in the sixth through eighth grades attending schools sixteen percent black or less scored the same on standardized tests as their counterparts in schools seventy-five percent black or greater. See J.A. XXXIII-16,215.
In order to shed light on the true causes of the achievement gap, Dr. Armor turned to socioeconomic factors. The data revealed startling differences between black and white children in CMS.
Average black family income is $31,000 compared to $59,000 for whites, and only 15 [percent] of black parents are college graduates, compared to 58 percent for white parents. A huge poverty gap is also revealed, with 63 percent of black students on free lunch compared to only 9 percent of white students. Finally, 83 percent of white students have both parents at home, compared to only 42 percent for black students.
J.A. XXXIII-16,179. According to Dr. Armor, the socioeconomic factors plus the second grade scores, which are the earliest available, explain “nearly 80 percent of the reading gap and over 70 percent of the math gap.” J.A. XXXIII-16,180. Former Superintendent Murphy testified that in his experience “[p]oor students come behind and stay that way. And in Charlotte, a majority of poor students happen to be African-American.” J.A. VI-2696. Dan Saltrick, former assistant superintendent for instructional services, also testified that in his experience low student test scores related to parental support which in turn was “a matter of ... socioeconomic levels.” J.A. VII-3280. While socioeconomic disparities between black and white pupils are troubling, they are not the result of CMS’s actions or inactions and therefore are beyond the scope of the original desegregation order. See Baliles, 829 F.2d at 1314 (“Educational deficiencies that result from problems such as poverty are best remedied by programs directed toward eliminating poverty, not by indirect solutions through school programs.”).8 Ac*332cordingly, the district court did not clearly err in finding that the achievement gap between black and white students is not a vestige of past discrimination or evidence of present discrimination.
8. Student Discipline
The district court found “that any disparities that exist in the area of discipline are not causally related to the dual system.” Capacchione, 57 F.Supp.2d at 281. In none of the court’s prior orders is there any indication that CMS has ever discriminated in meting out punishment for disruptive students. However, recent statistics show that of the 13,206 students disciplined from 1996-98, sixty-six percent were African-American. See J.A. XXIV-11,637. As the district court noted, “disparity does not, by itself, constitute discrimination.” Capacchione, 57 F.Supp.2d at 281. The idea that CMS should have a disciplinary quota is patently absurd, and there is no evidence in the record that CMS targets African-Anerican students for discipline. Instead, the evidence indicates that CMS has adopted guidelines whereby students receive the same level of punishment for certain offenses to ensure that the amount of punishment will not vary from school to school. A student charged with a disciplinary infraction may also appeal the charge “and may assert that the charge was due to racial bias.” Id. There is simply no evidence in the record that CMS treats Arican-Anerican students differently in disciplinary matters. Hence, the district court’s conclusion that the disciplinary disparities are unrelated to the former de jure system is not clearly erroneous.
H. Good Faith
Lastly, the district court found that CMS has complied with the desegregation decree in good faith. See Freeman, 503 U.S. at 491, 112 S.Ct. 1430 (requiring school board “to demonstrate its good-faith commitment to a constitutional course of action”). Seven factors supported the district court’s good-faith finding: (1) no further relief has been sought since the district court removed the case from the active docket in 1975; (2) CMS has gone above and beyond the court’s orders by continually striving to achieve balance even when the imbalance was unconnected to the dual system; (3) the board has been open to community input and sought community support for its integrative efforts; (4) the board has repeatedly reaffirmed its commitment to desegregation through various resolutions; (5) Arican-Anericans currently occupy four of the nine seats on the school board, including the chair; (6) the board’s actions over the past thirty years do not evince discriminatory motives; and (7) “no evidence has been presented that school authorities were guilty of easily correctable errors.” Capacchione, 57 F.Supp.2d at 282-83.
Testimony from former board members indicated that the court’s order has been “institutionalized,” J.A. V-2222, and that the board “always stuck to what the rules were.” J.A. V-2234. Former Superintendent Murphy testified that when he arrived in Charlotte-Mecklenburg he found a “unique” environment where “everybody wanted to make sure that their schools were racially balanced.” J.A. VI-2686. In 1992, Dr. Stolee suggested a magnet plan to increase integration, and, in the course of his recommendations, observed that “[f]or the last twenty years, the Charlotte-Mecklenburg Board of Education and the Charlotte-Mecklenburg community have, *333in good faith, complied with the orders of the court.” J.A. XXXII-15,570. He further observed “that the Charlotte-Meck-lenburg Board and community have a great deal of pride in the fact that they successfully met a challenge and made the solution work.” J.A. XXXII-15,571.
Of course, both in the district court and in appellate arguments, current CMS officials engaged in much self-recrimination and claimed that they had not pursued the dismantlement of the dual system with the requisite zeal. Right on cue, the Swann plaintiffs describe this case as “unique” because CMS “has acknowledged its own failure to comply with specific directives” of the district court. Swann Plaintiffs’ Response to Petition for Rehearing at 10. The district court gave little weight to CMS’s assertions that the board had not put forth enough effort, and the evidence presented at trial amply supports the district court in this regard. Former Superintendent Murphy testified that despite a report indicating that CMS was unitary and his belief that CMS “w[as] definitely in compliance,” no effort was made to dissolve the court order. J.A. VI-2706. Dr. Murphy gave three reasons for the avoidance of a unitary status hearing. First, he advised board members that the court hearing would be “a long, drawn-out process which would cost millions of dollars, and that would be money taken away from the instructional program.” J.A. VI-2706. Second, Dr. Murphy feared that if CMS was declared unitary “we would not be eligible for federal funding for our magnet schools.” J.A. VI-2706; see also J.A. XXII-10,563 (CMS report observing that “school districts that intend to use magnet schools for desegregation purposes can apply for grants from the federal government”); J.A. XXI-10,521 (1996-1997 Federal Magnet School Assistance Program Evaluation Report in which CMS describes federal funding as “an integral part” of its pupil assignment plan). Finally, Dr. Murphy thought it best to remain under court order so CMS could continue to racially balance schools even though the de jure violation had been remedied.
Dr. Susan Purser, the current associate superintendent of education services of CMS, expressed a similar desire for CMS to remain under court order. Though Dr. Purser testified that she believed that the school board, superintendent, and administration were dedicated to enhancing educational opportunities for all of CMS’s students regardless of race, she nonetheless expressed a preference for court supervision. Dr. Purser pointed out that the current “Board has only a limited time, because these are elected positions,” J.A. XVII-8076, and that over time “superintendents will change, [and] the people involved in [CMS] will change.” J.A. XVII-8077. At this point in the cross examination, counsel asked Dr. Purser: “But you don’t know what any future School Board or administration will do either way, do you?” J.A. XVII-8077. Dr. Purser responded: “That’s exactly my point.” J.A. XVII-8077. Dr. Purser’s testimony and that of Dr. Murphy exemplify why the Supreme Court has stressed that “federal supervision of local school systems was intended as a temporary measure to remedy past discrimination.” Dowell, 498 U.S. at 247, 111 S.Ct. 630; see also Coalition to Save Our Children, 90 F.3d at 761 n. 6 (warning of “the potential for the entrenchment of [a] putatively transitional desegregation scheme”). The district court’s desegregation orders were not intended to continue after CMS remedied the de jure violation, nor were they intended to suspend the democratic process with no prospect of restoration. Yet the orders have been institutionalized to the point that CMS officials cannot imagine life without them. Once a yoke meant to steer *334CMS towards compliance with the Constitution, the orders are now used by CMS officials as mechanisms for the attainment of different goals. In truth, CMS officials have little desire for a unitary status determination and are struggling to keep the orders firmly in place.
Ironically, CMS’s clinging to the temporary desegregation orders buttresses the district court’s finding that it is unlikely “CMS would return to an intentionally-segregative system.” Capacchione, 57 F.Supp.2d at 284. If CMS will go to such lengths to keep the court’s orders in place so that it may continue racial balancing and other policies, it is unthinkable that CMS will attempt to revive the dual system. Accordingly, the district judge’s finding of good faith is not clearly erroneous.
I. CMS’s Remedial Plan
As a response to the plaintiff-interve-nors’ push for unitary status, CMS developed a “remedial plan” addressing many of the Green factors and other ancillary factors. See J.A. XXIII-11,028. The district court dismissed the remedial plan as a “ ‘litigation strategy’ plan” and declined to consider it. Capacchione, 57 F.Supp.2d at 256. CMS and the Swann plaintiffs characterize the district court’s treatment of the remedial plan as a fundamental error of law that requires reversal of the unitary status determination. First, CMS and the Swann plaintiffs aver that the district court misconstrued the test for unitary status. Adopting the test crafted by the panel opinion, CMS and the Swann plaintiffs assert that a district court must consider (1) what a school district has done, and (2) what a school district may do in the future. See Belk, 233 F.3d at 252-53. Because the district court did not undertake the latter inquiry as to the remedial plan, CMS and the Swann plaintiffs argue that the district court’s order must be reversed. This proffered two-part test is divined from Supreme Court cases which have instructed district courts to ask “whether the Board ha[s] complied in good faith with the desegregation decree since it was entered, and whether the vestiges of past discrimination ha[ve] been eliminated to the extent practicable.” Dowell, 498 U.S. at 249-50, 111 S.Ct. 630; see also Freeman, 503 U.S. at 491, 112 S.Ct. 1430.
