RONEY, Chief Judge:

In this Savannah public school desegregation case, the central issue on this appeal is whether the district court was within its discretion in approving as a remedy for past discrimination, a desegregation plan which relies heavily on magnet programs in predominantly black schools and voluntary "majority to minority" transfers. Inasmuch as prior plans which provided for mandatory busing, mandatory assignment, and pairing of black schools with white schools has failed to result in the level of desegregation the parties desire, and there is sufficient evidence in the record to support a prediction that this new plan will work, we affirm.

This is the latest in a series of decisions rendered over the last 25 to 30 years in an effort to desegregate the public schools of the City of Savannah and the County of Chatham County, Georgia, as a remedy for prior discrimination. The history of the case is fully outlined in the district court opinion. Stell v. Savannah-Chatham County Bd. of Public Education, 724 F. Supp. 1384 (S.D.Ga. 1988).

Since the 1971-72 school year, schoolchildren in Savannah-Chatham County have been assigned to schools pursuant to a school desegregation plan implemented in compliance with the specific order of the district court. That plan called for the pairing and clustering of all-black and all-white schools, for mandatory assignment, and for extensive busing to achieve a unitary school system. The plan achieved immediate and successful albeit ephemeral results. After institution of the plan, the school system lost approximately 10,000 children, predominantly white and middle class, to private and other area schools. This "white flight" continued until what had been a majority white school district became predominantly black.

Recognizing that segregated schools had again taken root, in June 1985, at the request of the plaintiffs, the district court directed the school board to submit a redrawn desegregation plan. An initial plan, approved by the parties, was defeated by the voters of Chatham County who rejected a proposed bond issue necessary to finance the plan. The parties could not agree on another plan, so they submitted alternative proposals to the court. The plans target elementary schools and one high school. The middle schools and the other high school are currently desegregated. The Justice Department did not submit a proposal.

The school board's plan abandons mandatory busing, the use of pairing of one black school to one white school, and mandatory reassignment. The heart of the board's plan centers on revised attendance zones, voluntary "magnet programs" and a "majority-to-minority transfer" option.

A magnet is an educational program which, in addition to a basic curriculum, offers an additional highly specialized curriculum centered on a theme. Such a program is designed to meet the needs of children as well as providing an added attraction that will draw students from around the school district. These programs would be instituted in predominantly black schools along with traditional educational courses, with each magnet program to enroll fifty percent white and fifty percent black students. Because of the residential pattern in the district, the non-magnet programs would be made up of predominantly black students.

The district court, in a thorough opinion entered after an evidentiary hearing and allowing the parties additional time to submit supplemental materials and memoranda, found the school board's plan, "while perhaps not the most desirable plan which could have been selected if financial limitations were not a consideration, . . . should be effective in achieving the desegregation of the Savannah-Chatham County public schools," and adopted the school board's proposal. It is from this order that plaintiffs, representing the black, minor schoolchildren in the Savannah-Chatham County School District, and the Justice Department appeal.

We review the district court's order for abuse of discretion, and we are bound by the district court's findings unless clearly erroneous. Milliken v. Bradley, 433 U.S. 267, 288, 97 S.Ct. 2749, 2761, 53 L.Ed.2d 745 (1977); Lee v. Anniston City School Sys., 737 F.2d 952 (11th Cir. 1984). The school board authorities are charged with making "every effort to achieve the greatest possible degree of actual desegregation," but the board must do so "taking into account the practicalities of the situation." Davis v. Board of School Comm'rs of Mobile County, 402 U.S. 33, 37, 91 S.Ct. 1289, 1292, 28 L.Ed.2d 577 (1971).

The district court relied heavily on the school board's witnesses in making its determination. A number of witnesses, including the school system superintendent, the president of the Savannah-Chatham County School Board, and the principals of the schools hosting the pilot magnet programs, testified as to the planning, design, goals, and feasibility of the school board's plan. Two other witnesses, whom the court accepted as credible experts in the field of school desegregation, testified as to the validity of the school board's survey techniques, the reliability of the board's enrollment and transfer projections, and the possibilities of success of the school board's plan.

