The McKenzie School, a public kindergarten through twelfth grade school, is located in the southern portion of Butler County, Alabama, approximately one mile from the county line. The sole issue in this appeal is whether the district court erred in failing to enjoin the continued transfer of certain students, who are virtually all white, from the neighboring Alabama counties of Cone-cuh and Covington to McKenzie, a majority white school.
The chain of events leading to this lawsuit began in 1986 when Lowndes County, which straddles Butler County’s northern border, sought discharge from its 1973 desegregation decree. Because the Lowndes County school system had become increasingly black in the years following the entry of its desegregation order, and the exodus of white students from the district could not be completely accounted for by movement into the private schools, the government suspected that some of these students were “zone-jumping” to predominantly white public schools in other counties. The government’s resulting investigation uncovered a comprehensive network of in-terdistrict transfers among Lowndes, Cren-shaw, and Butler Counties, each of which were operating under a court-ordered desegregation plan that contained a standard Singleton-type transfer provision.1 The government moved to enforce the transfer clause of Lowndes County’s desegregation order, and the court later granted the government’s motion to consolidate the cases in the other counties to determine whether their transfer provisions similarly had been violated. The boards of education in Montgomery, Pike, Coffee, Cov-ington, Conecuh, and Wilcox Counties and Elba City were later joined as defendants for the limited purpose of considering inter-district transfers. During the course of the litigation, all counties except Conecuh, Wilcox, Covington and Butler entered into consent decrees.2
*1303Prior to the 1987-1988 school year, the Butler County Board of Education freely granted transfers into its school system under an informal freedom of choice plan. In August, 1987, after the government filed this suit, the school board adopted a formal interdistrict transfer policy aimed at reducing the total number of out-of-county pupils accepted by Butler County. The new policy effectively reduces the number of incoming transfer students from Lowndes and Wilcox counties, but does not prohibit the continued transfer of students from the counties of Conecuh and Coving-ton to the McKenzie School.3
The case then went to trial with respect to Conecuh, Wilcox, Covington, and Butler Counties, limited to transfers either to or from the latter two counties. Following the district court’s decision in favor of all defendants, the government appealed that part of the order allowing Conecuh and Covington County students to continue attending McKenzie.
The Butler County school district is comprised of three attendance zones. Although the district-wide ratio of students in the county is 55% black and 45% white, the southern attendance zone has an overall population of between 30-37% black and 63-70% white. All students residing in the southern zone are assigned to McKenzie, the only predominantly white school in a system that is majority black. During the 1987-88 academic year, the McKenzie School accepted 81 children from Conecuh County, of whom 78 were white, and 13 white pupils from Covington County. As the accompanying chart illustrates, without these white out-of-county transfers, McKenzie would have a racial mix of 54.6% white students and 45.4% black students. With the transfers, as presently allowed, the school is 63.9% white and 36.1% black. Thus, the cumulative effect of nonresident enrollment on the McKenzie School is 9.3%.4
*1304White % White Black % Black
Present Enrollment 285 63.9 161 36.1
(Students from Conecuh) (78) (3)
(Students from Covington) (13) (0)
Enrollment Without White Out-of-County Students 194 54.6 161 45.4
The transfer provision in Butler County’s terminal desegregation order, which incorporates the language of the former Fifth Circuit Court of Appeals in Singleton v. Jackson Municipal Separate School District, provides:
If the school district grants transfers to students living in the district for their attendance at public schools outside the district, or if it permits transfers into the district of students who live outside the district, it shall do so on a non-discriminatory basis, except that it shall not consent to transfers where the cumulative effect will reduce desegregation in either district or reenforce the dual school system.5
Nothing in the record suggests that Butler County ever accepted out-of-county transfers on a racially discriminatory basis, either under its previous freedom of choice plan or under its new transfer policy. Even if transfers have been nondiscriminatorily granted or received, however, the Singleton provision also obligates the school district to monitor the effect of such transfers, both on its own desegregation efforts and on the desegregation process of the school district from which it receives, or to which it sends, its students. The district court found that Butler County had utterly failed in this respect prior to 1987, but that the subsequent implementation of its new transfer policy brought the district within full compliance. Because the new policy does not affect transfers from Conecuh and Covington Counties, the remaining question on appeal is whether the cumulative impact of these interdistrict pupil transfers to the McKenzie School in Butler County serves to reduce desegregation in any of the school districts or to promote a dual school system. Inasmuch as the government concedes that the transfers in issue did not impact negatively upon the Conecuh and Covington schools, we do not take issue with the district court’s finding that the cumulative effect on these sending districts was de minimis and thus not in violation of Singleton.6 Accordingly, our review is restricted to the impact of transfers upon McKenzie, the receiving school.