While we agree with the first prong of the test, we do not agree that examining “whether the vestiges of past discrimination ha[ve] been eliminated to the extent practicable,” Dowell, 498 U.S. at 249-50, 111 S.Ct. 630, requires a district court — as a matter of law — to consider a remedial plan conceived, drafted, and offered by one of the parties during the lawsuit as an obvious defense to it. The plain meaning of the relevant language is that in some desegregation cases simple compliance with the court’s orders is not enough for meaningful desegregation to take place. See Swann, 402 U.S. at 25, 91 S.Ct. 1267 (stating that “a district court’s remedial decree is to be judged by its effectiveness”). For example, a decree entered in the 1960s or 1970s could have underestimated the extent of the remedy required, or changes in the school district could have rendered the decree obsolete. In either case, a district court must look beyond mere compliance with the original decree and ask whether the vestiges of the dual system have been eliminated to the extent practicable. In the present case, the district court undertook such an inquiry. Not only did the district court address compliance, but it also looked beyond the original decree and examined how the extensive changes in the Charlotte-Mecklenburg area have affected the dismantling of the former dual system. Hence, the district court was not required under Dowell and *335Freeman to have considered CMS’s eleventh-hour remedial plan.
Likewise, the district court did not run afoul of Federal Rule of Evidence 402 when it refused to consider the remedial plan. Rule 402, of course, declares that “[a]ll relevant evidence is admissible.” Fed.R.Evid. 402. Even relevant evidence may be excluded, however, when its probative value is substantially outweighed by considerations of the needless presentation of cumulative evidence. See Fed.R.Evid. 403. And CMS’s remedial plan was certainly cumulative, citing and summarizing several expert reports which had been admitted into evidence. For example, the plan’s discussion of faculty assignment is based on the reports of Dr. William Trent, Dr. Robert Peterkin, and Dr. Roslyn Mickelson; the plan’s discussion of facilities is based on Dr. Gardner’s report; the plan’s discussion of the achievement gap between blacks and whites is based on the reports of Dr. Trent, Dr. Peterkin, and Dr. Mickelson; and the plan’s student assignment discussion is based on Dr. Gordon Foster’s report. All of the aforementioned reports were admitted into evidence and the authors of the reports testified at the hearing and were subject to cross-examination. Hence, much of the remedial plan was cumulative, providing the district court with but a rehashing of expert reports and testimony.
To the extent that the remedial plan contained relevant evidence appearing nowhere else in the record, we hold that the exclusion of such evidence was harmless. According to Federal Rule of Civil Procedure 61, a “court at every stage of the proceeding must disregard any error or defect in the proceeding which does not affect the substantial rights of the parties.” Listing myriad deficiencies, objectives, and strategies, the thirty-one page remedial plan is often short on specifics. Considering the amount of evidence presented on every aspect of CMS’s operations during other phases of the two-month bench trial, we cannot hold that the. exclusion of the remedial plan affected CMS’s substantial rights. See Ingram Coal Co. v. Mower, L.P., 892 F.2d 363, 366 (4th Cir.1989) (applying Rule 61). Because the exclusion of the remedial plan in no way renders the judgment below suspect, the district court’s treatment of the plan cannot support reversal.
J. Conclusion
Pursuant to the foregoing, we affirm the district court’s unitary status determination in toto. The district court’s findings on the Green factors and the ancillary factors are bereft of clear error and we cannot discern any error of law affecting the substantial rights of the parties. After more than three decades of federal court supervision, CMS has complied in good faith with the mandate of Brown embodied in the district court’s desegregation orders to achieve a unitary school system. The dual system has been dismantled and the vestiges of prior discrimination have been eliminated to the extent practicable.
This is not to say that CMS is a perfect school system — it is not. Like school systems across the nation, CMS faces an expanding pupil population, aging facilities, and a scarcity of funds. These difficulties, however, are not vestiges of the former de jure system and therefore do not have constitutional implications. Considering CMS’s exemplary efforts in eradicating the segregated school system, we are confident that de jure segregation is history.
III. Magnet Schools
I turn now to Capacchione’s challenge to CMS’s 1992 magnet schools plan. Specifically, Capacchione contends that his daughter Cristina was unconstitutionally *336denied admission to a magnet school program on account of her race. Capacchione does not argue that race should not have been a factor in the magnet admissions process, but that the inflexible quotas, which operated to leave seats in these specialized schools vacant despite long waiting lists, went beyond what was permissible under prior court orders and the Constitution.
As noted previously, CMS operated its schools in nearly perfect racial balance for almost twenty years under a pupil assignment plan, adopted by the board and approved by the district court in 1974, which primarily utilized paired elementary schools, satellite attendance zones, a feeder system, and three experimental “optional schools.” See Swann, 379 F.Supp. at 1103-05; J.A. XXVIII-13,536-44. In 1991, however, CMS hired Dr. Stolee to examine racial imbalance that was being caused anew by the demographic shifts and population growth in Mecklenburg County. The result of Dr. Stolee’s labors was a new pupil assignment plan, entitled “CMS Student Assignment Plan: A New Generation of Excellence.” This new plan emphasized the use of magnet schools, which would allow CMS to phase out the unpopular paired elementary schools. Magnet schools, many of which were located in predominately black neighborhoods, offered a specialized curriculum or innovative instructional styles not found in the other schools in the system.
Former Superintendent Murphy oversaw implementation of the Stolee plan and testified that the magnet program was adopted because CMS “wanted to attract more white youngsters into the inner city schools” in order to meet CMS’s racial-balance goals. J.A. VI-2709. Dr. Stolee observed in his report that “Charlotte-Mecklenburg has had a long and successful experience with mandatory school assignments,” but that in order to combat demographic shifts CMS should adopt a plan based on voluntarism. J.A. XXXII-15,581; see also Missouri v. Jenkins, 515 U.S. 70, 92, 115 S.Ct. 2038, 132 L.Ed.2d 63 (1995) (Jenkins III) (“Magnet schools have the advantage of encouraging voluntary movement of students within a school district in a pattern that aids desegregation on a voluntary basis, without requiring extensive busing and redrawing of district boundary lines.”); J.A. XXVIII-13,796 (student assignment plan boasting that “Charlotte, the city which prides itself on leading the nation in integration through busing, now has the opportunity to become the city to lead the nation in voluntary busing”). A desegregation plan using magnet schools, according to Dr. Stolee, would “give[ ] each parent an opportunity to make a choice between a school serving the area in which the family resides, a school in some other area, or a school offering a very specific attractive program.” J.A. XXXII-15,580. Dr. Stolee also recognized that the magnet-centered plan would be a dramatic shift from the prior desegregation plan which featured paired elementary schools, satellite attendance zones, and a feeder system. Thus, as part of the plan, he recommended that CMS secure approval from the district court before making any changes. Indeed, Dr. Stolee’s “RECOMMENDATION # 1,” out of forty-four, read:
THE SCHOOL BOARD, THROUGH LEGAL COUNSEL, SHOULD APPROACH THE FEDERAL COURT TO SECURE APPROVAL TO CHANGE THE COURT-ORDERED DESEGREGATION PLAN.
J.A. XXXII-15,578. This recommendation was consistent with the prior district court order directing CMS to apply to the district court “before making any material departure” from the approved desegregation plan. Swann, 311 F.Supp. at 270; see *337also J.A. XXVIII-13,790 (board member requesting that Dr. Stolee “review the federal court order” to determine if the magnet plan was permissible). However, CMS ignored Dr. Stolee’s advice and the district court’s instruction, choosing instead to withhold these changes in the desegregation plan from the district court.
The crux of the problem with CMS’s magnet school plan is its admissions process. As aptly described by the district court, it operates as follows:
At the start of the process, CMS first fills seats with preferences based on whether the applicant lives in close proximity to the school and whether the applicant has any siblings in the school. CMS then fills the remaining seats by selecting students from a black lottery and a non-black lottery until the precise racial balance is achieved.
Capacchione, 57 F.Supp.2d at 287 (internal citations omitted). As originally explained to the board, the plan sought a balance of sixty percent white and forty percent black in the magnet schools with a plus or minus fifteen percent deviation. See J.A. XXVI-11-13,705. Unfortunately, CMS opted for a strict ratio of sixty percent white and forty percent black, and decreed in its 1992 student assignment plan that magnet “slots reserved for one race will not be filled by students of another race.” J.A. XXXII-15,702. The result of this policy was that if a sufficient number of blacks or whites did not apply and fill the seats allotted to their respective races, then those seats would be left vacant. Though some exceptions were made, Superintendent Eric Smith testified that CMS generally adhered to the policy. See J.A. XV-7217.
