GRUENDER, Circuit Judge.

In these consolidated appeals regarding continuing school desegregation efforts in the Little Rock, Arkansas metropolitan area, North Little Rock School District ("NLRSD") and Pulaski County Special School District ("PCSSD") each appeal the district court's denial of their petitions for a declaration of unitary status, opposed by appellee intervenors representing the class of black children harmed by segregation ("Joshua Intervenors"). In addition, NLRSD and PCSSD join with Little Rock School District ("LRSD"), several local teachers' unions, and a union member1 in appealing the district court's decision to terminate certain funding obligations of the State of Arkansas through its Department of Education (collectively, "the State") arising from a previous settlement agreement in this case. We reverse the partial denial of NLRSD's petition for unitary status, affirm the partial denial of PCSSD's petition for unitary status, and vacate the portion of the order terminating the State's funding obligations.

I. BACKGROUND

This is another chapter of a case that has been before us repeatedly since 1982, when LRSD sued NLRSD, PCSSD, and the State to obtain a remedy for the effects of their segregation practices on LRSD. The Joshua Intervenors soon intervened to protect the interests of local black students. We have held that "the long history of concurrent actions on the part of the state, PCSSD, and NLRSD exerted an unmistakable interdistrict effect on the schools of the [Little Rock] metropolitan area by singling out LRSD as the school district which provided some educational opportunities for black students and by identifying PCSSD and NLRSD as white districts." Little Rock Sch. Dist. v. Pulaski Cnty. Special Sch. Dist. No. 1, 778 F.2d 404, 427 (8th Cir. 1985) (en banc). Under a comprehensive series of settlement agreements reached in 1988 and 1989 and adopted as consent decrees, NLRSD and PCSSD each agreed to implement detailed plans to remedy segregation violations within their districts, while the State agreed to provide funding for an interdistrict remedy among the three districts, including the creation of magnet schools and majority-to-minority student transfers. See Little Rock Sch. Dist. v. Pulaski Cnty. Special Sch. Dist. No. 1, 921 F.2d 1371 (8th Cir.1990).2

After some court-approved or stipulated revisions, NLRSD's plan to remedy its intradistrict violations reached its current form in 1992 ("1992 Plan"), and PCSSD's plan to remedy its intradistrict violations reached its current form in 2000 ("Plan 2000"). Over the years, each district has obtained partial release from court supervision in some of the problem areas addressed by their intradistrict plans, and NLRSD and PCSSD filed the instant petitions for unitary status in all remaining areas in 2007. At the time of filing, NLRSD remained non-unitary in a total of nine areas and PCSSD remained non-unitary in a total of twelve areas.

The district court held a combined hearing on both petitions in early 2010 and accepted extensive testimony and evidence. In May 2011, the district court granted the petitions in part but denied unitary status for NLRSD in one area and for PCSSD in nine areas. In addition, the district court sua sponte proclaimed an end to most of the State's funding obligations under the 1989 settlement agreement. This set of appeals followed.

II. DISCUSSION

We review the district court's legal conclusions de novo, its findings of fact for clear error, and its modification of a consent decree for abuse of discretion. Little Rock Sch. Dist. v. North Little Rock Sch. Dist., 451 F.3d 528, 531 (8th Cir.2006). A constitutional violator seeking relief from a desegregation plan adopted as a consent decree must show that it "complied in good faith with the desegregation decree since it was entered" and that "the vestiges of past discrimination ha[ve] been eliminated to the extent practicable." Freeman v. Pitts, 503 U.S. 467, 492, 112 S.Ct. 1430, 118 L.Ed.2d 108 (1992) (quoting Bd. of Educ. of Okla. City Pub. Sch. v. Dowell, 498 U.S. 237, 249-50, 111 S.Ct. 630, 112 L.Ed.2d 715 (1991)). We evaluate the school districts' objective compliance with the terms of their respective plans in the same fashion as we would evaluate compliance with the terms of any contract, simply by "applying the terms of [the] contract between the parties to facts that have arisen since its creation." Little Rock Sch. Dist. v. Pulaski Cnty. Special Sch. Dist. No. 1, 83 F.3d 1013, 1017 (8th Cir.1996). The "good faith" aspect of compliance with a desegregation decree, however, is more specific than in contract law, and has been described as

Freeman, 503 U.S. at 491, 112 S.Ct. 1430 (addressing good faith as a factor in considering a declaration of partial unitary status). Finally, we are mindful that "federal supervision of local school systems was intended as a temporary measure to remedy past discrimination." Missouri v. Jenkins, 515 U.S. 70, 88, 115 S.Ct. 2038, 132 L.Ed.2d 63 (1995) (quoting Dowell, 498 U.S. at 247, 111 S.Ct. 630) (emphasis added).

