with whom Justice Stevens and Justice O’Connor join, concurring in the judgment.
It is almost 38 years since this Court decided Brown v. Board of Education, 347 U. S. 483 (1954). In those 38 years the students in DeKalb County, Ga., never have attended a desegregated school system even for one day. The majority of “black” students never have attended a school that was not disproportionately black. Ignoring this glaring dual character of the DeKalb County School System (DCSS), part “white” and part “black,” the District Court relinquished control over student assignments, finding that the school district had achieved “unitary status” in that aspect of the system. No doubt frustrated by the continued existence of duality, the Court of Appeals ordered the school district to take extraordinary measures to correct all manifestations of this racial imbalance. Both decisions, in my view, were in error, and I therefore concur in the Court’s decision to vacate the judgment and remand the case.
I also am in agreement with what I consider to be the holdings of the Court. I agree that in some circumstances the District Court need not interfere with a particular portion of the school system, even while, in my view, it must retain jurisdiction over the entire system until all vestiges of state-imposed segregation have been eliminated. See ante, at 490-491. I also agree that whether the District Court must order DCSS to balance student assignments depends on whether the current imbalance is traceable to unlawful state policy and on whether such an order is necessary to fashion an effective remedy. See ante, at 491, 493-494, 497-498. Finally, I agree that the good faith of the school board is relevant to these inquiries. See ante, at 498-499.
*510I write separately for two purposes. First, I wish to be precise about my understanding of what it means for the District Court in this case to retain jurisdiction while relinquishing “supervision and control” over a subpart of a school system under a desegregation decree. Second, I write to elaborate on factors the District Court should consider in determining whether racial imbalance is traceable to board actions and to indicate where, in my view, it failed to apply these standards.
I
Beginning with Brown, and continuing through the Court’s most recent school-desegregation decision in Board of Ed. of Oklahoma City Public Schools v. Dowell, 498 U. S. 237 (1991), this Court has recognized that when the local government has been running de jure segregated schools, it is the operation of a racially segregated school system that must be remedied, not discriminatory policy in some discrete subpart of that system. Consequently, the Court in the past has required, and decides again today, that even if the school system ceases to discriminate with respect to one of the Green-type factors, “the [district] court should retain jurisdiction until it is clear that state-imposed segregation has been completely removed.” Green v. School Bd. of New Kent County, 391 U. S. 430, 439 (1968) (emphasis added); Raney v. Board of Ed. of Gould School Dist., 391 U. S. 443, 449 (1968); see ante, at 491.
That the District Court’s jurisdiction should continue until the school board demonstrates full compliance with the Constitution follows from the reasonable skepticism that underlies judicial supervision in the first instance. This Court noted in Dowell: “A district court need not accept at face value the profession of a school board which has intentionally discriminated that it will cease to do so in the future.” 498 U. S., at 249. It makes little sense, it seems to me, for the court to disarm itself by renouncing jurisdiction in one aspect of a school system, while violations of the Equal Protec*511tion Clause persist in other aspects of the same system. Cf. Keyes v. School Dist. No. 1, Denver, 413 U. S. 189, 207 (1973). It would seem especially misguided to place unqualified reliance on the school board’s promises in this case, because the two areas of the school system the District Court found still in violation of the Constitution — expenditures and teacher assignments — are two of the Green factors over which DCSS exercises the greatest control.
The obligations of a district court and a school district under its jurisdiction have been clearly articulated in the Court’s many desegregation cases. Until the desegregation decree is dissolved under the standards set forth in Dowell, the school board continues to have “the affirmative duty to take whatever steps might be necessary to convert to a unitary system in which racial discrimination would be eliminated root and branch.” Green, 391 U. S., at 437-438. The duty remains enforceable by the district court without any new proof of a constitutional violation, and the school district has the burden of proving that its actions are eradicating the effects of the former de jure regime. See Dayton Board of Education v. Brinkman, 443 U. S. 526, 537 (1979); Keyes, 413 U. S., at 208-211; Swann v. Charlotte-Mecklenburg Board of Education, 402 U. S. 1, 26 (1971); Green, 391 U. S., at 439.