Two former Fifth Circuit Court of Appeals decisions serve as our guideposts in determining whether Butler County has violated the terms of its transfer provision by sanctioning the continued attendance of these Conecuh and Covington County students at the McKenzie School. In Lee v. Eufala City Board of Education, 573 F.2d 229 (5th Cir.1978), the government alleged that the city school system of Eufala, Alabama breached the Singleton transfer provision of its desegregation order by permitting white children from predominantly black, neighboring Alabama and Georgia counties to attend Eufala’s majority white schools. The court explicitly rejected the government’s contention that Singleton permitted the transfer of white students only on a strict one-for-one exchange basis, and held instead that “[i]n measuring the cumulative effect of a student transfer program on desegregation, the Court must do so from a qualitative viewpoint, without blind deference to an objective mathematical formula.” Eufala, 573 F.2d at 232. *1305The court noted that although a percentage or quantitative analysis could be considered as a factor, the focal point in determining the qualitative effect on desegregation was whether the transfers “increase the racial identifiability of the schools.” Id. at 233 (quoting N.A.A.C.P. v. Lansing Board of Education, 559 F.2d 1042, 1051 (6th Cir.), cert. denied, 434 U.S. 997, 98 S.Ct. 635, 54 L.Ed.2d 491 (1977)). Several years later, the court held in Lee v. Lee County Board of Education, 639 F.2d 1243, 1261 (5th Cir.1981), that an index of racial identifia-bility was whether the “increment of change in the racial composition of a school [is likely] to alter significantly general perceptions of a school’s racial identity or the behavior of persons who rely on such factors in determining whether or not to send their children to a particular school.”
Eufala also articulates the applicable model of analysis to be used in assessing the cumulative effect of interdistrict transfers:
The ‘cumulative effect’ of the transfer program must be measured on a school-by-school basis. This is the only operational level on which actual segregative effect can be measured, and upon which it can be determined whether the transfer policy reduces desegregation or reinforces the existence of a constitutionally impermissible dual school system.
Eufala, 573 F.2d at 233 (footnote omitted). Thus, Eufala mandates a comparison of the racial composition of the McKenzie School as it would exist without the transfers with the present enrollment including the transfers, and then a determination, from a qualitative standpoint, of whether the difference undermines desegregation efforts at the school. The government would also have us consider the fact that McKenzie’s 63.9% white enrollment deviates almost 20 percentage points from the district-wide ratio of 45% white to 55% black. However, “[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 v. Charlotte-Mecklenburg Board of Education, 402 U.S. 1, 24, 91 S.Ct. 1267, 1280, 28 L.Ed.2d 554, 571 (1971); accord Milliken v. Bradley, 418 U.S. 717, 740-41, 94 S.Ct. 3112, 3125, 41 L.Ed.2d 1069, 1089 (1974). Nor does Eufala envision a district- or county-wide analysis.7 Rather, the court held that “it may be appropriate to enjoin the transfer of students ... if the transfer policy has the effect of reducing desegregation or reinforces the existence of a dual school system in each attendance zone within the county.” Id. at 233 n. 10. As we previously observed, Butler County’s southern attendance zone, from which McKenzie draws its students, is approximately 63-70% white. The school would be predominantly white even without the contested transfers. Thus, to require McKenzie’s enrollment ratios to approximate the county’s overall racial makeup would be inappropriate where, as here, the attendance zones are drawn along nonracial lines and the disproportionately higher percentage of whites in the southern zone is due to demographic factors.8 Racial imbalance in the public schools amounts to a constitutional violation only if it results from some form of state action and not from factors, such as residential housing patterns, which are beyond the control of state officials.9 See Swann, supra, 402 U.S. at 28, 91 S.Ct. at 1282, 28 L.Ed.2d at 573 (1971).