A. NLRSD's Petition for Unitary Status

The sole area in which the district court denied a finding of unitary status to NLRSD is the area of staff recruitment. Section 2 of the 1992 Plan requires NLRSD to take numerous actions intended to increase the number of black teachers, principals, and administrators throughout the district, including conducting surveys within the district to identify black teachers interested in advancement to administrative or principal positions; targeting "regional colleges and universities which have strong teacher education programs with significant black enrollment" with promotional recruiting kits and on-campus recruiting visits; contacting "[b]lack community leaders" and "community organizations" to obtain employment referrals for black teachers and administrators; providing professional associations of teachers and administrators with recruiting materials; producing a ten-minute slide or video presentation for use in recruitment activities; and advertising vacancies in "newspapers and other media in communities that have colleges and universities targeted for recruitment efforts." In addition, section 2A of the 1992 Plan requires NLRSD to employ a labor economist "to conduct an analysis of the pool of qualified applicants in the relevant labor market" for each hiring category (teachers, principals, administrators, and coaches), and then to set numerical hiring goals for blacks in each category and a timetable for reaching those goals.

There appears to be little dispute that NLRSD has satisfied the 1992 Plan with respect to black principals, administrators, and coaches. The parties' dispute turns instead on NLRSD's efforts to recruit black teachers. The district court denied unitary status in this area based on the good-faith prong of Freeman, finding that NLRSD had failed to document its compliance efforts sufficiently to prove its commitment.

NLRSD presented extensive testimony and other evidence regarding its efforts to comply with the teacher recruitment provisions of the 1992 Plan, which may be summarized as follows. NLRSD complied with the 1992 Plan's requirements regarding job postings and advertisements, publicity materials, and internal training and promotion opportunities. For five or six years, NLRSD also recruited heavily at colleges throughout the South with "strong teacher education programs with significant black enrollment," as required by the 1992 Plan. NLRSD's recruiters made many on-campus recruiting visits to such schools and even were authorized to hire black applicants "on the spot," but NLRSD discovered that qualified black graduates from outside Arkansas generally were not receptive to relocating to the Little Rock area. Market competition for black teachers is intense, as other school districts throughout the South likewise have been making efforts to increase black staff. One NLRSD recruiter described the typical on-campus minority recruiting visit as involving "a hundred recruiters and [twenty] graduates." Moreover, NLRSD did not have the resources to compete with many other districts. For example, some school districts in Texas were offering signing bonuses to minority candidates, while others offered starting annual salaries as much as $11,000 higher than NLRSD.

Thereafter, NLRSD refocused its efforts to in-state colleges and had slightly more success, although it faced similar market pressure from out-of-state districts recruiting at Arkansas colleges. In addition, NLRSD faced in-state market pressure, as Arkansas itself has about 280 school districts, and only about 100 black education majors graduate from Arkansas colleges each year. Through 2006, NLRSD's starting salaries remained significantly lower than LRSD and PCSSD, hampering its efforts to compete even locally.

Around 2006, NLRSD changed strategies again. It subscribed to a number of advanced tracking services for potential black applicants, reduced costs, and redirected funds to increase its starting salaries. The number of newly hired black teachers has increased markedly as a result. While the percentage of newly hired employees who were black was 9.9 and 6.0 percent in 2004-2005 and 2005-2006 respectively, in the next four fiscal years the percentages rose to 21.3, 12.6, 17.7, and 23.3 percent respectively. As a result, blacks now comprise 16.7 percent of the certified educators at NLRSD (up from 13 percent), while only about 9 percent of the certified educators in the state are black.

Despite this extensive testimony and evidence describing NLRSD's efforts, the district court found that NLRSD did not submit sufficient documentation of those efforts and that its evidence was too anecdotal to prove good faith. For example, NLRSD's chief recruiter could not recite yearly statistics for the number of offers made to, and rejected by, black candidates. As a result, the district court ordered NLRSD to continue its efforts, but keep more detailed records, for 24 additional months and then ask for reconsideration of unitary status.

The Joshua Intervenors bemoan the absence of firm statistics on employment offers, but they do not direct us to any evidence that would tend to discredit the extensive evidence of NLRSD's efforts.3 Notably, the district court did not suggest that any portion of NLRSD's testimony was not credible, nor did it suggest that some different or additional action by NLRSD would be necessary for compliance. Instead, the district court simply asked for the same actions, but with more detailed documentation, for an additional 24 months.

While the goal of documenting compliance over time is a laudable one, the district court abused its discretion by imposing new data collection and reporting requirements with respect to staff recruitment that were not agreed to by the parties in the 1992 Plan. "When construing a consent decree, courts are guided by principles of contract interpretation and, where possible, will discern the parties' intent from the unambiguous terms of the written consent decree, read as a whole." Pure Country, Inc. v. Sigma Chi Fraternity, 312 F.3d 952, 958 (8th Cir.2002). Here, the 1992 Plan contains a separate section on desegregation monitoring that required NLRSD to "design a detailed reporting/monitoring system describing the information to be collected and reported, the format of reports and the frequency of reports." The district court found that NLRSD followed these requirements in good faith, stating that "nothing came close to controverting this finding." Because NLRSD satisfactorily complied with its duty to specify and maintain "the information to be collected and reported" as expressed in the 1992 Plan, there was no basis for the district court to impose upon NLRSD, with no advance notice, a more extensive set of collection and reporting requirements with respect to staff recruiting. See Holland v. N.J. Dep't of Corr., 246 F.3d 267, 281 (3d Cir.2001) ("A court should interpret a consent decree as written and should not impose terms when the parties did not agree to those terms."); EEOC v. N.Y. Times Co., 196 F.3d 72, 78 (2d Cir.1999) ("[A] court may not replace the terms of a consent decree with its own, no matter how much of an improvement it would make in effectuating the decree's goals." (quotation omitted)).