The district court appropriately examined the magnet schools through a pre-unitary status lens, observing “that the current litigation started not as a petition for unitary status but as a discrimination suit arising out of Cristina Capacchione’s denial of admission to a magnet school based on her race.” Capacchione, 57 F.Supp.2d at 284. The district court recognized that school officials acting pursuant to a desegregation order were immune from liability for actions taken consistent with that order. See Fowler v. Alexander, 478 F.2d 694, 696 (4th Cir.1973) (law enforcement officials who confined the plaintiff pursuant to a court order were immune from § 1983 suit); see also Wolfe v. City of Pittsburgh, 140 F.3d 236, 240 (3d Cir.1998) (officials acting pursuant to court order establishing quotas for promotions are not subject to § 1983 liability); Turney v. O’Toole, 898 F.2d 1470, 1472-73 (10th Cir.1990) (holding that so long as a court order is facially valid, officials acting pursuant to that order are immune from a damages suit); Coverdell v. Department of Soc. & Health Servs., 834 F.2d 758, 764 (9th Cir.1987) (social worker is immune from § 1983 liability when executing a facially valid court order). However, the district court concluded that the use of magnet schools had never been approved and that the rigid racial limitations of the magnet admissions policy were “beyond the scope of the Court’s mandate.” Capacchione, 57 F.Supp.2d at 285. The district court then subjected the admissions policy to strict scrutiny, holding that the policy violated the Equal Protection Clause of the Fourteenth Amendment because it was not narrowly tailored to achieve the compelling state interest of remedying past discrimination. This court reviews the district court’s findings of fact for clear error and its legal conclusions de novo. See Rutherford Hosp., Inc. v. RNH Partnership, 168 F.3d 693, 698 (4th Cir.1999).
A. Immunity
I begin with the question of whether CMS officials are entitled to immunity be*338cause their actions in adopting and implementing the Stolee magnet program in 1992 were taken pursuant to and were consistent with the desegregation orders and opinions issued by the district court and Supreme Court in the early 1970s. In the main, CMS asserts that it is entitled to immunity for its act of implementing the 1992 magnet schools program without court approval because the prior desegregation orders authorized the use of “optional schools” and a racial balance goal for filling them. Like the district court, I conclude that the magnet schools plan, as implemented, was not authorized by the prior court orders and that, for the reasons stated hereafter, the CMS officials are not entitled to immunity.
1. MagneNCentered Program
As an initial matter, I note that prior court orders did not countenance implementation of a desegregation plan based primarily on magnet schools. Never was CMS given carte blanche to adopt such a program absent court review and approval. CMS counters that a magnet-centered plan was permissible insofar as the district court approved the establishment of a few experimental optional schools in 1974 as part of a plan utilizing paired elementary schools, satellite attendance zones, and a feeder system. See Swann, 379 F.Supp. at 1103-04. What CMS fails to recognize is that optional schools were but a small part of the plan approved in 1974, likely because the district court was very skeptical about their efficacy as a desegregation technique. In the course of its order, the district court noted that the history of optional schools was marked by “failure” in a number of regards and warned CMS to be cautious in creating them. Id. at 1103. Consequently, CMS began with three experimental optional schools in 1974 and increased the number to only six by the early 1990s.
The optional schools created in the wake of the 1974 order placed more “emphasis on open or traditional education than normally offered in conventional schools.” J.A. XXXII-15,683. The optional schools’ traditional programs “offer[ed] an enriched and highly structured education,” J.A. XXXII-15,732, whereas the open programs offered a “student-centered” environment that “encouraged [students] to take responsibility for their behavior and for their own learning.” J.A. XXXII-15,-733. The optional schools approved by the 1974 order were not as diverse and specialized as the magnet school program implemented in 1992. The program suggested by Dr. Stolee offered schools specializing in traditional and open educational methods and created specialized schools featuring the Montessori method; science, mathematics, and technology; foreign language immersion; learning immersion programs for young children; enhanced education for academically gifted students; and communication studies programs. See J.A. XXXII-15730-41. However, both the optional schools and the magnet schools were designed to achieve the same end result — the attraction of students to a school in a particular location by using a specialized curriculum or teaching technique. Thus, Dr. Stolee, in recommending the magnet program in 1992, observed that CMS, via its optional schools, “had some experience in such specialized schools.” J.A. XXXII-15,580.
Despite the district court’s 1970 directive that CMS obtain court approval for material modifications to the court-imposed desegregation plan, the court’s skepticism of optional schools, the approval process that took place in the ensuing years, and Dr. Stolee’s specific recommendation in 1992 that CMS seek court approval for the new magnet schools program, CMS inexplicably chose not to re*339turn to the district court to obtain approval of the magnet schools plan. At appellate argument before the entire court, CMS contended that the language in the 1970 order requiring court approval for material departures was superceded by the 1974 order. CMS points to no language in the 1974 order supporting this argument and its repeated citations to and rebanee on pre 1974 orders regarding other aspects of this case further call into doubt this new line of argument. Moreover, the 1974 order made clear that “[e]xcept as modified herein, all previous orders of court remain in effect.” Swann, 379 F.Supp. at 1105 (emphasis added). Hence, the 1970 order’s requirement that CMS obtain leave of court “before making any material departure from any specific requirement set out in the order” remained binding on school officials. Swann, 311 F.Supp. at 270.
Nevertheless, I recognize that magnet schools are frequently used by school districts under a desegregation order, see Milliken v. Bradley, 433 U.S. 267, 272, 97 S.Ct. 2749, 53 L.Ed.2d 745 (1977) (Milliken II) (approving of magnet schools as a desegregation tool), and that the district court “encouraged [CMS officials] to use their full ‘know-how’ and resources to attain” a desegregated school system, Swann, 311 F.Supp. at 269. Indeed, the plaintiff-intervenors’ own expert has touted magnet programs as an “effective way to attract sizable numbers of white students to predominately minority schools.” David J. Armor, Forced Justice: School Desegregation and the Law 223 (1995). Thus, a magnet schools program, properly implemented, can no doubt be an effective desegregation tool. However, a conclusion that CMS was free to adopt any form of magnet school program it might wish to see in place does not flow from this general proposition.
I must forcefully disagree with CMS’s contention that the mention of optional schools in the 1974 order provided legal cover for the implementation of an assignment plan depending almost entirely on magnet schools. The portions of the district court order authorizing “optional schools” could perhaps be read in isolation as authorizing CMS’s use of “magnet schools” in more diverse, specialized areas, but the order did not authorize CMS to unilaterally abandon pairing, satellites, and feeders in exchange for a magnet-centered plan. Despite the import of the 1974 order, and without even a nod to the district court, CMS in 1992 abandoned the approved desegregation plan in favor of magnets. By the end of the decade CMS had created fifty-eight magnet programs—a far cry from the six optional schools in operation in the school year just prior to the adoption of the Stolee plan. See J.A. XXXIV-16,721-30. CMS describes this abandonment of the prior plan as but an expansion of the approved use of optional schools. Clearly, this “expansion” was in reality a substantial restructuring and cannot be squared with the unambiguous directives of prior orders.
2. Strict Ratios
Even if I could conclude that a magnet-centered plan was permitted under prior court orders, the plan implemented by CMS is nonetheless ultra vires because it combines a rigid ratio of sixty percent white and forty percent black with a pobey decreeing that “slots reserved for one race will not be filled by students of another race.” J.A. XXXII-15,702.9 In 1970, the *340district court issued a desegregation order to CMS, noting that the order was “not based upon any requirement of ‘racial balance.’” Swann, 311 F.Supp. at 267 (emphasis added). The court reiterated “that efforts should be made to reach a 71-29 ratio in the various schools so that there will be no basis for contending that one school is racially different from the others, but ... that variations from the norm may be unavoidable.” Id. at 267-68 (internal quotation marks omitted). On appeal, the Supreme Court affirmed the guidelines set forth in the district court’s order and also addressed the subject of racial quotas. See Swann, 402 U.S. at 23-25, 91 S.Ct. 1267.
With regard to the district court’s goal of achieving a racial balance of seventy-one percent white and twenty-nine percent black, the Court took care to note that “[t]he constitutional command to desegregate schools does not mean that every school in every community must always reflect the racial composition of the school system as a whole.” Swann, 402 U.S. at 24, 91 S.Ct. 1267. But central to the issue now before us, the Court held that had the district court
require[d], as a matter of substantive constitutional right, any particular degree of racial balance or mixing, that approach would be disapproved and we would be obliged to reverse.