*1306Eufala and Lee County provide this court with the appropriate legal framework, but divulge no easy answer. Whether a particular school has become racially identifiable, to the extent that such an assessment relies upon largely subjective perceptions, can present a formidable task. The fact that the McKenzie School is 9.3% more white than it would be without the transfers from Conecuh and Covington Counties is not per se indicative of a Singleton violation.10 What compels our decision in this case is whether that increment of change is likely to aggravate or alter popular perceptions of McKenzie’s racial identity and whether it affects the decision-making process of white students considering where to attend school.
In the 1987-88 school year, McKenzie enrolled 78 white transfer students from the adjoining Conecuh County. By informal agreement between the two counties, Butler County buses come into Conecuh to transport these out-of-district students to McKenzie.11 Most of the children reside in a small, confined area within the northeastern corner of the county, which is almost exclusively white. Ordinarily, they would attend Conecuh County’s predominantly black Evergreen area schools if they were to go to school in their home zone.12 The defendants contend, however, that children from this part of the county have always attended McKenzie and consider it their “home school.” The heart of their argument is that these particular residents traditionally have considered themselves, and are regarded by McKenzie residents, to be part of the McKenzie community. The McKenzie School is less than five miles away from the northeastern section of Co-necuh County in which the students reside. To require them to attend school in their own zone, the defendants maintain, would subject the students to a longer bus ride, depriving the students of valuable time and the Conecuh school district of much needed funds.
The Singleton transfer provision was designed to prevent white students from fleeing predominantly black, often urban schools to majority white public schools in neighboring counties. This pattern of “white flight” often results in de facto resegregation. See, e.g., N.A.A.C.P. v. Lansing Board of Education, supra, 559 F.2d at 1050-51. Here, however, students from this part of the county, a sparsely populated rural area, have crossed the county line to attend McKenzie since the 1980’s, well before the advent of desegregation. They cannot literally be said to be “fleeing” the Evergreen schools since they have never gone there.13 We note also that McKenzie’s total enrollment, as well as the *1307racial composition of its student body, has remained relatively constant since Butler County’s implementation of its desegregation order in 1970. Thus, the McKenzie School does not appear to have become a “white haven,” at least in the usual sense of the term.
Nonetheless, neither tradition nor convenience justify constitutional infirmity. While these arguments may signify some evidence of lack of present discriminatory intent, they do not legitimize what was in reality purposeful racial discrimination in years past. McKenzie was an all-white school during the era of state-imposed segregation and has remained predominantly white following desegregation of the school. Moreover, Singleton requires us to assess the segregative impact of the transfers. The fact that the transfers from this area traditionally associated with the McKenzie community may not now be racially motivated loses its relevance if the result of the transfers is to impede desegregation or to perpetuate a dual school system. Nor is the issue settled by evidence that McKenzie’s white enrollment has not appreciably increased throughout the years. A Singleton violation has still occurred if the 9.3% increment of change has resulted in a perception of the school as being more “white.”