Contrary to the Court of Appeals’ conclusion, however, retaining jurisdiction does not obligate the district court in all circumstances to maintain active supervision and control, continually ordering reassignment of students. The “duty” of the district court is to guarantee that the school district “ ‘eliminate^] the discriminatory effects of the past as well as to bar like discrimination in the future.’” Green, 391 U. S., at 438, n. 4. This obligation requires the court to review school-board actions to ensure that each one “will further rather than delay conversion to a unitary, nonracial nondiscriminatory school system.” Monroe v. Board of Comm’rs of Jackson, 391 U. S. 450, 459 (1968); see also Dayton Board of Education, 443 U. S., at 538; United States v.
*512Scotland Neck Board of Education, 407 U. S. 484, 489 (1972). But this obligation does not always require the district court to order new, affirmative action simply because of racial imbalance in student assignment.
Whether a district court must maintain active supervision over student assignment, and order new remedial actions, depends on two factors. As the Court discusses, the district court must order changes in student assignment if it “is necessary or practicable to achieve compliance in other facets of the school system.” Ante, at 497; see also ante, at 507 (Sou-ter, J., concurring). The district court also must order affirmative action in school attendance if the school district’s conduct was a “contributing cause” of the racially identifiable schools. Columbus Board of Education v. Penick, 443 U. S. 449, 465, n. 13 (1979); see also Keyes, 413 U. S., at 211, and n. 17 (the school board must prove that its conduct “did not create or contribute to” the racial identifiability of schools or that racially identifiable schools are “in no way the result of” school board action). It is the application of this latter causation requirement that I now examine in more detail.
HH
A
DCSS claims that it need not remedy the segregation in DeKalb County schools because it was caused by demographic changes for which DCSS has no responsibility. It is not enough, however, for DCSS to establish that demographics exacerbated the problem; it must prove that its own policies did not contribute.1 Such contribution can occur in at *513least two ways: DCSS may have contributed to the demographic changes themselves, or it may have contributed directly to the racial imbalance in the schools.
To determine DCSS’ possible role in encouraging the residential segregation, the court must examine the situation with special care. “[A] connection between past segregative acts and present segregation may be present even when not apparent and . . . close examination is required before concluding that the connection does not exist.” Keyes, 413 U. S., at 211. Close examination is necessary because what might seem to be purely private preferences in housing may in fact have been created, in part, by actions of the school district.
“People gravitate toward school facilities, just as schools are located in response to the needs of people. The location of schools may thus influence the patterns of residential development of a metropolitan area and have important impact on composition of inner-city neighborhoods.” Swann, 402 U. S., at 20-21.
This interactive effect between schools and housing choices may occur because many families are concerned about the racial composition of a prospective school and will make residential decisions accordingly.2 Thus, schools that are demonstrably black or white provide a signal to these families, perpetuating and intensifying the residential movement. See Keyes, 413 U. S., at 202; Columbus Board of Education, 443 U. S., at 465, n. 13; ante, at 507-508 (Souter, J., concurring).
School systems can identify a school as “black” or “white” in a variety of ways; choosing to enroll a racially identifiable *514student population is only the most obvious. The Court has noted: “[T]he use of mobile classrooms, the drafting of student transfer policies, the transportation of students, and the assignment of faculty and staff, on racially identifiable bases, have the clear effect of earmarking schools according to their racial composition.” Keyes, 413 U. S., at 202. Because of the various methods for identifying schools by race, even if a school district manages to desegregate student assignments at one point, its failure to remedy the constitutional violation in its entirety may result in resegregation, as neighborhoods respond to the racially identifiable schools. See ante, at 508-509 (Souter, J., concurring). Regardless of the particular way in which the school district has encouraged residential segregation, this Court’s decisions require that the school district remedy the effect that such segregation has had on the school system.