Id.See also Winston-Salem/Forsyth County Bd. of Educ. v. Scott, 404 U.S. 1221, 1227, 92 S.Ct. 1236, 31 L.Ed.2d 441 (1971) (Burger, C.J., in chambers) (describing as “disturbing” the school board’s “understanding that it was required to achieve a fixed ‘racial balance’ that reflected the total composition of the school district”). The goal was upheld, only upon the condition that “use made of mathematical ratios was no more than a starting point in the process of shaping a remedy, rather than an inflexible requirement.” Swann, 402 U.S. at 25, 91 S.Ct. 1267.
Just two years after the Supreme Court, in this very case, made clear that strict ratios were unacceptable, the district court, in a carefully worded order permitting CMS to create optional schools, approved an intentionally flexible enrollment formula of “about or above 20% black students.” Swann, 379 F.Supp. at 1104 (emphasis added). The district court recognized that the “actual enrollment of the optional school may have to be guided by its racial composition and by the number drawn from each other school area, not by considerations of space and program only.” Id. at 1108. Additionally, the district court’s order directed that “[r]eassign-ments to optional schools must not jeopardize the racial composition of any other school.” Id. These modifications, however, at no time set a racial ratio of the type disapproved of by the district court in its earlier orders and by the Supreme Court in its 1971 review of the district court’s 1970 order.10
*341CMS asserts that the inflexible racial limits adopted in the 1992 magnet-centered plan were countenanced by the 1974 order discussing optional schools.11 In making this argument, CMS ignores the district court’s choice of words in the 1974 order (“about or above 20% black students”), see Swann, 379 F.Supp. at 1104, and points to an attachment to the order designated as Exhibit A. This exhibit, a proposed pupil assignment plan drafted by CMS and a citizens advisory group, called for optional school enrollment “at or above approximately a 20% black ratio.” Id. at 1108 (emphasis added). From this language, CMS concludes that strict quotas were permitted. CMS’s concentration on just a portion of the relevant language (“at or above”) edits out the word “approximately,” which does not suggest rigidity. Even if Exhibit A could be read as requiring rigid quotas, CMS disregards the fact that the district court approved the guidelines “subject to the further conditions stated” in the 1974 order. Id. at 1103. With the Supreme Court’s admonition about strict quotas in mind, the district court chose its language carefully, observing that optional schools should “have about or above 20% black students.” Id. at 1104. Hence, it is the district court’s understanding and modification of the pupil assignment plan that controls, not CMS’s tortured reading. Under a just construction, it is clear that the 1974 order did not approve a use of race to the extent that CMS could deny eager applicants an otherwise available slot in a magnet program solely on account of the applicant’s race. Both the district court and the Supreme Court in this very case consistently rejected the use of such rigid racial quotas.
I also find no authorization for the board’s adoption of the magnet schools program in the Supreme Court’s 1971 approval in Swann of a majority-to-minority transfer policy that would prevent, for example, an African-American child in a majority white school from transferring to a majority black school because the transfer would increase the degree of segregation in the affected schools. See Swann, 402 U.S. at 26, 91 S.Ct. 1267. Because the majority-to-minority transfer policy, like the magnet admissions policy, prevents a child from enrolling in the public school of his choice, CMS argues that the magnet admissions policy is permissible. By definition, however, CMS’s specialized magnet programs are not tantamount to conventional public schools. While a child denied a transfer from one conventional school to another still receives the same general education, a child denied admission to a specialized magnet program does not receive *342a similar benefit in a conventional school. In other words, an education in a magnet school offering, for example, foreign language immersion, is not interchangeable with an education in a conventional public school.12 Hence, the effect of the magnet admissions policy is far different from the majority-to-minority transfer policy.
Unfortunately, the end result of the challenged magnet schools admissions policy is placement of racial quotas ahead of educating students — an inappropriate result nowhere countenanced in the district court’s orders or in the Supreme Court’s desegregation decisions. Cf. Wright v. Council of the City of Emporia, 407 U.S. 451, 463, 92 S.Ct. 2196, 33 L.Ed.2d 51 (1972) (holding that courts should not approve a desegregation plan if the plan offers “ ‘quality education’ to some children, [but] has a substantial adverse effect upon the quality of education available to others”). In fact, Brown I struck down segregated schooling because children were denied equal educational opportunities. See Brown I, 347 U.S. at 493, 74 S.Ct. 686. While school boards were permitted to use race in assigning students in order to convert to a unitary system, see North Carolina State Bd. of Educ. v. Swann, 402 U.S. 43, 46, 91 S.Ct. 1284, 28 L.Ed.2d 586 (1971) (holding that the use of race in pupil assignments is “one tool absolutely essential to fulfillment of [a school board’s] constitutional obligation to eliminate existing dual school systems”), neither the Brown opinions nor the district court orders implementing them ever contemplated that remedial use of race, like the old dual system, would deny some students educational opportunities solely because of their race. See Brown I, 347 U.S. at 493, 74 S.Ct. 686 (holding that an educational opportunity provided by the state “must be made available to all on equal terms”); see also Bakke, 438 U.S. at 305, 98 S.Ct. 2733 (Powell, J.) (“When a classification denies an individual opportunities or benefits enjoyed by others solely because of his race or ethnic background, it must be regarded as suspect.”).13 Indeed, in bringing suit in 1965, the Swann plaintiffs, in accord with the Brown opinions, simply asked that CMS convert “into a unitary nonracial system wherein the educational opportunities offered by [CMS] are made available to students without regard to race or color.” J.A. XXXIII-16,162 (original complaint filed by the Swann plaintiffs).
An admissions policy that uses rigid racial quotas to deny an available, unclaimed slot in a specialized magnet school to a child, whether black or white, on account of the child’s race cannot be squared with the district court’s orders or the Supreme Court’s desegregation decisions. Since 1971 it has been perfectly clear that mathematical ratios may be used as “a starting point in the process of shaping a remedy,” *343but not as “an inflexible requirement.” See Swann, 402 U.S. at 25, 91 S.Ct. 1267. The district court took heed of this admonition in 1974 when it permitted the creation of optional schools with “about or above 20% black students.” Swann, 379 F.Supp. at 1104 (emphasis added). However, CMS in 1992 ran afoul of the rule announced by the Supreme Court when it crafted strict racial ratios designed to leave open magnet school seats empty, rather than permitting waitlisted students to compete for the slots. Because nothing short of intellectual gymnastics can transform the clear meaning of the Supreme Court’s Swann opinion or the district court’s 1974 order into vehicles countenancing the rigid use of racial ratios, I agree with the district court that the policy is ultra vires and that CMS officials are not entitled to immunity.
B. Equal Protection
Having determined that the CMS officials are not entitled to immunity for the implementation of the strict race-based magnet school assignment policy, I now turn to the question of whether the offi-dais’ act of implementing the policy without prior court approval, albeit while under an order to desegregate schools, runs afoul of the Equal Protection clause. I would hold that it does.
Under the Fourteenth Amendment, “[n]o State shall ... deny to any person within its jurisdiction the equal protection of the laws.” U.S. Const, amend XIV, § 1. By guaranteeing equal protection, the Amendment recognizes that “[d]istinctions between citizens solely because of their ancestry are by their very nature odious to a free people whose institutions are founded upon the doctrine of equality.” Hirabayashi v. United States, 320 U.S. 81, 100, 63 S.Ct. 1375, 87 L.Ed. 1774 (1943). The Supreme Court has refused to make exceptions for so-called “benign” racial classifications, see Adarand Constructors, Inc. v. Pena, 515 U.S. 200, 227, 115 S.Ct. 2097, 132 L.Ed.2d 158 (1995), and the Court has made clear that “all racial classifications, imposed by whatever federal, state, or local governmental actor, must be analyzed by a reviewing court under strict scrutiny,” id.14
*344To survive strict scrutiny, CMS’s use of race in the magnet admissions program “must (1) serve a compelling governmental interest and (2) be narrowly tailored to achieve that interest.” Tuttle v. Arlington County Sch. Bd., 195 F.3d 698, 704 (4th Cir.1999), cert. dismissed, 529 U.S. 1050, 120 S.Ct. 1552, 146 L.Ed.2d 364 (2000). CMS avers that the magnet admissions policy was adopted to remedy the effects of the dual school system previously operated in Mecklenburg County. Without question, remedying the effects of past discrimination is a compelling state interest. See City of Richmond v. J.A. Croson Co., 488 U.S. 469, 493, 109 S.Ct. 706, 102 L.Ed.2d 854 (1989).
In reviewing whether a policy is narrowly tailored to serve a compelling state interest, a court considers factors such as:
(1) the necessity of the policy and the efficacy of alternative race neutral policies;
(2) the planned duration of the policy;
(3) the relationship between the numerical goal and the percentage of minority group members in the relevant population;
(4) the flexibility of the policy, including the provision of waivers if the goal cannot be met; and
(5) the burden of the policy on innocent third parties.
