concurring in part and dissenting in part,
joined by FAGG, Circuit Judge.Although I agree with much of what the court does today, I do not believe there is sufficient evidence in the record to support that part of the court’s decision requiring *444that the boundary lines of the City of Little Rock and the LRSD be made coterminous. Therefore, I respectfully dissent. Furthermore, while I agree with nearly all of Judge Arnold’s persuasive discussion and reasoning, I believe that his opinion likewise fails to demonstrate that there has been sufficient proof of significant interdistrict segregative effects to justify realigning the boundaries of the LRSD and the City of Little Rock.
The record makes plain, and I fully concur in the court’s conclusion, that there have been substantial and in fact egregious intradistrict constitutional violations as a result of segregative acts on the part of the NLRSD and the PCSSD, the effects of which must be remedied. I also agree with the court’s conclusion that the record does not reveal sufficient interdistrict segregative effects to justify consolidation of the three districts.
I further agree with the court today that the Granite Mountain transfer, which occurred in 1953, had an interdistrict segregative effect. However, I believe that the current effects of this violation are not clearly delineated either in the district court’s findings of fact and conclusions of law, or in this court’s conclusions today. Contrary to the court’s decision today, as well as Judge Arnold’s views, I believe that this issue should be remanded to the district court for more precise and specific findings as to the current interdistrict segregative effect of the 1953 transfer, and for consideration of an appropriate remedy tailored to the constitutional violation that is found to exist.
My chief concern with the opinion of the court is that it reads too broadly the principles which govern the federal equitable remedial power in Milliken v. Bradley, 418 U.S. 717, 94 S.Ct. 3112, 41 L.Ed.2d 1069 (1974) (Milliken I). Judge Arnold, while recognizing these principles and cogently pointing out the manner in which the court misapplied them, improperly relies upon his own factual conclusions based on an overly generous interpretation of the record to justify the remedy ordered today.
Milliken I established that an interdistrict remedy is appropriate only upon a showing of “a constitutional violation within one district that produces a significant segregative effect in another district;” specifically, that “racially discriminatory acts * * * have been a substantial cause of interdistrict segregation.” Id. at 744-45, 94 S.Ct. at 3127. Further, an interdistrict remedy is appropriate only “to eliminate the interdistrict segregation directly caused by the constitutional violation.” Id. at 745, 94 S.Ct. at 3127 (emphasis added). Two courts of appeals have read this language to require clear proof of cause and effect of a constitutional violation and a careful delineation of the extent of the effect before an interdistrict remedy may be involved. In Lee v. Lee County Board of Education, 639 F.2d 1243, 1256 (5th Cir.1981), the court stated:
We believe the Court’s deliberate choice of phrases such as “substantial” or “direct cause” and “significant segregative effect” also expresses an insistence that in cases where an interdistrict remedy is requested, there must be clear proof of cause and effect and a careful delineation of the extent of the effect. In the absence of such a showing, school district lines are to be carefully observed and desegregation remedies confined to orders affecting the school district in which the condition of segregation is manifest.
Accord Goldsboro City Board of Education v. Wayne County Board of Education, 745 F.2d 324, 332 (4th Cir.1984).
The Fifth Circuit also emphasized in Lee that there must be “a substantial, direct and current segregative effect,” 639 F.2d at 1260 (emphasis in original), before an interdistrict remedy may be ordered. This argument is persuasive, for a remedy cannot be tailored to correct a condition, Milliken I, 418 U.S. at 738, 94 S.Ct. at 3124, unless it currently offends the Constitution. As the court stated in Milliken I: “A federal remedial power may be exercised ‘only on the basis of a constitutional violation’ and, ‘[a]s with any equity case, the *445nature of the violation determines the scope of the remedy.’ ” Id. at 738, 94 S.Ct. at 3124 (quoting Swann v. Charlotte-Mecklenburg Board of Education, 402 U.S. 1, 16, 91 S.Ct. 1267, 1276, 28 L.Ed.2d 554 (1971)). The Court reemphasized this important limitation on the federal remedial power in General Building Contractors v. Pennsylvania, 458 U.S. 375, 399, 102 S.Ct. 3141, 3154, 73 L.Ed.2d 835 (1982), in which it cautioned that a remedial decree should “extend no farther than required by the nature and the extent of that violation.”
It is also well to observe before we turn to specific issues that the court today adopts its own remedy, which has not been addressed by the parties in their arguments or briefs, and which differs substantially from that ordered by the district court. The findings of fact necessary to support this court’s remedy simply do not exist. Specifically, there is no finding by the district court of a current segregative effect to support the conclusion that the boundaries of the City of Little Rock and LRSD be made coterminous.