Moreover, we are disturbed by evidence which reveals that at least eleven white children from other areas of Conecuh County’s Evergreen attendance zone attend the McKenzie School. The record reflects that in some instances these children are privately transported by car to designated Butler County pick-up points, despite the proximity of Conecuh County bus routes, and in other cases are driven in carloads directly to McKenzie. These children do not live within the confines of the McKenzie pocket, so their attendance cannot be rationalized by tradition or convenience. Although the district court apparently did not consider these transfers to be significant, we find them to be highly suspect. This evidence strongly suggests that McKenzie’s racial composition did influence the behavior of white residents of other areas of Conecuh County who found it advantageous to send their children to a predominantly white school, an impermissible result under Eufala and Lee County.
The Covington County transfers also present an area of concern. A Butler County school bus transports thirteen white students eight miles from the Brooks community in the northwest corner of Cov-ington County to the McKenzie School. These transfers are not explained by convenience. The record reflects that these students would otherwise attend the Red Level School in Covington, which is equidistant, if not closer, to the Brooks area. Furthermore, although Jimmy Lawrence, Butler County’s Superintendent of Education, testified that students from the Brooks area have traditionally attended McKenzie, such historical attendance at the school simply might be a remnant from the days of de jure segregation when McKenzie was exclusively white and would not justify the present transfers. However, although the record is unclear on this point, it appears that the Red Level School, which ordinarily would accommodate these students, is majority white.14 If true, this would tend to negate the inference that the Covington County transfers were motivated by general perceptions of McKenzie as a “white” school.
Eufala reminds us that “ ‘school desegregation can first be measured quantitatively, using percentages as a rough rule of thumb, but ultimately must be measured qualitatively, judging whether schools are still identifiable as white or Negro.’ ” 573 F.2d at 233 (quoting Davis v. Board of School Commissioners of Mobile County, 393 F.2d 690, 693 (5th Cir.1968)). In its opinion, however, the district court equated racial identifiability with quantitative percentages. The court found that the McKenzie School was integrated 36.1% black and 63.7% white with an 8.4% quantitative cumulative effect of transfers from *1308Conecuh and Covington Counties, and also that the quantitative cumulated effect on these transferring schools was between 3.4% and 6.6%. Based on these findings, the court determined that the transfers had not significantly altered general perceptions of any school’s racial identity or the behavior of persons who rely on such factors in determining whether or not to send their children to such schools. Eufala, however, prohibits strict reliance on figures in deciding whether a school has become racially identifiable. Racial identifiability involves a qualitative analysis of whether a school has assumed a racial identity. One cannot decide the issue based on percentages and ratios alone.
We conclude from our independent review of the record that the McKenzie School has become significantly more racially identifiable as a result of the transfers from Conecuh and Covington. We acknowledge that McKenzie, due to the residential makeup of the southern zone, would be majority white even without these students. We also note that the 9.3% cumulative difference in this case, from 54.6% white to 63.9% white, is not of itself indicative of a Singleton violation, as it might be if the school were to change, for example, from 91% white to 100%. Nonetheless, we find that this increment of change is likely to alter significantly general perceptions of McKenzie’s racial identity or the behavior of persons who rely on such factors in determining whether or not to send their children to that particular school. Lee v. Lee County, supra, 639 F.2d at 1261. The key to our determination is the government’s assertion, well supported in the record but apparently ignored by the district court, that white Conecuh County students from areas outside the traditional McKenzie pocket are also permitted to attend the McKenzie School. Such transfers intimate that McKenzie is currently perceived as a “white school” and that this status attracts students who otherwise would attend the predominantly black, Evergreen area schools. Similarly, although evidence of discriminatory preference for the McKenzie School is not as clear with the thirteen Covington County transfers, they also should be disallowed. Requiring these students to attend school in their home zone would be no more inconvenient than the current practice of transporting them roughly the same distance across the county line to McKenzie. In any event, their continued presence at the school only enhances the perception that McKenzie has become a refuge for white students.