See United States v. Paradise, 480 U.S. 149, 171, 107 S.Ct. 1053, 94 L.Ed.2d 203 (1987) (plurality opinion). Like the district court, I would hold that the CMS magnet admissions policy is not narrowly tailored to the compelling interest of remedying past discrimination.
First, the magnet admissions policy was not necessary to comply with the court’s order to dismantle the dual educational system. CMS had a number of options available to it that would not have deprived children, solely on account of their race, an available seat in a specialized magnet program. Instead, CMS opted for rigid racial limits that were clearly prohibited by the district court’s orders and the Supreme Court’s desegregation decisions. Nor is there evidence in the record that added flexibility or a waiver provision would have undermined the use of magnet schools as a desegregation technique. The evidence simply does not reveal that the magnet admissions policy used was the only efficacious option available to CMS.
Second, this circuit has emphasized that “[t]he use of racial preferences must be limited so that they do not outlast their need; they may not take on a life of their own.” Hayes v. North State Law Enforcement Ass’n, 10 F.3d 207, 216 (4th Cir.1993) (internal quotation marks omitted). Like the district court, I can find “no mention of the duration that CMS would use racially segregated lotteries, vacancies, and waiting lists.” Capacchione, 57 F.Supp.2d at 290. In light of CMS’s desire to remain under court order for the indefinite future, see supra Part II.H, the lack of a duration for the magnet admissions policy is not surprising. CMS was apparently content, in a number of instances, to leave available magnet seats empty despite the waiting lists.
*345Third, I agree with the district court that “the 60-40 numerical goal is related to the relevant population, i.e., the racial composition of schoolchildren in CMS.” Capacchione, 57 F.Supp.2d at 289. However, there is no evidence that CMS considered the “practicability of achieving this precise ratio in every magnet school,” id. at 290, or the very real danger that magnet schools would be underutilized because seats would be left open despite an abundance of applicants. The result of the admissions policy is but another indication that the CMS administration, in the words of former Superintendent Murphy, “was more focused on balance than on [educational] outcomes.” J.A. VI-2687.
Fourth, the district court aptly described the inflexibility in the magnet admissions policy: “The Court is hard-pressed to find a more restrictive means of using race than a process that results in holding seats vacant while long waiting lists full of eager applicants are virtually ignored.” Capacchione, 57 F.Supp.2d at 289. The policy is indeed “restrictive,” but it also borders on obduracy. The policy contained no written waiver provision which, once again, shows a lack of concern that these highly specialized schools could and would be underutilized.
Finally, the innocent parties affected are children denied magnet slots solely because of their race and parents who “must wait for months without knowing where their children eventually will be placed.” Id. at 290. A child’s education is one of the greatest concerns of the family, and CMS unnecessarily causes much agonizing when it places children of the “wrong col- or” on waiting lists while it actively recruits children of the “right color” to fill empty magnet school seats.
In sum, the magnet admissions policy is not narrowly tailored. The policy is not necessary to dismantle the de jure system, is for an unlimited duration, provides for virtually no flexibility, and burdens innocent children and their families. The policy quixotically purports to establish equal protection of the laws in the realm of public education by denying children an equal opportunity to compete for open, unclaimed slots in CMS’s extraordinary magnet schools. The withholding of seats from white students after all African-American children wishing seats have been given them is most certainly not a narrowly tailored program. Such a result calls to mind why strict scrutiny is used in the first place: “Of all the criteria by which men and women can be judged, the most pernicious is that of race.” Maryland Troopers Ass’n v. Evans, 993 F.2d 1072, 1076 (4th Cir.1993). Teaching young children that admission to a specialized academic program with available seats is contingent on their race is indeed pernicious, and CMS’s magnet admissions policy can in no way be described as narrowly tailored to achieve the compelling interest of remedying past discrimination.15
C. Award of Nominal Damages
After finding a constitutional violation in the magnet schools, the district court held *346CMS “nominally liable in the amount of one dollar.” Capacchione, 57 F.Supp.2d at 290. CMS argues that the nominal damages awarded were unjustified because the actions resulting in a constitutional violation were taken in good faith. CMS fears that the damages award will “open the door to numerous suits by other students who could claim that they did suffer actual damages and argue that collateral estoppel prevents CMS from denying liability.” Defendants-Appellants’ Brief at 24. Regarding nominal damages, the Supreme Court has observed:
Common-law courts traditionally have vindicated deprivations of certain “absolute” rights that are not shown to have caused actual injury through the award of a nominal sum of money. By making the deprivation of such rights actionable for nominal damages without proof of actual injury, the law recognizes the importance to organized society that those rights be scrupulously observed; but at the same time, it remains true to the principle that substantial damages should be awarded only to compensate actual injury or, in the case of exemplary or punitive damages, to deter or punish malicious deprivations of rights.
Carey v. Piphus, 435 U.S. 247, 266, 98 S.Ct. 1042, 55 L.Ed.2d 252 (1978) (nominal damages available for denial of procedural due process rights) (footnote omitted); see also Price v. City of Charlotte, 93 F.3d 1241, 1246 (4th Cir.1996) (stating that “the rationale for the award of nominal damages being that federal courts should provide some marginal vindication for a constitutional violation”).
In the present case there was indeed a constitutional violation. CMS ran afoul of the Equal Protection Clause when it adopted a strict racial quota designed to deny an available, unclaimed slot in a specialized magnet school to a child on account of the child’s race. In order to recover nominal damages, Cristina Capac-chione need not prove that absent the unconstitutional policy she would have been admitted to the magnet program. The injury in the present case is not the ultimate inability to enroll in the magnet school, but the inability to compete for seats on an equal basis. See Northeastern Florida Chapter of the Associated Gen. Contractors v. City of Jacksonville, 508 U.S. 656, 666, 113 S.Ct. 2297, 124 L.Ed.2d 586 (1993). Though the two open “black seats” at the Olde Providence magnet school were eventually awarded to white children, the fact remains that the official magnet admissions policy prohibited children like Cristina from competing for the open slots. In fact, CMS left the two available “black seats” at Olde Providence unfilled for most of the summer while Cristina and over one hundred other white children languished on a waiting list. In Orwellian fashion, CMS marketed Olde Providence as “a school to benefit everyone,” but in reality permitted only a select few to compete for the benefits bestowed.
The nominal award in this case recognizes the importance of equal protection under the law and provides some measure of vindication. As for CMS’s worry about collateral estoppel, liability has already been established, and vacating the nominal damages would not change this. Consequently, I would affirm the district court’s award of nominal damages.
IV. Injunctive Relief
After recounting the unitary status determination and the constitutional violation in the magnet admissions policy, the district court enjoined “CMS from any further use of race-based lotteries, preferences, and set-asides in student assignment.” Capacchione, 57 F.Supp.2d at 292. CMS challenges the district court’s injunc*347tion as unwarranted and overbroad. We review the grant of a permanent injunction for an abuse of discretion. See Tuttle, 195 F.3d at 703.
Before a court grants a permanent injunction, the court must first find necessity—a danger of future violations. See Connecticut v. Massachusetts, 282 U.S. 660, 674, 51 S.Ct. 286, 75 L.Ed. 602 (1931) (stating that an injunction “will not be granted against something merely feared as liable to occur at some indefinite time in the future”); United States v. Oregon State Med. Soc’y, 343 U.S. 326, 333, 72 S.Ct. 690, 96 L.Ed. 978 (1952) (“All it takes to make the cause of action for relief by injunction is a real threat of future violation or a contemporary violation of a nature likely to continue or to recur.”); Bloodgood v. Garraghty, 783 F.2d 470, 475 (4th Cir.1986) (“An injunction is a drastic remedy and will not issue unless there is an imminent threat of illegal action.”). Though a flexible tool, an injunction may not be used for “punishment or reparations for ... past violations.” Oregon State Med. Soc., 343 U.S. at 333, 72 S.Ct. 690.
The district court’s finding of a threat of future violations centered on CMS’s offering of diversity as a compelling state interest. This interest was offered after the district court decided that the admissions policy should be reviewed using strict scrutiny. Because in this circuit it is unsettled whether diversity may be a compelling state interest, see Eisenberg v. Montgomery County Pub. Schs., 197 F.3d 123, 130 (4th Cir.1999), cert. denied, 529 U.S. 1019, 120 S.Ct. 1420, 146 L.Ed.2d 312 (2000), it was improper for the district court to base its injunction on CMS’s unsuccessful defense of the policy. At this point, we can discern nothing in the record indicating that CMS will ignore the district court order and continue to use race in an unconstitutional manner in the operation of the magnet schools or other schools in the system. CMS represented to the district court both during and after trial that it had no intention of continuing the magnet plan. In moving for a stay of the injunction, CMS did not ask that the injunction be stayed as to the magnet schools, and was prepared to comply immediately with the court’s order. CMS requested a stay as to the non-magnet schools because over 50,000 students were likely to be reassigned in a short period of time. Moreover, there was no evidence presented at trial about what CMS proposed to do as a unitary school system. A post-unitary status student assignment plan was never given to the district court, and the evidence simply does not indicate that “there is an imminent threat of illegal action.” Bloodgood, 783 F.2d at 475.