Apart from the perceived deficiency in record-keeping, the district court found no fault with NLRSD's efforts to comply with the staff recruitment provisions of the 1992 Plan. Our examination of the record confirms that NLRSD's staff recruitment efforts, as described at the hearing, are sufficient to "demonstrate[], to the public and to the parents and students of the once disfavored race, its good-faith commitment to the whole of the court's decree." Freeman, 503 U.S. at 491, 112 S.Ct. 1430. Therefore, we turn to the question of whether "the vestiges of past discrimination ha[ve] been eliminated to the extent practicable" in the area of recruitment. Freeman, 503 U.S. at 492, 112 S.Ct. 1430 (quoting Dowell, 498 U.S. at 250, 111 S.Ct. 630).

The district court noted that, statistically, the percentage of teachers employed by NLRSD who are black actually declined from 20 percent in the 1984-85 school year4 to as low as 13 percent in the 2006-07 and 2007-08 school years. However, as the district court correctly observed, the proper standard under Freeman is not statistical outcomes, but rather good-faith compliance with the 1992 Plan and elimination of the vestiges of past discrimination to the extent practicable. See 503 U.S. at 492, 112 S.Ct. 1430. The required labor market study, performed in 1988, determined that the relevant labor market is the State of Arkansas and that the percentage of qualified blacks in that market was 11.9 percent.5 The fact that NLRSD has maintained levels of black teacher employment (and levels of new hiring of black teachers, at least since 2006) that exceed the percentages in the relevant labor market, viewed in the light of NLRSD's goodfaith efforts, suggests that the vestiges of past discrimination have been eliminated to the extent practicable in the area of recruitment.

As a result, we reverse the denial of unitary status for NLRSD in the area of staff recruitment.

B. PCSSD's Petition for Unitary Status

The district court denied PCSSD's petition for unitary status in nine areas addressed by Plan 2000: (1) student assignment, (2) advanced placement, gifted and talented, and honors programs, (3) discipline, (4) school facilities, (5) scholarships, (6) special education, (7) staff, (8) student achievement, and (9) monitoring. The district court relied in large part on evidence of PCSSD's lack of good faith, finding at the outset that "it seems [PCSSD] has given very little thought, and even less effort, to complying with its desegregation plan. Complying with its plan obligations seems to have been an afterthought." We find no reason to disagree with the district court's observation.

PCSSD concedes a "failure to perform some of the tasks to which it committed." However, it relies on a number of statistical comparisons to argue that its desegregation outcomes are better than, or at least comparable to, "outcomes reported by other unitary districts," including LRSD and NLRSD. PCSSD asserts that in Liddell v. Special Sch. Dist., 149 F.3d 862 (8th Cir.1998), we excused a similar lack of good faith in implementing a desegregation plan when we stated that "we did not hold or imply that each of the twelve ordered goals ... must be achieved to meet constitutional requirements for the establishment of a unitary district," id. at 871, purportedly to justify a finding of partial unitary status despite a lack of good-faith compliance with the twelve goals. PCSSD asks us to likewise hold that good-faith compliance with the elements of Plan 2000 is unnecessary for the establishment of a unitary district.

PCSSD's argument is misguided because it conflates three separate holdings from Liddell. Liddell addressed "a dual system of vocational education: a predominantly white district in St. Louis County and a predominantly black district in the city of St. Louis." Id. at 864. As a remedy, a Special School District, or SSD, had been designated as the sole provider of vocational education for both the city and county. Id. at 865. In the order being appealed, the district court found that "SSD's failure to exert good faith efforts to establish a city vocational education facility and failure to provide a quality program preclude a conclusion that SSD has achieved unitary status." Id. at 868. First, we reversed in part, finding that SSD had achieved "partial unitary status in operating the vocational education program in the county schools." Id. at 869 (emphasis added). Second, we affirmed "that full unitary status has not been achieved" until such time as a "vocational high school in the city is firmly established and functioning." Id. (emphasis added). Finally, addressing the ongoing court supervision of the city system, we held that twelve desegregation goals announced by the district court were unnecessary in light of constitutionally adequate state statutory requirements. Id. at 871.

Contrary to PCSSD's assertion, then, we did not declare unitary status for a district that failed to make good-faith efforts to comply with the twelve-goal plan. Instead, we declared partial unitary status for the county portion of the district, affirmed the denial of unitary status for the city portion due to a lack of good faith in establishing that system, and then separately addressed the utility of the twelve goals in the continuing supervision of the city system. Indeed, part of our prior mandate to the district court had been to determine in the first instance if achievement of those twelve goals, adopted by the district court from an earlier report of a desegregation monitoring agency, would even be necessary to establish unitary status. See Liddell ex rel. Liddell v. Bd. of Educ. of the City of St. Louis, 121 F.3d 1201, 1216 & n. 2 (8th Cir.1997). Because it was clear prior to the district court's order that the twelve goals had not yet received court approval, SSD's compliance with the twelve goals could not possibly have been at issue.