Two principals where the magnet programs are in place testified that magnet and non-magnet students participated in numerous assemblies, lunch hours, remedial reading classes, band, chorus, playground and physical education, and that there was interaction among the magnet and non-magnet students.

The black students in non-magnet programs have several choices. First, resident children are given priority for admission to the magnet programs in their neighborhood school. The record indicates that in the two magnet programs instituted, no neighborhood black child seeking participation on time was denied admission. Second, resident children who do not wish to participate in the magnet program in their neighborhood may apply for admission to another magnet program, or they may voluntarily transfer to another school under the majority-to-minority program. Third, resident children who attend their neighborhood magnet school but do not participate in the magnet program nevertheless attend an integrated school and experience meaningful daily interaction with the children in the magnet classes.

Though plaintiffs' expert, Dr. Michael Stolee, testified that a plan proposed by plaintiffs would better serve to desegregate the schools, he also admitted that he was not an expert in statistical analysis or surveys, he has never done any research or scholarly writing in the areas of white flight, mandatory versus voluntary plans, majority-to-minority transfers or the characteristics of effective school desegregation plans. He also testified that he has consulted with one parent from the school system, but has consulted no school board members, no teachers, and no administrators, nor has he visited the two elementary schools piloting the magnet programs to ascertain their success.

The plaintiffs submitted to the court an alternative plan based largely on the school board's plan and enrollment projections, and relying on the same attendance zones. The fundamental difference in the plans is that instead of magnets within otherwise non-magnet schools, plaintiffs would require dedicated magnet schools, where the entire school would be devoted to magnet programs. Appellants would also implement a back-up plan involving mandatory pairing of schools in the event the magnet schools program failed. Like the board's plan, plaintiffs' proposal allows for the voluntary majority-to-minority transfer option.

The plaintiffs' contention that their plan is better is a fact question, and it was clearly within the district court's discretion to choose between conflicting testimony. Given the district court's opportunity to observe the witnesses' demeanor and level of experience, we cannot find clear error in the district court's holding that the school board's plan was more appropriate.

In challenging the school board's use of a limited magnet school program, plaintiffs contend that the school board's plan is fatally defective because it subjects black students attending magnet schools who are not enrolled in magnet classes to "in-school segregation." Plaintiffs rely principally on two cases, Davis v. East Baton Rouge Parish School Bd., 514 F. Supp. 869, 872 (M.D.La. 1981), aff'd 721 F.2d 1425 (5th Cir. 1983), and United States v. Pittman, 808 F.2d 385 (5th Cir. 1987), to support this contention.

In both cases, the courts rejected voluntary magnet program proposals similar to the one adopted in the instant case. Both Davis and Pittman are factually distinguishable in two important respects. In both cases, the courts determined that the respective school boards had made no serious effort in the past to desegregate their schools. No such finding was made in the instant case. To the contrary, the school board appears here to be fully committed to the success of this plan and to making adjustments from time to time to make it so. Also, the rejected programs in Davis and Pittman, both left in place some one-race schools while the school board's program in this case will integrate all the schools.

While the rejection of the school board's plan was certainly an option within the district court's discretion, the fact that similar programs were rejected in Davis and Pittman is not a valid reason for the district court to exercise that option, particularly in light of the factual distinctions.

Plaintiffs contend that under the approved plan, the district has not met its goal of a racial composition within twenty percentage points above or below the systemwide student ratio. Such statistics are not independently controlling. The Supreme Court has made clear that while racial ratios provide a useful tool for analyzing various plans, they are not to be used as rigid barriers. See Swann v. Charlotte-Mecklenburg Bd. of Educ., 402 U.S. 1, 22-25, 91 S.Ct. 1267, 1279-1281, 28 L.Ed.2d 554 (1971). While the magnet schools may not have yet met the goal established by the school board, they are not "segregated" schools in which minority children are isolated. There is ample evidence in the record to support the district court's opinion that the school board plan realistically promises to effectively desegregate the school district.