In addition to exploring the school district’s influence on residential segregation, the District Court here should examine whether school-board actions might have contributed to school segregation. Actions taken by a school district can aggravate or eliminate school segregation independent of residential segregation. School-board policies concerning placement of new schools and closure of old schools and programs such as magnet classrooms and majority-to-minority (M-to-M) transfer policies affect the racial composition of the schools. See Swann, 402 U. S., at 20-21, 26-27. A school district’s failure to adopt policies that effectively desegregate its schools continues the violation of the Fourteenth Amendment. See Columbus Board of Education, 443 U. S., at 458-459; Dayton Board of Education, 443 U. S., at 538. The Court many times has noted that a school district is not responsible for all of society’s ills, but it bears full responsibility for schools that have never been desegregated. See, e. g., Swann, supra.
*515B
The District Court’s opinion suggests that it did not examine DCSS’ actions in light of the foregoing principles. The court did note that the migration farther into the suburbs was accelerated by “white flight” from black schools and the “blockbusting” of former white neighborhoods. It did not examine, however, whether DCSS might have encouraged that flight by assigning faculty and principals so as to identify some schools as intended respectively for black students or white students. See App. 226-231. Nor did the court consider how the placement of schools, the attendance zone boundaries, or the use of mobile classrooms might have affected residential movement. The court, in my view, failed to consider the many ways DCSS may have contributed to the demographic shifts.
Nor did the District Court correctly analyze whether DCSS’ past actions had contributed to the school segregation independent of residential segregation. The court did not require DCSS to bear the “heavy burden” of showing that student assignment policies — policies that continued the effects of the dual system — served important and legitimate ends. See Dayton Board of Education, 443 U. S., at 538; Swann, 402 U. S., at 26. Indeed, the District Court said flatly that it would “not dwell on what might have been,” but would inquire only as to “what else should be done now.” App. 221. But this Court’s decisions require the District Court to “dwell on what might have been.” In particular, they require the court to examine the past to determine whether the current racial imbalance in the schools is attributable in part to the former de jure segregated regime or any later actions by school officials.
As the Court describes, the District Court placed great emphasis on its conclusion that DCSS, in response to the court order, had desegregated student assignment in 1969. DCSS’ very first action taken in response to the court decree, however, was to shape attendance zones to result *516in two schools that were more than 50% black, despite a district-wide black student population of less than 6%. See ante, at 477-478. Within a year, another school became majority black, followed by four others within the next two years. App. 304, 314, 350, 351, 368. Despite the existence of these schools, the District Court found that DCSS effectively had desegregated for a short period of time with respect to student assignment. See ante, at 478. The District Court justified this finding by linking the school segregation exclusively to residential segregation existing prior to the court order. See ibid.
But residential segregation that existed prior to the desegregation decree cannot provide an excuse. It is not enough that DCSS adopt race-neutral policies in response to a court desegregation decree. Instead, DCSS is obligated to “counteract the continuing effects of past school segregation.” Swann, 402 U. S., at 28. Accordingly, the school district did not meet its affirmative duty simply by adopting a neighborhood-school plan, when already existing residential segregation inevitably perpetuated the dual system. See Davis v. Board of School Comm’rs of Mobile County, 402 U. S. 33, 37 (1971); Swann, 402 U. S., at 25-28, 30.
Virtually all the demographic changes that DCSS claims caused the school segregation occurred after 1975. See ante, at 475; App. 215, 260. Of particular relevance to the causation inquiry, then, are DCSS’ actions prior to 1975; failures during that period to implement the 1969 decree render the school district’s contentions that its noncompliance is due simply to demographic changes less plausible.
A review of the record suggests that from 1969 until 1975, DCSS failed to desegregate its schools. During that period, the number of students attending racially identifiable schools actually increased, and increased more quickly than the increase in black students. By 1975, 73% of black elementary students and 56% of black high school students were attending majority black schools, although the percentages of black *517students in the district population were just 20% and 13%, respectively. Id., at 269-380.