In short, Liddell does not support the proposition that we can excuse a school district's broad failure to comply with a desegregation plan to which it agreed, even if its desegregation statistics appear favorable relative to other unitary districts. We reemphasize that a constitutional violator seeking relief from a desegregation plan adopted as a consent decree must show both that it "complied in good faith with the desegregation decree since it was entered" and that "the vestiges of past discrimination ha[ve] been eliminated to the extent practicable." Freeman, 503 U.S. at 492, 112 S.Ct. 1430. With this in mind, we address each disputed area of Plan 2000 in turn.

1. Student assignment

Section C of Plan 2000 imposes two requirements on student assignment. The first requirement, carried forward from a previous settlement agreement, is that black students comprise at least 20 percent of the student body of each of PCSSD's schools, with exceptions for remotely located schools.6 In addition, the maximum allowable percentage of black students at any individual elementary or secondary school is the district-wide percentage of black elementary or secondary enrollment, respectively, for the school year, multiplied by 1.25. The district court found that, although a few schools fell outside the allowable range in particular years, PCSSD substantially complied with this requirement. The Joshua Intervenors do not dispute that finding.

The second requirement is that PCSSD prepare "one-race reports" in October of each year, listing every class that is comprised entirely of students of a single race, a description of the steps taken to eliminate the single-race aspect of the class, and the reason why this proved infeasible. The district court found that a number of the required reports were missing and that the majority of the reports filed failed to include any description of the steps taken to eliminate the one-race class and why it proved infeasible to do so. As a result, the district court found that PCSSD had not substantially complied with this section of Plan 2000.

PCSSD does not argue that the district court's findings as to the one-race reports were clearly erroneous. Instead, it argues that few students are affected by singlerace classrooms and that single-race classes occur because some elective classes, such as automotive shop, are "selected primarily by non-black students." The parties' agreement in Plan 2000, however, contains no reporting exception for such cases. It is not clear to us how PCSSD's opinion that the explanations for single-race classrooms are non-discriminatory would excuse PCSSD's failure to produce the plain and simple reports expressly required by Plan 2000.

PCSSD also asks us to separate the two requirements of Section C of Plan 2000 and declare PCSSD unitary in the sub-area of assignment of students to schools. While we have "discretion to order the incremental withdrawal of [court] supervision in a school desegregation case," in doing so we must consider factors such as "whether retention of judicial control is necessary or practicable to achieve compliance with the decree in other facets of the school system ... and whether the school district has demonstrated, to the public and to the parents and students of the once disfavored race, its good-faith commitment to the whole of the court's decree." Freeman, 503 U.S. at 491, 112 S.Ct. 1430. Initially, we note that the parties agreed in Plan 2000 to place the two requirements together under a single section entitled "Assignment of Students," indicating that the parties understood the two requirements to be connected. See Pure Country, Inc. v. Sigma Chi Fraternity, 312 F.3d at 958 ("[We] discern the parties' intent from the unambiguous terms of the written consent decree, read as a whole."). This makes sense, as the goal of preventing segregation of black students through disproportionate placement into individual schools within the district would be severely undermined if the district could then, without explanation, isolate black students in single-race classrooms within each school.

Under the circumstances, the release of judicial control over the sub-area of assignment of students to schools would not be conducive to achieving compliance with at least one other facet of the decree, the sub-area of reporting on single-race classrooms. See Freeman, 503 U.S. at 491, 112 S.Ct. 1430. In addition, PCSSD's dismissive approach to the one-race reporting requirement has done nothing to "demonstrate[], to the public and to the parents and students of the once disfavored race, its good-faith commitment to the whole of the court's decree." Id. As a result, we deny PCSSD's request to declare PCSSD unitary in the sub-area of assignment of students to schools.

Accordingly, we affirm the denial of unitary status for PCSSD in the area of student assignment.

2. Advanced placement

Section D of Plan 2000 requires PCSSD to implement certain standards to promote racial diversity in its advanced placement, gifted and talented, and honors programs. Some of those standards are embodied in PCSSD's 1998-1999 Advanced Placement Guidelines, which set an aspirational 8 percent limit on the variance between overall black enrollment at a school and advanced-placement black enrollment at each school. As the district court explained, for example, if a given school's enrollment is 60 percent white and 40 percent black, then the guidelines are satisfied as long as the school's black enrollment in advanced placement courses exceeds 32 percent. The district court found that "a substantial portion" of PCSSD's advanced programs failed to meet this standard, and PCSSD does not challenge that finding.

The district court correctly noted that failure to meet the aspirational 8 percent goal could be excused if PCSSD demonstrated that it had complied in good faith with its plan obligations. The 1998-1999 Advanced Placement Guidelines list eight recruitment strategies that PCSSD agreed to pursue in order to improve black enrollment in advanced placement courses. However, the district court found that PCSSD "simply failed to show that it has done anything to implement the eight goals set forth in the 1988-1999 Guidelines." As a result, the district court denied unitary status in the area.