A finding of unitariness brings a fresh start for the school board—an opportunity to operate a school system in compliance with the Constitution. The prospective relief awarded by the district court is in tension with the resumption of local control, which is one of the ultimate goals of any desegregation order. See Freeman, 503 U.S. at 490, 112 S.Ct. 1430. Freeing the school district from one court order only to shackle it with another was here an abuse of the district court’s discretion, and we therefore vacate the grant of injunctive relief.
V. Discovery Sanctions
The district court sanctioned CMS for failing to supplement its answers to interrogatories that sought a list of witnesses. We review the district court’s management of discovery under the abuse of discretion standard. See Anderson v. Foundation for Advancement, Educ. & Employment of Am. Indians, 155 F.3d 500, 504 (4th Cir.1998). The record re*348veals that no list of fact witnesses was presented to the plaintiff-intervenors until five days before the trial date. At that time, CMS presented a list of 174 witness, which was later cut to twenty-six. The plaintiff-intervenors moved for sanctions and the district court granted the motion in part. The district court continued the trial for one week so that the plaintiff-intervenors could depose the newly disclosed witnesses, and the court held CMS accountable for the fees and expenses of these depositions.
We have developed a four-part test for a district court to use when determining what sanctions to impose under Federal Rule of Civil Procedure 37. Specifically, “[t]he court must determine (1) whether the non-complying party acted in bad faith, (2) the amount of prejudice that noncompliance caused the adversary, (3) the need for deterrence of the particular sort of non-compliance, and (4) whether less drastic sanctions would have been effective.” Id. An examination of the four factors reveals no abuse of discretion by the district court.
First, there is ample evidence of bad faith. Early in the case, the plaintiff-inter-venors presented CMS with an interrogatory asking for disclosure of trial witnesses. In response to the interrogatory, CMS stated that it would provide appropriate information concerning witnesses at the time and in the manner specified by the district court. The plaintiff-interve-nors moved to compel discovery, and the court agreed with CMS that the request was premature. However, the court instructed CMS to “supplement its responses [to the interrogatories], as it promised, when such information becomes known.” J.A. 1-195. As an excuse for its untimely disclosure of fact witnesses, CMS relies on the district court’s pre-trial order, which provides that “[a] witness list containing the name of every proposed witness” should be filed with the court on the first day of trial. J.A. 1-150. This provision of the pre-trial order was clearly for the court’s convenience and could not reasonably be interpreted to apply to disclosures to the other parties. Besides, even if such an interpretation were reasonable, the district court’s command to supplement interrogatories superceded the pre-trial order. Accordingly, bad faith is evident.
Second, the presentation of such a lengthy witness list on the eve of trial to the plaintiff-intervenors was prejudicial. Without the action of the court, the plaintiff-intervenors would have had no opportunity to depose the witnesses, much less properly prepare for trial. Thus, CMS’s failure to supplement interrogatories was prejudicial.
Third, such non-compliance with the district court’s orders certainly needed to be deterred. The district court’s condonation of CMS’s bad faith at a time so close to the beginning of trial could have encouraged repetition of improper conduct. As found by the district court, the record indicates that the failure to supplement interrogatories was not the first time CMS “was lacking in candor in disclosing relevant and important information.” J.A. 1-305. Hence, deterrence was essential to a proper management of this case.
Finally, less drastic sanctions would not have been effective. Permitting the plaintiff-intervenors to depose witnesses and requiring CMS to pay fees and expenses for the depositions was appropriate. CMS was fortunate to receive such a light sanction, and it is doubtful whether lesser measures would have had any effect on CMS’s conduct.
In sum, the discovery sanctions imposed did not amount to an abuse of the district court’s discretion.
*349VI. Attorney Fees
CMS argues that the district court erred in awarding attorney fees to the plaintiff-intervenors. While conceding that Grant is entitled to fees if the district court’s unitary status finding is upheld, CMS argues that Capacchione cannot be a prevailing party on this issue. CMS also challenges Capacchione’s receipt of fees based on the district court’s magnet schools ruling because (1) Capacchione received only nominal damages, and (2) young Capac-chione would not have been admitted to the magnet program even if race was not a factor insofar as her lottery number was so high. The district court’s decision to award attorney fees is reviewed for an abuse of discretion. See Hitachi Credit Am. Corp. v. Signet Bank, 166 F.3d 614, 631 (4th Cir.1999). I would affirm.
A. Attorney Fees for Unitary Status
1.
In my view, the plaintiff-intervenors are entitled under § 1988 to attorney fees for their successful litigation of the unitary status issue. Indeed, CMS has conceded that if we upheld the declaration of unitary status, Grant would be entitled to attorney fees.
Unlike Capacchione, the Grant interve-nors were granted declaratory and in-junctive relief related to the issues of unitary status and CMS’ magnet school admission policies. Therefore, the entitlement of the Grant intervenors to recover attorneys’ fees is tied directly to the merits of those claims.
CMS’s Brief at 39-40. Surprisingly, despite CMS’s concession, a majority of this court vacates the award of fees to both Capacchione and Grant.
By obtaining a declaration of unitary status, the Grant plaintiffs, along with Ca-pacchione, finished what the original Swann plaintiffs started. If we deny the plaintiff-intervenors the ability to be compensated in a situation such as this— where an incredible amount of legal work is required and the board, for improper reasons, clings to the court’s order — then we give to litigants like the Swann plaintiffs effective control over the decision of “when” or even “if’ a unitary status hearing will be sought because they would be the only ones who could ever obtain reimbursement for their legal fees. The practical consequences are scarcely more apparent than in this case, where the only party ruled entitled to obtain attorney fees for finishing the job was opposed to seeing unitary status declared. This, coupled with the fact that the plaintiff-intervenors received a court order in their favor on the unitary status question as a continuation of the original § 1983 action, leaves me at a loss to see how the district court’s award of attorney fees to them under § 1988 can be reversed.
Under 42 U.S.C.A. § 1988(b) (West Supp.2000), “[i]n any action or proceeding to enforce a provision of [§ 1983 and other civil rights laws] ... the court, in its discretion, may allow the prevailing party, other than the United States, a reasonable attorney’s fee as part of the costs.” To be considered a prevailing party, a party must “succeed on any significant issue in litigation which achieves some of the benefit the parties sought in bringing suit.” Farrar v. Hobby, 506 U.S. 103, 109, 113 S.Ct. 566, 121 L.Ed.2d 494 (1992) (internal quotation marks omitted).
This case began in 1965 as a § 1983 action with the Swann plaintiffs seeking conversion of CMS “into a unitary nonracial system wherein the educational opportunities offered by [the board] are made available to students without regard to race or color.” J.A. XXXIII-16,162 (original Swann complaint commencing an ac*350tion under § 1988); see also Monell v. Department of Social Servs., 436 U.S. 658, 697, 98 S.Ct. 2018, 56 L.Ed.2d 611 (1978) (observing that school desegregation actions “have almost without exception been § 1983 suits”). In essentially a continuation of what was begun in 1965, Capac-chione, believing that CMS had established a unitary school system, brought suit pursuant to § 1983 and prayed that the district court enter a declaration of unitary status. See J.A. I — 110 (Capacchione amended complaint seeking a declaration of unitary status); see also J.A. 1-140 (Grant complaint seeking a declaration of unitary status); cf. Waste Mgmt. Holdings, Inc. v. Gilmore, 252 F.3d 316 (4th Cir.2001) (§ 1983 action seeking declaratory relief). Shortly after Capacchione filed suit, the Swann plaintiffs moved to reactivate Swann and to consolidate it with Ca-pacchione’s action. The district court granted the Swann plaintiffs’ motion and later permitted Capacchione to intervene in Swann. Grant, who also sought a declaration of unitary status, then moved to intervene in the consolidated action, and the district court granted his motion. After months of litigation, the plaintiff-inter-venors succeeded in having CMS declared unitary, and this court has affirmed on appeal.
With the prior court orders now dissolved, CMS must cease using the orders to assign Grant’s children as well as all other public school students on account of their race.16 A unitary school system is what the original plaintiffs sought in 1965, and the plaintiff-intervenors have greatly assisted in the final stages of this litigation in making unitariness a reality. Henceforth, unless CMS’s use of racial classifications satisfies the requirements of strict scrutiny, the color of a child’s skin will no longer be a permissible basis for assigning, or refusing to assign, a child to a conventional public school or a specialized magnet program. See Farrar, 506 U.S. at 110, 113 S.Ct. 566 (observing that declaratory relief may constitute relief under § 1988 “if[ ] it affects the behavior of the defendant toward the plaintiff’) (internal quotation marks omitted). Moreover, this declaration of unitary status is enforceable against CMS in the unlikely event it later attempts to continue prior assignment polices, say, on the ground that vestiges of prior discrimination have not been eradicated. See id. at 111, 113 S.Ct. 566 (explaining that to be a prevailing party an enforceable judgment must be obtained).