The court, to support that portion of its discussion mandating that the boundary of the City of Little Rock and the LRSD be coterminous, simply catalogs a portion of the district court’s findings in a footnote. See ante at 435 n. 20. It must be observed, however, that the court, at least with respect to the district court’s findings on student assignments, special education, transportation, and employment of faculty and administrators, simply has enumerated intradistrict violations that may require intradistrict remedies within the PCSSD. Neither this court’s nor the district court’s opinion indicates in any way that these mira district violations manifested an inter district effect.
Similarly, neither the opinion of this court nor the record from the district court reveals the conduct of the PCSSD which has resulted in the lack of low income housing in the PCSSD. The only exception with respect to public housing is the evidence concerning the Granite Mountain deannexation in 1953 and the role that the PCSSD may have played at that time.
The court also relies on violations relating to school sitings to justify an interdistrict remedy. These violations which are discussed at some detail in the court’s opinion involve decisions by the PCSSD to build some twelve schools away from the centers of black population. This, it is argued, served to attract whites to the outlying areas, resulting in a number of schools with over 90% white enrollment. This is a weak foundation for the remedy the court today orders. Adjusting the boundaries of the PCSSD and the LRSD so that the latter are coterminous with Little Rock city limits will not affect the twelve schools in question, which are located far from the city limits. There simply is no indication in the record that the suspect school sitings had any impact on the schools within the Little Rock city limits that, as a result of the court’s decision today, will now be turned over to the LRSD.
Judge Arnold’s opinion treats in detail what interdistrict effects might result from the history of annexations and deannexations. The district court’s order stressed the freezing of the boundaries. The court today emphasizes testimony that “the boundaries hardened],” ante at 419, and were “maintained to keep the LRSD predominantly black and the PCSSD predominantly white,” and concludes that these manipulations had a substantial interdistrict segregative effect. Admittedly there was expert testimony to support this conclusion. The court relies heaviliy on Dr. Dentler’s testimony that beginning in 1968 the board of PCSSD refused to modify its boundaries. Ante at 420. To the contrary, the United States argued that PCSSD has not refused a single deannexation petition since it allegedly froze its boundaries. In its brief, LRSD attempts to refute this argument and support the theory that the PCSSD intentionally froze its boundaries:
The record reveals, however, on 4/13/65 the PCSSD notified the adjacent Bryant and Cabot school districts that it would *446not accept any more black students. PX 10. The minutes further reveal the county refused to meet with LRSD officials to discuss consolidation on 5/14/68, and refused NLRSD’s attempt to annex the Spring Hill area on 4/8/69. Further, informal efforts of the LRSD to discuss •cooperative interdistrict agreements failed.
Brief for Appellee at 56.
After argument, this court specifically inquired as to the record support for these statements. LRSD answered that the first sentence had no record support. As to the second sentence, it answered that the PCSSD board had, on May 14, 1968, voted in favor of consolidation with LRSD.1 On this critical point, therefore, LRSD simply was forced to admit that its argument had collapsed. This collapse not only removes the factual underpinnings from Dr. Dentler’s opinion, but makes it directly contrary to the evidence. Judge Arnold is correct in his conclusion that “it is true * * * that no deannexations have taken place since 1968, but to blame this on PCSSD’s desire not to increase its black student percentage is not plausible.” Ante at 440. On the record before us, I can only conclude that the district court’s finding that the PCSSD intentionally “froze” or “hardened” its boundaries, which the court today accepts, is contrary to the evidence, and thus is clearly erroneous.
The only other support for the court’s order today is “the pre- and post-Brown interdistrict transfers and the other historical violations with continuing effect.” Ante at 435 n. 20. Absent a more specific demonstration of the continuing effects of historical violations, see ante at 436 (Arnold, J., concurring and dissenting), I believe that this evidentiary foundation, as well as the entire stated basis for the court’s remedy, fails to meet the standards of Milliken I and the clarifying interpretation of those standards in Lee and Goldsboro.
To Judge Arnold’s credit, he squarely faces the weaknesses of the court’s opinion. His further observations on the support for making the boundaries of LRSD and the City of Little Rock coterminous deserve further comment.