PCSSD makes no representation that it attempted to implement the agreed-upon strategies. Instead, it merely argues that its advanced-placement enrollment disparities are less pronounced than those of NLRSD, which achieved unitary status in this area. Unfortunately, mere comparisons to other unitary districts are insufficient to satisfy Freeman. Because PCSSD has not even attempted to implement its plan, we have no basis on which to decide if "the vestiges of past discrimination ha[ve] been eliminated to the extent practicable" at PCSSD. See Freeman, 503 U.S. at 492, 112 S.Ct. 1430. More importantly, PCSSD, an adjudged constitutional violator, has done nothing to assure the public of any permanent good-faith commitment to promoting racial diversity in its advanced placement courses. See id. at 491, 112 S.Ct. 1430. Therefore, we affirm the denial of unitary status for PCSSD in the area of advanced placement.

3. Discipline

Section F of Plan 2000 requires PCSSD to (1) collect detailed data on each disciplinary incident that occurs within the district, (2) develop criteria for identifying individual teachers with problematic disciplinary practices, individual schools with atypically high discipline rates, and individual schools with atypically high racial disparities in discipline, (3) make specific efforts to work with the identified teachers and schools to eliminate racial disparities, (4) conduct a comprehensive study of the disciplining of black male secondary-school students which suggests prevention and intervention measures, (5) develop initiatives to reduce disciplinary rates within the district, and (6) adhere to certain handbook policies with respect to the disciplinary hearing and appeals process. The district court found that PCSSD failed to comply with items (3), (4), and (5).

PCSSD does not challenge the district court's compliance findings. Instead, it argues that racial disparities in discipline within the district, measured by comparing the rate of suspensions of black students to the rate for non-black students, are lower than the national average. Once again, PCSSD contends that its outcomes are sufficient to demonstrate unitary status, regardless of its failure to implement the relevant sections of the plan to which it agreed. Once again, because PCSSD has not even attempted to implement its plan, we have no basis on which to decide if "the vestiges of past discrimination ha[ve] been eliminated to the extent practicable" at PCSSD and no evidence of a permanent good-faith commitment to the goal of eliminating racial disparities in discipline. See Freeman, 503 U.S. at 491-92, 112 S.Ct. 1430. Accordingly, we affirm the denial of unitary status for PCSSD in the area of discipline.

4. School facilities

Section H of Plan 2000 requires PCSSD to (1) prepare a plan, with the help of consultants as needed, "so that existing school facilities are clean, safe, attractive and equal," (2) build two new schools in specific areas, (3) not close schools in predominantly black areas absent compelling necessity, and (4) notify the Joshua Intervenors of any plan to build a new school or add student capacity to an existing school. The district court found that PCSSD failed to act in good faith to make its school facilities clean, safe, attractive, and equal.

To satisfy the first requirement, PCSSD commissioned a construction company in 1999 to analyze existing school facilities in the district and to list major concerns and estimated repair or replacement costs for each facility. The 1999 study recommended building ten new schools at a projected cost of more than $230 million. However, Pulaski County voters rejected a millage of $110 million for new construction, and PCSSD completed only three new schools before the 2010 hearings—the two required by item (2) above, Bates Elementary and Maumelle Middle School, plus one additional school, Chenal Elementary, an interdistrict school. Breaking down the cost of each of those facilities on a per-student basis reveals that PCSSD spent only $8,150 per student on Bates, located in a predominantly black area, but spent about $22,000 per student and $25,000 per student respectively on Maumelle and Chenal, located in predominantly white areas.7 PCSSD baldly states that inflation is to blame for the cost difference, but it makes no attempt to quantify how inflation caused costs to almost triple between 2001, when Bates was opened, and 2005, when Maumelle was opened, or to more than triple between 2001 and 2008, when Chenal was opened. Moreover, the district court credited testimony that Chenal, for example, was built to "Mercedes-Benz" standards, while Bates was not.

Similarly, PCSSD was in the process of replacing Oak Grove High School, also located in a predominantly white area, with an entirely new facility costing approximately $58 million, despite the 1999 study's finding that Oak Grove High School could be made adequate with renovations. Meanwhile, several schools in predominantly black areas that were identified as in need of replacement by the 1999 study continue to "languish in relatively poor condition," as the district court found, with "broken commodes, falling ceiling tiles, holes in the ceiling, [and] exposed wiring." Likewise, PCSSD justified its plan to construct a new wing on Pine Forest Elementary, located in a predominantly white area, as necessary to eliminate the use of trailers as classrooms, while continuing to install trailers as classrooms at a number of schools located in predominantly black areas.

In response to these detailed factual findings, PCSSD relies on a scatter plot of the year each of its schools was built against the percentage of black enrollment at that school. The scatter plot shows that PCSSD has both older school facilities with relatively low black enrollment and older facilities with relatively high black enrollment, and PCSSD argues that it does not have the funds to replace all of them at once. These facts, however, are not inconsistent with the district court's findings with respect to PCSSD's recent efforts under Plan 2000. PCSSD also argues that the Joshua Intervenors should not be allowed to dispute the chosen location of each new school because the Joshua Intervenors did not file separate motions to challenge the plans for each new school, as they were permitted to do under Plan 2000. However, whether the Joshua Intervenors chose to devote their efforts to challenging each instance of construction has nothing to do with whether PCSSD has demonstrated good-faith compliance with its duties under Plan 2000.