Plaintiffs argue that the approved plan was proposed because of a perceived fear of "white flight." Because the board and its experts see white flight as an inevitable consequence of any mandatory desegregation plan, according to plaintiffs, the board refused to consider any such plan. Plaintiffs are correct that fear of "white flight" cannot justify delaying desegregation. United States v. Scotland Neck Bd. of Educ., 407 U.S. 484, 491, 92 S.Ct. 2214, 2218, 33 L.Ed.2d 75 (1972); Tasby v. Wright, 713 F.2d 90, 99 (5th Cir. 1983). Neither may it be appropriate to design a plan controlled by predicted white flight. The situation is different, however, where white flight has occurred, the school system has become predominantly black, and the problem is how to attract white students back into the program. Plans like the school board's that are designed to attract white students into predominantly black schools are suitable as long as they do not frustrate desegregation efforts. Tasby, 713 F.2d at 99; Davis v. East Baton Rouge Parish School Bd., 721 F.2d 1425, 1438 (5th Cir. 1983).

In fact, we were told at oral argument that a substantial number of both white students and black students have re-entered the school system since the plan was implemented. The goal is education. A school system should be designed to attract motivated students of all races if it is to provide quality education.

The district court found here that the plan proposed by the school board relying on voluntary desegregation techniques is particularly appropriate in light of the previous failure of mandatory pairing and busing. The district court committed no clear error in making this determination.

The court stated that the primary purpose of the magnet program as designed for this desegregation plan is to draw white students to the almost exclusively black inner city schools in substantial enough numbers to desegregate those schools. The plaintiffs' contention that magnet schools, rather than magnet programs, will produce more effective integration could properly be rejected. Both of the school board's school desegregation experts testified that dedicated magnet schools would require mandatory reassignment of resident black students who do not enroll in the magnet programs, which would place an unfair burden on those students.

Neither the district court nor the school board was unmindful of plaintiffs' concerns about students in the non-magnet programs having the opportunity to interact with magnet program students. The court acknowledged this concern and stated that because it retains jurisdiction over this case, alterations can be made in the future if it proves necessary. We note that while we recognize the potential problems in the proposed plan pointed out by plaintiffs, we cannot base a finding of abuse of discretion upon these flaws.

The district court correctly stated that the measure of a desegregation plan is its ultimate effectiveness. Davis v. Board of School Comm'rs of Mobile County, 402 U.S. at 37, 91 S.Ct. at 1292. The type of plan employed is of little consequence so long as it effectively achieves the constitutionally required result that public schools be conducted on a unitary basis. Valley v. Rapides Parish School Bd., 434 F.2d 144, 146 (5th Cir. 1970). There is ample evidence to support the district court's conclusion that this plan "is valid and would lead to effective and stable desegregation if given an opportunity to succeed."

The United States does not endorse the argument that plaintiffs' proposed plan should be accepted rather than the school board's. Rather it presents a limited challenge on this appeal. First, the United States seeks a plan requiring the school board to integrate magnet and resident students attending the same school in non-magnet courses (academic courses as well as physical education, assembly and shared lunch hours). The schools presently do this pursuant to stipulation between the parties. Since the following of this stipulation will make it more likely that the result sought by all parties will be achieved, we remand this case for the district court to implement the stipulation into a court order. Second, the Government seeks magnet programs at predominantly white schools as well, to draw black students to these schools. Although we make no decision as to its merit at this time, this is the type of recommendation that the court should review on remand and consider whether its implementation as part of the court-ordered plan would enhance the chances of a successful result. It is understood that adjustments may have to be made from time to time that will be worthwhile in making improvements in the school board's plan. That fact, however, does not constitute proof that the adopted plan is constitutionally defective.

AFFIRMED and REMANDED.