While we realize that students from the northeastern corner of Conecuh County might naturally desire to go to their “neighborhood” school, independent of any racial animus, we cannot entertain such a preference where the result only contributes to McKenzie’s white image. Transfers from this portion of the county cannot be justified in the interests of tradition and convenience without residual harmful effects to Butler County’s desegregation efforts. The influx of an additional 78 white students from this area into McKenzie does make the school significantly whiter in a school that is already predominantly white and was exclusively white before desegregation. As such, they obstruct the goals of desegregation. Although McKenzie’s present and past enrollment figures convince us that the school has not yet become a white haven, there is nothing to prevent such an eventual occurrence should we allow these transfers to continue. Our obligation is to ensure that schools do not devolve into “white schools” or “black schools,” but remain simply schools. See Green v. School Board of New Kent County, 391 U.S. 430, 442, 88 S.Ct. 1689, 1696, 20 L.Ed.2d 716, 726 (1968).
Accordingly, we reverse and remand. Being mindful of need to implement whatever changes may be necessary before the onset of the 1989-1990 academic year, we direct the district court to enter judgment consistent with this opinion as soon as practicable to achieve this end.
REVERSED and REMANDED.
. See Singleton v. Jackson Municipal Separate School District, 419 F.2d 1211, 1218-19 (5th Cir.1969), cert. denied, 396 U.S. 290, 90 S.Ct. 608, 24 L.Ed.2d 530 (1970).
. Wilcox and Conecuh Counties were no longer under a desegregation order, as these districts had earlier been declared unitary. Butler County, however, has never been subject to a judicial determination that its school system has achieved unitary status. Although the district court referred to Butler County as "unitary," we surmise that the court did not equate the term with the unitary status requiring dismissal of the action. Rather, "[t]he court may have been stating merely that a constitutionally acceptable desegregation plan was implemented ... thus making the school unitary in some respects.” Pitts v. Freeman, 755 F.2d 1423, 1426 (11th Cir.1985); accord United States v. Lawrence County School District, 799 F.2d 1031, 1037-38 (5th Cir.*13031986). Thus, it was not necessary for, and the district court did not compel, the government to prove discriminatory intent, a requirement appropriate only after a judicial finding of full unitary status. Pitts v. Freeman, supra, 755 F.2d at 1426.
. Under the terms of the new policy, the County, as either the sending or receiving district, must deny all transfer requests or justify them by finding that desegregation efforts will not be impeded. Even if such transfers potentially would thwart desegregation, transfers are still permitted under certain limited exceptions. The policy was neither submitted to the district court for approval nor embodied into a court decree. However, we do not pass today on the legality of Butler County’s new transfer policy, since Conecuh and Covington County transfers to the McKenzie School are not affected by this policy.
. The government challenges the district court’s finding that McKenzie accepted 78 white students, rather than 85, from Conecuh County. The dispute arose over Butler County’s assertion that 7 of the 85 transferring students from Cone-cuh to Butler attend R.L. Austin Elementary School in Georgiana, which is predominantly black, rather than the McKenzie School. Our review of the record indicates that the court based its finding on conflicting evidence and was not clearly erroneous. We disagree, however, with the court’s finding that the cumulative effect of nonresident enrollment on McKenzie was 8.4%, which it computed by disallowing the 3 black transfers from Conecuh County (a .5% change) and subtracting the percentage of Butler County residents attending school in Conecuh and Covington Counties who otherwise would have gone to McKenzie (.4%). The cumulative effect of the transfers, without the 7 students contested by the government, is 9.3%. Our finding of a Singleton violation in this case would not preclude the continued transfer of the 3 black students from Conecuh County into the McKenzie School. In Lee v. Eufala City Board of Education, 573 F.2d 229, 232 n. 6 (5th Cir.1978), the court, in discussing a one-for-one exchange policy, stated that "a ruling that only black students may transfer into the ... school system would itself run afoul of Singleton, which clearly prohibits acceptance of transfers on a racially discriminatory basis.” The court went on to explain, however, that "[i]t may well be the case ... that the ratio of black to white students in any given county may dictate that a larger number of black students may transfer into the Eufala system without reducing desegregation or reinforcing the existence of a dual school system in the districts involved." Id. Nor do we consider the effect of those Butler County students transferring out of McKenzie. Although the .9% difference between the district court’s computation and our calculation is negligible in this case, we note that only those transfers adversely affecting integrative efforts are prohibited.