We find no clear error in the district court's factual findings that PCSSD has devoted a disproportionate share of its facilities spending to predominantly white areas. See Little Rock Sch. Dist., 451 F.3d at 531. We also agree with the district court that these findings demonstrate an absence of good faith in PCSSD's efforts to comply with the facilities requirements of Plan 2000. As a result, we affirm the denial of unitary status for PCSSD in the area of facilities.

5. Scholarships

Section I of Plan 2000 requires PCSSD to "establish a bi-racial committee to explore a program for providing college scholarships to designated PCSSD students" within 30 days after LRSD established such a program. LRSD established its program in April 1999, but PCSSD did not establish its exploratory committee until 2002. In April 2003, PCSSD's committee finally stated that funds were not available to finance any scholarships and that it would award scholarships when funds were available. To date, no scholarships have been awarded. The district court found that PCSSD failed to act in good faith to explore ways to make college scholarships available.

PCSSD argues that it complied with Plan 2000 simply by establishing the committee, albeit three years late. However, merely establishing a committee is insufficient under Section I unless the committee made a genuine good-faith effort to "explore a program for providing college scholarships to designated PCSSD students." PCSSD directs us to no evidence that its committee made any effort whatsoever after April 2003 to balance potential scholarship funding against other priorities or to develop any parameters for evaluating how or when such a program might become feasible. Yet again, PCSSD has done nothing to demonstrate to the public and the parents and students of the once disfavored race that it intends to honor its commitment in good faith. See Freeman, 503 U.S. at 491, 112 S.Ct. 1430. Accordingly, we affirm the denial of unitary status for PCSSD in the area of scholarships.

6. Special education

Section K of Plan 2000 requires PCSSD to identify standards for preventing inappropriate referrals of black males and black kindergarten students to special education and to develop a "specific plan" for monitoring of "schools where there are atypically high racial disparities in special education classification, generally or as to black male students," including "criteria for identifying schools for monitoring." Section K also requires PCSSD to submit copies of these standards and criteria to the Joshua Intervenors.

Pursuant to Plan 2000, PCSSD notified the Joshua Intervenors that an 8.3 percent standard deviation would serve as its criterion for identifying schools with "atypically high racial disparities in special education classification." In other words, a school has an "atypically high" racial disparity under Plan 2000 if the percentage of the school's students in special education who are black exceeds the percentage of its students overall who are black by more than 8.3 percentage points. The district court found that, although multiple individual schools each year failed to satisfy the 8.3 percent standard, PCSSD failed to develop a specific plan for those individual schools as required by Section K.

In response, PCSSD notes primarily that its district-wide percentage of students in special education who are black has not exceeded the district-wide percentage of students overall who are black by more than 8.3 percentage points while Plan 2000 has been in effect. PCSSD fails to explain, however, how this fact would relieve PCSSD of its obligation to develop specific plans for each individual school within the district that does not meet the standard each year, as expressly required by Section K. PCSSD also argues that its non-compliance is excusable because, over the past ten years, there have been only nine failures of individual high schools to meet the 8.3 percent standard in a given year, all occurring prior to 2006. The evidence cited by PCSSD, however, does not support PCSSD's summarization.8 Moreover, a much larger number of elementary and middle schools also have failed to meet the 8.3 percent standard. PCSSD agreed in Plan 2000 to develop specific plans for each of these instances, and it indisputably failed to do so. Therefore, we affirm the denial of unitary status for PCSSD in the area of special education.

7. Staff

Section L of Plan 2000 requires PCSSD to (1) recruit applicants for each available administrative position in a manner designed to develop a racially diverse pool of applicants, targeting applicants both external and internal to PCSSD, (2) recruit new teachers from a racially diverse pool of applicants and monitor the process to ensure that no policy, practice, or custom has the purpose or effect of limiting the proportion of black teachers, (3) implement programs, policies, and procedures to increase the number of black teachers in certain under-represented disciplines, including offering incentives to black teachers to obtain certification in those areas, and (4) allocate teachers and staff in a manner that avoids racially identifiable schools. The district court found that PCSSD failed to comply in good faith with any of the four requirements.

On appeal, PCSSD continues its strategy of focusing on outcomes, noting that, as at NLRSD, the percentage of its administrators and staff who are black exceeds the percentage of the relevant labor market that is black. As discussed above, however, outcomes alone are insufficient; PCSSD must also show that it "complied in good faith with the desegregation decree since it was entered." Freeman, 503 U.S. at 492, 112 S.Ct. 1430. The district court did not clearly err in finding that PCSSD has failed to comply with the express requirements of Section L.