Of the 13 new elementary schools DCSS opened between 1969 and 1975, 6 had a total of four black students in 1975. Id., at 272, 299, 311, 316, 337, 353. One of the two high schools DCSS opened had no black students at all.3 Id., at 367, 361. The only other measure taken by DCSS during the 1969-1975 period was to adopt the M-to-M transfer program in 1972. Due, however, to limitations imposed by school-district administrators — including a failure to provide transportation, “unnecessary red tape,” and limits on available transfer schools — only one-tenth of 1% of the students were participating in the transfer program as of the 1975-1976 school year. Id., at 75, 80.
In 1976, when the District Court reviewed DCSS’ actions in the M-to-M program, it concluded that DCSS’ limitations on the program “perpetuate the vestiges of a dual system.” Id., at 83. Noting that the Department of Health, Education, and Welfare had found that DCSS had ignored its responsibility affirmatively to eradicate segregation and perpetuate desegregation, the District Court found that attendance zone changes had perpetuated the dual system in the county. Id., at 89, 91.
Thus, in 1976, before most of the demographic changes, the District Court found that DCSS had not complied with the 1969 order to eliminate the vestiges of its former de jure school system. Indeed, the 1976 order found that DCSS had contributed to the growing racial imbalance of its schools. Given these determinations in 1976, the District Court, at a minimum, should have required DCSS to prove that, but for the demographic changes between 1976 and 1985, its actions would have been sufficient to “convert promptly to a system without a ‘white’ school and a ‘Negro’ school, but just *518schools.” Green, 391 U. S., at 442. The available evidence suggests that this would be a difficult burden for DCSS to meet.
DCSS has undertaken only limited remedial actions since the 1976 court order. The number of students participating in the M-to-M program has expanded somewhat, composing about 6% of the current student population. The district also has adopted magnet programs, but they involve fewer than 1% of the system’s students. Doubtless DCSS could have started and expanded its magnet and M-to-M programs more promptly; it could have built and closed schools with a view toward promoting integration of both schools and neighborhoods; redrawn attendance zones; integrated its faculty and administrators; and spent its funds equally. But it did not. DCSS must prove that the measures it actually implemented satisfy its obligation to eliminate the vestiges of de jure segregation originally discovered in 1969, and still found to exist in 1976.
Ill
The District Court apparently has concluded that DCSS should be relieved of the responsibility to desegregate because such responsibility would be burdensome. To be sure, changes in demographic patterns aggravated the vestiges of segregation and made it more difficult for DCSS to desegregate. But an integrated school system is no less desirable because it is difficult to achieve, and it is no less a constitutional imperative because that imperative has gone unmet for 38 years.
Although respondents challenged the District Court’s causation conclusions in the Court of Appeals, that court did not reach the issue. Accordingly, in addition to the issues the Court suggests be considered in further proceedings, I would remand for the Court of Appeals to review, under the foregoing principles, the District Court’s finding that DCSS has met its burden of proving the racially identifiable schools are in no way the result of past segregative action.
The Court’s cases make clear that there is a presumption in a former de jure segregated school district that the board’s actions caused the racially identifiable schools, and it is the school board’s obligation to rebut that presumption. See Dayton Board of Education v. Brinkman, 443 U. S. 526, 537 (1979); Keyes v. School Dist. No. 1, Denver, 413 U. S. 189, 208, 211 (1973); Swann v. Charlotte-Mecklenburg Board of Education, 402 U. S. 1, 26 (1971); ante, at 494-495.
See Taeuber, Housing, Schools, and Incremental Segregative Effects, 441 Annals Am. Acad. Pol. & Soc. Sci. 157 (1979); Orfield, School Segregation and Residential Segregation, in School Desegregation: Past, Present, and Future 227, 234-237 (W. Stephan & J. Feagin eds. 1980); Elam, The 22nd Annual Gallup Poll of Public’s Attitudes Toward the Public Schools, 72 Phi Delta Kappan 41, 44-45 (1990).
By 1986, one of those two high schools was 2.4% black. The other was 91.7% black. Of the 13 elementary schools, 8 were either virtually all black or all white and all were racially identifiable. App. 269-359.