In the present case, with its counter-intuitive alignment of parties, the plaintiff-intervenors have stepped into the shoes of the Swann plaintiffs, continued the original § 1983 action, and brought this case to a close. Just as the Swann plaintiffs, prior to removal of the case from the active docket, were compensated for their services, see Swann v. Charlotte-Mecklenburg Bd. of Educ., 66 F.R.D. 483 (W.D.N.C.1975) (awarding the Swann plaintiffs $204,072.33 in fees and costs), so too *351should the plaintiff-intervenors be compensated for a continuation of the legal efforts to achieve a unitary school system and to remove federal court control. In this regard the plaintiff-intervenors have acted as “private attorney[s] general,” Independent Fed’n of Flight Attendants v. Zipes, 491 U.S. 754, 758-60, 109 S.Ct. 2732, 105 L.Ed.2d 639 (1989) (internal quotation marks omitted) (alteration in original), and obtained a finding of unitariness, which was the ultimate objective of the original action. See J.A. XXXIII-16,162 (original complaint seeking “reorganization of the school system into a unitary nonracial system”). Indeed, no one disputes that the Swann plaintiffs would have been entitled to fees if they had taken the initiative to petition for a declaration of unitary status instead of acting to oppose the result we reach today.
Without question, the monitoring of a school desegregation decree is crucial to the dismantling of the dual system. See Jenkins v. Missouri, 967 F.2d 1248, 1251 (8th Cir.1992) (awarding fees in desegregation case pursuant to § 1988). Accordingly, efforts “to insure full compliance and to ensure that the plan is indeed working to desegregate the school systemf] are com-pensable services.” Northcross v. Board of Educ., 611 F.2d 624, 637 (6th Cir.1979). Here, the plaintiff-intervenors observed CMS’s progress in dismantling the dual system, and once convinced that full compliance had been achieved, they moved for a declaration of unitary status. ’ In so doing, the plaintiff-intervenors were faced with a recalcitrant school board that insisted none of the Green factors had been satisfied. The Sivann plaintiffs, though having never returned to court to complain about the continuation or revival of segre-gative practices, suddenly claimed that the dual system was not being dismantled and joined the school board in the quest for continuation of court supervision. Though most of the vital information was in the hands of CMS’s officials, who were often uncooperative in the discovery process, see Capacchione, 57 F.Supp.2d. at 292-293 (cataloging sanctions and threats of sanctions against CMS), the plaintiff-intervenors persevered and ultimately obtained a declaration of unitary status. But for the actions of the plaintiff-intervenors, CMS, though having dismantled the dual system, would still be using the district court’s orders as mechanisms for attaining other goals. Despite the progress of the last three decades, CMS was apparently content to forestall a finding of unitariness for the foreseeable future. See Freeman, 503 U.S. at 490, 112 S.Ct. 1430 (observing that restoration of local control “at the earliest practicable date” is a goal of any desegregation order). Consequently, I believe that the plaintiff-intervenors, for stepping in and finishing what was begun in 1965, are entitled to attorney fees under § 1988 for their litigation of the unitary status issue.
2.
I would also affirm the plaintiff-interve-nors’ award of attorney fees, based on the unitary status declaration, under this circuit’s exceptional circumstances doctrine. See Rolax v. Atlantic Coast Line R. Co., 186 F.2d 473, 481 (4th Cir.1950) (holding that absent a statute attorney fees are normally unavailable unless “the taxation of such costs is essential to the doing of justice ... in exceptional cases”). To avoid a declaration of unitary status, CMS has clung to the desegregation decree for improper reasons, see supra part II.H, and the equitable remedy ordered in 1969 “would be far from complete, and justice would not be attained, if reasonable counsel fees were not awarded” to the plaintiff-intervenors. Bell v. School Bd. of Powhatan County, 321 F.2d 494, 500 (4th Cir.*3521963) (en banc) (awarding attorney fees in school desegregation case based on exceptional circumstances when the school board engaged in a “pattern of evasion and obstruction” which “cast[ ] a heavy burden on the children and their parents”). A contrary result would hamper the involvement of concerned citizens in school desegregation litigation and permit school boards that are inclined to remain under court order to eschew a unitary status hearing.
I recognize that the Supreme Court recently rejected the catalyst theory as a basis for awarding attorney fees. See Buckhannon Bd. & Care Home v. West Virginia Dep’t of Health & Human Res., 532 U.S. 598, 121 S.Ct. 1835, 149 L.Ed.2d 855 (2001). However, the award of fees in the present case has never been based on the catalyst theory, “which posits that a plaintiff is a prevailing party if it achieves the desired result because the lawsuit brought about a voluntary change in the defendant’s conduct.” Id. at 1838 (internal quotation marks omitted). In this case, there was no voluntary change in CMS’s conduct. CMS clung to the desegregation orders and put up a vigorous defense in the course of a two-month trial. A final judgment was handed down, and any change in CMS’s behavior will be due to the district court’s decree, not a voluntary act. While a “voluntary change in conduct ... lacks the necessary judicial imprimatur on the change” for a plaintiff to be considered a prevailing party, a declaration of unitary status is far different. Id. at 1840. Once found to be in violation of the Constitution, a school district cannot be declared unitary without the order of a court. Because the district court’s order, and not a voluntary act, is the impetus behind any change, the rejection of the catalyst theory in Buckhannon does not undermine an award of attorney fees based on the exceptional circumstances doctrine of Rolax.
B. Attorney Fees for the Magnet Schools Litigation
I would also find that Capacchione is entitled to fees because he is a prevailing party on the magnet schools issue. The district court held that the magnet schools admissions policy violated the Equal Protection Clause of the Fourteenth Amendment and awarded nominal damages in light of the constitutional violation. In Farrar, the Supreme Court specifically addressed the issue of nominal damages and prevailing party status:
We therefore hold that a plaintiff who wins nominal damages is a prevailing party under § 1988.... A plaintiff may demand payment for nominal damages no less than he may demand payment for millions of dollars in compensatory damages. A judgment for damages in any amount, whether compensatory or nominal, modifies the defendant’s behavior for the plaintiffs benefit by forcing the defendant to pay an amount of money he otherwise would not pay.
506 U.S. at 112-13, 113 S.Ct. 566 (internal citations omitted). The award of nominal damages constitutes relief on the merits and affects CMS’s behavior toward Capacchione if only by forcing CMS to pay. Hence, Capacchione is a prevailing party. See also Shaw v. Hunt, 154 F.3d 161, 164 (4th Cir.1998) (noting that “persons within the generic category of plaintiff-interve-nors have often been found by courts to fit within the rubric ‘prevailing party’ for fees purposes”). That young Capacchione had a high lottery number is irrelevant for a determination of prevailing party status. As previously stated, the injury in this case was the inability to compete for open magnet seats, not the denial of admission to a magnet program. Because I would *353find that Capaechione rightly prevailed on the magnet schools issue, I would affirm the district court’s award of attorney fees for work in this area as well.
VII.
For the foregoing reasons, a majority of this court affirms the district court’s declaration of unitary status and the imposition of discovery sanctions. We vacate the district court’s injunction because we can discern no danger of future violations. Additionally, I would affirm the finding of a constitutional violation in the magnet schools admissions policy, the award of nominal damages, and the attorney fees awarded pursuant to 42 U.S.C.A. § 1988.
. Those represented by Capacchione and Grant will be referred to as the "plaintiff-intervenors."
. At trial, Dr. Eric Smith, the current superintendent of CMS, testified that unitary status depended on every school being in balance. See J.A. XV-7187 & 7239. This is not the law. See Swann, 402 U.S. at 24, 91 S.Ct. 1267 ("The constitutional command to desegregate schools does not mean that every school in every community must always reflect the racial composition of the school system as ao whole.”).
We find equally erroneous the Swann plaintiffs’ assertion at appellate argument before the panel that "[t]he issue of how many schools are balanced has never been a question in this case.” App. Tr. 91. The racial composition of schools goes to the heart of a desegregation case, and is very much key to a review of the district court’s declaration of unitary status. See Swann, 311 F.Supp. at 268 (ordering CMS to assign pupils "in such a way that as nearly as practicable the various schools at various grade levels have about the same proportion of black and white students”).
. For example, the population of DeKalb County grew from 70,000 in 1950 to 450,000 in 1985, and the percentage of black students in the district grew from 5.6 percent in 1969 to forty-seven percent in 1986. See Freeman, 503 U.S. at 475, 112 S.Ct. 1430.