With regard to recruitment of black administrators, the head of PCSSD's human resources department conceded at the 2010 hearing that she was unaware of any policies or standards at PCSSD designed to implement the requirements of Plan 2000, despite reports from the Office of Desegregation Monitoring in both 2003 and 2006 decrying the lack of such compliance. The district court also credited the testimony of a monitor for the Joshua Intervenors that several of PCSSD's former superintendents lacked enthusiasm for the idea of increasing the number of black administrators and that the applicant pools she was allowed to review were not, in fact, racially diverse. The district court's credibility determination is virtually unreviewable on appeal. All Am. Life Ins. Co. v. Billingsley, 122 F.3d 643, 648 (8th Cir. 1997). Likewise, with regard to recruitment of black teachers, the head of PCSSD's human resources department conceded that she was unaware of any policies or standards at PCSSD designed to implement the express affirmative monitoring requirements for the recruitment and hiring of new teachers. PCSSD also did not even track the number of black teachers in some of the listed under-represented disciplines it was required to address in subsection L(3). PCSSD makes no showing of clear error in these bases for district court's finding of non-compliance.

With regard to the allocation of teachers to avoid racially identifiable schools, PCSSD showed that the percentage of black staff at each school was within 15 percent of the district-wide average. The district court agreed that compliance with a 15-percent standard might be sufficient to meet the requirements of subsection L(4) had PCSSD expressly adopted such a "standard" at the outset of the plan, but found that PCSSD only adopted the 15-percent standard "at the eleventh hour" before the hearing as an ad hoc attempt to justify its outcomes, rendering it insufficient to prove a long-term good-faith commitment to the requirements of the plan. See Freeman, 503 U.S. at 491, 112 S.Ct. 1430. We do not disagree with the district court's application of the good-faith standard to these circumstances.

Finally, PCSSD challenges the district court's brief statement at the end of its analysis that the percentage of black staff at PCSSD has remained at 21 percent since 1984, while the percentage of black students has increased from 23 to 44 percent. To the extent, if any, that the district court implied the percentage of black students should be a guideline for establishing the percentage of black staff, the district court erred. Cf. Hazelwood Sch. Dist. v. United States, 433 U.S. 299, 308, 97 S.Ct. 2736, 53 L.Ed.2d 768 (1977) ("There can be no doubt ... that the District Court's comparison of Hazelwood's teacher work force to its student population fundamentally misconceived the role of statistics in employment discrimination cases. The Court of Appeals was correct in the view that a proper comparison was between the racial composition of Hazelwood's teaching staff and the racial composition of the qualified public school teacher population in the relevant labor market."); see also Wygant v. Jackson Bd. of Educ., 476 U.S. 267, 276, 106 S.Ct. 1842, 90 L.Ed.2d 260 (1986) ("Carried to its logical extreme, the idea that black students are better off with black teachers could lead to the very system the Court rejected in Brown v. Board of Education, 347 U.S. 483, 74 S.Ct. 686, 98 L.Ed. 873 (1954)"). Nevertheless, this error does not affect the bases discussed above for the district court's findings of an absence of good-faith compliance. As a result, we affirm the denial of unitary status for PCSSD in the area of staff.

8. Student achievement

Section M of Plan 2000 requires PCSSD to implement a plan designed by Dr. Stephen Ross ("the Ross Plan") to improve student achievement. The Ross Plan requires PCSSD to "improve educational achievement by all students, with special attention to African-American students and others who are at-risk of academic failure due to socioeconomic disadvantages, or other factors" and to "decrease the performance gap between white students and African-American students through the systematic design/selection and implementation of intervention programs that provide effective remediation and/or adaption to individual or group needs." The district court found that PCSSD failed to provide "special attention" to black students in its performance-enhancement strategies and failed to systematically design, select, and implement effective intervention programs.9

The Ross Plan required each school in PCSSD to prepare a Formative Education Plan for School Improvement that would outline how the school intended to increase student achievement and close the performance gap between black and white students. Because many of the requirements overlapped with plans later required under the State's Arkansas Consolidated School Improvement Planning ("ACSIP") program, PCSSD decided to implement its obligations under the Ross Plan within each school's ACSIP plan. Unfortunately, the ACSIP plans as implemented do not meet the express requirements of the Ross Plan. An independent study commissioned by PCSSD in 2006 found that the ACSIP plans, while addressing student achievement generally, in no way devoted "special attention" to black students as required by the Ross Plan.

To counter the findings of the independent study, PCSSD produced its own report in 2009. The highlight of the report is that the gap between the percentage of black students and white students scoring "proficient or above" on Arkansas benchmark tests decreased by 9 percent in math and 6 percent in literacy from 2005 to 2009. While this is commendable, it is a limited statistic. Individual scores on the test are grouped into four categories, "below basic," "basic," "proficient," and "advanced." See Testing: Student Assessment, http://arkansased.org/testing/ assessment.html (last visited Dec. 16, 2011). A wide range of outcomes are encompassed in each category, and the statistics for "proficient and above" cited by PCSSD do not reveal relative achievement levels within the categories. For example, if the scores of all white students had made a large jump from the low end of the "proficient" level in 2005 to the "advanced" level in 2009, while the scores of all black students made a small jump from the high end of the "basic" level in 2005 to the low end of the "proficient" level in 2009, PCSSD's chosen statistic would suggest that the achievement gap had closed enormously, when in reality the gap had increased. Similarly, comparing changes in the "proficient and above" range completely ignores relative movement within the "below basic" and "basic" categories. The Joshua Intervenors note that PCSSD has the individual scores and could have presented a more informative analysis of the achievement gap, but PCSSD chose not to do so.