. 419 F.2d 1211, 1218-19 (5th Cir.1969), cert. denied, 396 U.S. 1032, 90 S.Ct. 611, 24 L.Ed.2d 530 (1970). This court adopted as binding precedent all decisions of the former Fifth Circuit Court of Appeals in Bonner v. City of Prichard, 661 F.2d 1206, 1207 (11th Cir.1981) (en banc).
. The district court found that the cumulative effect upon the various schools from which Co-necuh and Covington students were transferring ranged from 3.4% to 6.6%. We note also that Conecuh County already has been declared fully unitary. Receipt of transfer students from that district cannot be said to impede Conecuh’s desegregation efforts where a dual system already has been disestablished. Otherwise, the scope of the remedy would impermissibly exceed the scope of the original constitutional violation. Cf. Milliken v. Bradley, 418 U.S. 717, 744-45, 94 S.Ct. 3112, 3127, 41 L.Ed.2d 1069, 1091 (1974).
. The former Fifth Circuit Court of Appeals noted in Lee v. Lee County Board of Education, supra, that “prior to the gloss which Eufala placed on the meaning of the language in the Singleton transfer clause, it was not unreasonable for a school district to interpret that language as prohibiting only those transfers which had a district-wide adverse impact.” 639 F.2d at 1262 n. 13.
. The government does not dispute that Butler County’s attendance zones were properly drawn along constitutionally permissible guidelines. In fact, a sizeable black residential community, Pigeon Creek, was intentionally incorporated within the southern zone in order to increase that zone’s black percentage.
.Local school boards are state agents for purposes of the fourteenth amendment. Cooper v. Aaron, 358 U.S. 1, 16, 78 S.Ct. 1401, 1408, 3 L.Ed.2d 5, 15 (1958).
. McKenzie's enrollment without the out-of-county transfers would be 55.1% white. The inclusion of the 94 transfer students raises this figure to 63.9% In Eufala, supra, the court noted that although a strict quantitative analysis was improper, "the range of deviation may be significant in measuring qualitative segregative effect. For example, a transfer program which has the effect of increasing the black student population in a particular school from 90% to 100% may be more suspect than a corresponding 10% increase from 50% to 60%." 573 F.2d at 233 n. 9.
. Butler County’s terminal desegregation order does not proscribe such transportation agreements per se.
. The government does not appeal the district court's determination that the cumulative effect of returning these children to schools in the Evergreen attendance zone was legally insignificant. Also, although not a defense to this suit, we take notice of the fact that Conecuh County has scheduled the opening of a new central high school in the fall of 1989, which consolidates the Evergreen, Repton, and Conecuh high schools. At the time of trial, Evergreen was 69% black and 30% white, while the other two high schools in the district were majority white. The new high school will produce a better mix with a racial balance of approximately 50% black and 50% white. Evergreen Elementary, Southside Elementary, and Marshall Middle School, however, which are all within the Evergreen attendance zone, remain majority black.
.However, we reject Butler County’s argument that these nonresident students are not really transfers at all, but rather "full-fledged historical members of Butler County's educational community.” As the former Fifth Circuit Court of Appeals stated in Lee v. Lee County, supra, “Milliken explicitly teaches that school district lines are not to be casually ignored or considered as ‘no more than arbitrary lines on a map drawn for "political convenience’’.'" 639 F.2d at 1255 (quoting Milliken v. Bradley, supra, 418 U.S. at 741, 94 S.Ct. at 3125, 41 L.Ed.2d at 1089).
. The government’s trial exhibits show extensive transfers of white students from Conecuh County’s majority black schools, Evergreen Elementary, Southside Elementary, Marshall Middle, and Evergreen High, to the Red Level School in Covington.