. Dr. Armor did not include the predominantly white schools in this analysis on three grounds:
(1) the court order did not establish a minimum percent black enrollment, (2) the half-dozen schools that have had low black enrollment for the past three or more years and that were operating in 1972 have been racially balanced for at least ten years[,] and (3) the demographic analysis of Dr. Clark shows that these schools have become imbalanced or were opened imbalanced because of the substantial white enrollment growth in the outskirts of the county.
J.A. XXXIII-16,174 (footnote omitted); see also Swann, 402 U.S. at 26, 91 S.Ct. 1267 (observing "that the existence of some small number of one-race, or virtually one-race, schools within a district is not in and of itself the mark of a system that still practices segregation by law”).
. Even the Swann plaintiffs admit that ten years after the district court charged the board with taking affirmative steps to desegregate schools, the system "w[as] nearly 100% statistically compliant with the court’s orders.” Plaintiff Appellants’ Brief at 38.
. Given the counter-intuitive alignment of the parties in this case, it could be argued that the presumption and burden allocation set forth in Balites should not be applied, and that CMS should instead be required to prove the existence of racial disparity in its facilities. See United States v. City of Yonkers, 181 F.3d 301, 309-11 (2d Cir.1999), vacated on reh'g, 197 F.3d 41 (2d Cir.1999).
. From this footnote and the district court's detailed discussion about the cause of any disparity in CMS’s facilities, it appears that the district court really made alternative rulings on the facilities question: The court first concluded that CMS and the Swann plaintiffs bore the burden of proof with regard to facilities and that they failed to carry that burden. See Capacchione, 57 F.Supp.2d at 267 (”[T]he Swann Plaintiffs have failed to overcome the Court’s previous findings on facilities by establishing the requisite discriminatory intent and causation.”). The court then ruled in the alternative, as indicated by the footnote and the findings, that the plaintiff-intervenors proved that any disparities were the result of factors unrelated to state action.
. Despite evidence that the achievement gap results from factors outside CMS's control, the district court found that CMS has undertaken sundry measures to eliminate the gap. For example, CMS adopted financial incentives for teachers and principals tied to student performance, urged black students to take advanced placement and other higher level classes, challenged all students by removing "fluff courses” from the curriculum, provided tutors and other forms of staff support to accelerate student preparedness, and *332adopted pre-kindergarten programs to accelerate preparedness for the youngest of students. See Capacchione, 57 F.Supp.2d at 273-275.
. My colleagues in the majority on this issue eloquently argue that CMS was permitted to take race-conscious measures when complying with desegregation orders. With this I agree — a school district under order to desegregate must of course take race into account *340when assigning students. The primary question regarding the magnet program, however, is whether CMS ran afoul of the Supreme Court's prohibitions against inflexible ratios, not whether race-conscious measures are permissible.
. 1 also disagree with the assertion that the Supreme Court’s disapproval of inflexible racial quotas as a desegregation tool is solely a limitation on a district court’s remedial power. While the Swann Court did imply that a school board, exercising its discretion, could "conclude ... that in order to prepare students to live in a pluralistic society each school should have a prescribed ratio of Negro to white students reflecting the proportion for the district as a whole,” 402 U.S. at 16, 91 S.Ct. 1267, this is certainly not the state of the law today nor was it the state of the law in 1992 when the magnet plan was adopted. At the very least, the Supreme Court decisions in Wygant v. Jackson Board of Education, 476 *341U.S. 267, 283, 106 S.Ct. 1842, 90 L.Ed.2d 260 (1986) (plurality opinion applying strict scrutiny to a school board's race-based layoff program), and City of Richmond v. J.A. Croson Co., 488 U.S. 469, 494, 109 S.Ct. 706, 102 L.Ed.2d 854 (1989) (applying strict scrutiny to a racial set-aside program), should have alerted CMS that it could not rely on the “pluralistic society” passage from the 1971 opinion when crafting a magnet admissions policy that was outside the scope of the desegregation orders. By 1992 such a use of race was not merely discretionary. Prevailing case law required that the racial classification be narrowly tailored to achieve a compelling state interest. See J.A. Croson Co., 488 U.S. at 494, 109 S.Ct. 706. And as demonstrated in section III.B, the admissions policy was in no sense narrowly tailored.
. In contending that rigid ratios were not used by CMS, several of my colleagues observe that not a single magnet school achieved the precise ratio of sixty percent white students and forty percent black students. This is not surprising insofar as the policy was designed to leave seats vacant. The very act of leaving seats vacant will compel a deviation from the stated goal. However, this in no way undermines a finding of rigidity. Instead, such a result illustrates the policy’s inflexibility.
. I recognize that parents might perceive that one "fungible” conventional school is superior to another because of a number of intangibles such as the reputation of teachers or the newness of facilities. However, these "personal preferences” do not rise to a level of constitutional significance. See Hampton v. Jefferson County Bd. of Educ., 102 F.Supp.2d 358, 380 n. 43 (W.D.Ky.2000). Magnet schools, on the other hand, are a completely different animal and therefore the admissions process used must be more closely scrutinized.
. Though the present case was brought on behalf of a white child denied admission to a magnet school, the policy as written could have just as easily denied a black child admission to the magnet school. See Hampton v. Jefferson County Bd. of Educ., 102 F.Supp.2d 358, 377 (W.D.Ky.2000) (racial quota in a magnet school resulting in black students being denied admission even though the school was several hundred students below capacity).
. The Supreme Court’s application of strict scrutiny has indeed been unwavering. In Adarand, the Court refused to apply a lesser standard of scrutiny to racial classifications enacted by Congress. Though Congress itself is charged with enforcing the Fourteenth Amendment's promise of equal protection via "appropriate legislation,” U.S. Const, amend. XIV, § 5, the Supreme Court in interpreting the Fifth Amendment held Congress to the same rigorous standards applicable to states and localities. See Adarand, 515 U.S. at 224, 115 S.Ct. 2097 (observing "that any person, of whatever race, has the right to demand that any governmental actor subject to the Constitution justify any racial classification subjecting that person to unequal treatment under the strictest judicial scrutiny”).
CMS and the Swann plaintiffs contend that strict scrutiny does not apply when a school district is under court order to dismantle the dual system. Such an approach, however, ignores two of the three pillars of Supreme Court's equal protection analysis: skepticism of all racial preferences and consistent application of heightened scrutiny regardless of the race of the person burdened or benefitted. See Adarand, 515 U.S. at 223-24, 115 S.Ct. 2097. Contrary to the assertions of CMS and the Swann plaintiffs, the approach I would adopt does not deprive a school board under court order of the necessary tools required to establish a unitary school system.
The point of carefully examining the interest asserted by the government in support of a racial classification, and the evidence offered to show that the classification is needed, is precisely to distinguish legitimate from illegitimate uses of race in governmental decisionmaking.... Strict scrutiny does not “trea[t] dissimilar race-based decisions as though they were equally objectionable”; to the contrary, it evaluates carefully all governmental race-based deci*344sions in order to decide which are constitutionally objectionable and which are not.
Id. at 228, 115 S.Ct. 2097 (internal citations omitted) (alteration in original). This careful evaluation demanded by the Supreme Court will preserve inviolate proper desegregation remedies while ensuring that in the process of desegregating a government actor does not stand equal protection on its head by denying some students educational opportunities solely because of their race.
. CMS also presented diversity as an alternative compelling state interest. See Capacchione, 57 F.Supp.2d at 289. In this circuit, it is unsettled whether diversity may be a compelling state interest. See Eisenberg v. Montgomery County Pub. Schs., 197 F.3d 123, 130 (4th Cir.1999), cert. denied, 529 U.S. 1019, 120 S.Ct. 1420, 146 L.Ed.2d 312 (2000). Assuming without deciding whether diversity may be a compelling state interest, I would hold that-the magnet admissions policy again fails because it is not narrowly tailored. Whether the interest is remedying past discrimination or diversity, the admissions policy as currently written is in no sense narrow. It is difficult to imagine any interest for which the magnet admissions policy is narrowly tailored.
. CMS argues that because Capacchione no longer resides in North Carolina the unitary status declaration does not alter CMS's behavior toward young Capacchione and consequently Capacchione is not a prevailing party entitled to fees. In other words, CMS contends that Capacchione's lack of standing counsels against an award of fees. I disagree. At the very least, because of Capacchione’s status as a plaintiff-intervenor in Swann, Ca-pacchione is still entitled to fees. See Shaw v. Hunt, 154 F.3d 161, 167 (4th Cir.1998) (noting that when plaintiffs with standing "secure[ ] precisely the relief that they sought,” plaintiff-intervenors who lack standing but contributed to the litigation may also be awarded attorney fees). Grant, who CMS concedes has standing and is entitled to fees, achieved the relief originally sought — a declaration of unitary status. Capacchione greatly contributed to this result, and under Shaw is entitled to fees just as Grant.