In any event, we must note again that outcomes are not determinative of good-faith compliance. More significant is the fact that PCSSD's 2009 report fails to prove that PCSSD complied with its obligation under Plan 2000 to design, select, and implement specific "intervention programs [to] decrease the performance gap." Instead, the district court found that PCSSD reported primarily on programs that existed prior to the adoption of Plan 2000 (and thus could not have been designed and implemented to comply with it), plus a few programs that were established only after PCSSD decided to file the instant motion for unitary status. Moreover, the district court found that, even when specific programs were shown to be effective, PCSSD failed to expand them beyond a limited number of students.

PCSSD does not challenge these findings, contending instead that it should be excused from its obligations because of a previous finding that "socioeconomic factors are the root cause for most, if not all, of the achievement gap." See Little Rock Sch. Dist., 237 F.Supp.2d at 1074. Regardless of whether the specific intervention programs required by Plan 2000 eventually bear fruit, however, PCSSD cannot disavow its agreed-upon obligation to make a good-faith effort. Accordingly, we affirm the denial of unitary status for PCSSD in the area of student achievement.

9. Monitoring

Section N of Plan 2000 requires PCSSD to monitor its efforts in three ways: (1) develop a plan for particular staff members to focus on monitoring and complying with specific elements of Plan 2000 and inform the Joshua Intervenors as to which staff member is responsible for which element, (2) permit the Joshua Intervenors to meet with staff members and to examine and copy certain records related to Plan 2000, and (3) submit certain statistical reports. The district court found that PCSSD satisfied the second and third requirements but failed to satisfy the first requirement.

Once again, PCSSD does not challenge the district court's underlying factual findings, but argues briefly that its compliance with requirements (2) and (3) provides a basis to excuse its failure to comply with requirement (1). We disagree. As should be clear by now, PCSSD cannot pick and choose among the requirements of Section N and expect to achieve unitary status in the area of monitoring. Therefore, we affirm the denial of unitary status for PCSSD in the area of monitoring.

C. The State's Funding Obligations

Under the 1989 settlement agreement, the State agreed to provide funding for an interdistrict remedy among the three districts. For example, the State must pay half the cost of operating magnet schools, the full cost of transporting students to those schools, a portion of teacher retirement and health benefits, and the cost of recruiting and transporting majority-to-minority student transfers. These obligations currently amount to about $38 million per year. The State has long made known its view that such funding will become unnecessary after all three districts are declared fully unitary. However, the State has not yet moved for relief from its funding obligations, and the scheduling order for the 2010 hearings on NLRSD's and PCSSD's petitions for declaration of unitary status did not provide for the presentation of any evidence regarding such relief. In addition, although the State participated in the hearings, it objected frequently that the State's own duties and obligations were not the subject of the hearings.

Nevertheless, the district court sua sponte released the State from the funding obligations listed above,10 finding this change necessary to avoid "an absurd outcome in which the districts are rewarded with extra money from the state if they fail to comply with their desegregation plans and they face having their funds cut by the state if they act in good faith and comply." It did not make specific findings of fact to support the remedy, but rather stated in general terms that a carrot and stick approach would no longer work with "these districts" because the districts "are wise mules that have learned how to eat the carrot and sit down on the job." Given that PCSSD is the only district involved that continues to show a broad lack of good faith, we presume that the district court was referring primarily to PCSSD. Of course, the termination of State funding heavily impacts LRSD and NLRSD as well.

The district court's frustration is understandable, and its conclusions regarding the perverse incentives created by the State's funding may well have some merit. Nevertheless, notice and a formal hearing are required before the court terminates a constitutional violator's desegregation obligations. See Jenkins v. Missouri, 216 F.3d 720, 727 (8th Cir.2000). While the State correctly notes that the issue of termination of State funding was presented to the district court through other avenues, including briefing in response to a July 29, 2009 request for information from the district court and discussion at a September 30, 2009 status hearing, we have rejected the notion that such other avenues may substitute for a formal evidentiary hearing:

Liddell, 121 F.3d at 1216.

Similarly, the briefings and status hearing referred to by the State in this case cannot be cobbled together to form an "adequate record," particularly in the absence of detailed findings by the district court. Instead, if the State wishes to obtain relief from its funding obligations, there must be a formal evidentiary hearing on the issue "followed by comprehensive and detailed findings of fact and conclusions of law," as envisioned by Jenkins and Liddell. We express no opinion on what the outcome of such a hearing should be. In the absence of these procedures, the portion of the district court's order terminating the State's funding obligations under the 1989 Settlement Agreement is vacated.

III. CONCLUSION

For the foregoing reasons, we reverse the partial denial of NLRSD's petition and direct the district court to declare unitary status for NLRSD, affirm the partial denial of PCSSD's petition for unitary status, vacate the portion of the order terminating of the State's funding obligations, and remand for further proceedings consistent with this opinion.