From the decisions of the PCSSD locating schools in areas of white suburban expansion farthest from black population centers, Judge Arnold argues “I believe these factors are having a substantial current interdistrict effect.” Ante at 441. He argues that the movement of families from outside the entire Pulaski County area to areas within the City of Little Rock also within the PCSSD has been substantially influenced by the unconstitutional siting decisions, causing a phenomenon labeled “white overflight.” It must be observed that Judge Arnold does not point to any “substantial evidence” proving this “significant segregative effect,” Milliken I, 418 U.S. at 744-45, 94 S.Ct. at 3127, or as expressed in Lee, “clear proof of cause and effect and a careful delineation of the extent of the effect.” 639 F.2d at 1256. Rather, he relies upon belief. Such beliefs do not, within the limitations of our judicial power, serve as an appropriate consideration in reviewing the order before us. Judge Arnold makes no effort to determine whether there are findings of the district court which support these conclusions. There simply are none. The imprecise nature of his conclusions and his departure from the principles of Milliken, are best evidenced by his conclusion that the ten percent change in racial makeup which results from realigning the LRSD’s boundaries “seems a fair approximation of what the racial percentages would have been absent the influence of PCSSD’s violations.” Ante at 441. Thus, his vigorous effort to support the court’s conclusion ultimately rests on speculation, belief, and fair approximation, and not upon the principles of Milliken I which must guide us.
*447Judge Arnold’s final observation conclusively demonstrates the tenuous nature of his conclusion. He points to “pro-white emanations” that PCSSD has given off over a period of years as a factor which has attracted white parents into the PCSSD areas within the City of Little Rock. There is no indication as to the record source of these emanations. I have substantial question whether such subjectively perceived emanations are sufficiently palpable to make the showing required by Millilcen I or to constitute the clear proof of cause and effect which Lee and Goldsboro correctly hold to be necessary.
Judge Arnold’s particular approach to these issues must rely upon his own findings. This is contrary to the Supreme Court’s teaching in Anderson v. City of Bessemer City, — U.S. -, 105 S.Ct. 1504, 1510-11, 84 L.Ed.2d 518 (1985), that the trial judge must play the principal role in the determination of facts.
A final observation is in order. The court stretches mightily to find a basis for making the boundaries of the City of Little Rock and the LRSD coterminous. This result will make the LRSD 60 percent black and 40 percent white. However, one most significant factor has been omitted from this equation. In 1980 there were 3,632 white students enrolled in private schools located in the LRSD, 2,794 white students enrolled in private schools in PCSSD, and 1,086 white students enrolled in private schools in the NLRSD. In the LRSD alone, if the white students in private schools attended public schools, the district would be approximately 52 percent rather than 69 percent black. Another significant factor which is not taken into account is that the number of white students in private schools in LRSD increased by more than 1,000 between 1970 and 1980 and the number of white students in private schools in PCSSD increased by more than 1,700 during the same time period. It is evident from the sharp increase in enrollment in private schools in these ten years, and the impact of these numbers on the total student population, that private choice is having a far greater segregative effect than those factors the court points to in its opinion today.2 This is a factor, however, that at present, and in all likelihood in the future, will pose a significant impediment to any effort to achieve desegregation within any of the three school districts.
One of the great failings of the court’s opinion, as well as that of the district court order, is the failure to address in other than a most speculative way demographic factors, such as population movement and birth rate, common to major metropolitan centers, that have significant impact on the school problems in this and other communities. Pasadena City Board of Education v. Spangler, 427 U.S. 424, 436, 96 S.Ct. 2697, 2704, 49 L.Ed.2d 599 (1976); see also Bradley v. School Board of City of Richmond, 462 F.2d 1058, 1066 (4th Cir.1972), aff'd mem. by an equally divided court, 412 U.S. 92, 93 S.Ct. 1952, 36 L.Ed.2d 771 (1973).
I would remand the case to the district court only for further consideration of appropriate relief for intradistrict constitutional violations and for further consideration of the current segregative effect resulting from the Granite Mountain deannexation and for consideration of an appropri*448ate remedy tailored to correct any such violation.
. The informal efforts referred to in the last sentence occurred in a later period, shortly before the filing of this action.
. The private school problem seems to be an untouchable issue that none of the parties has evidenced any interest in addressing. It may well be that a substantial portion of the enrollment is in religious and parochial schools. From figures furnished to the court by the parties based upon publications of the state of Arkansas, however, there are nearly 3,000 students in the Little Rock area who attend private schools seemingly having no religious affiliation. There simply is no record before us to determine whether some of this enrollment is pretext for avoiding the impact of desegregation or springs from other motives. We do observe, however, that racial discrimination in private schools, including those with a religious affiliation, is deeply contrary to public policy, Bob Jones University v. United States, 461 U.S. 574, 595 & n. 32, 103 S.Ct. 2017, 76 L.Ed.2d 157 (1983), and that schools which are shown to employ discriminatory practices will be barred from enjoying the governmental privileges accorded their nondiscriminatory counterparts. Id. at 595-96, 103 S.Ct. at 2030-31.