Full opinion text
BERZON, Circuit Judge: On January 24, 2000, the District Court for the District of Arizona ruled that Arizona was violating the Equal Educational Opportunity Act of 1974, 20 U.S.C. §§ 1701 et seq. (“EEOA”), because the state’s funding for English language instruction for non-native speakers was “not reasonably calculated to effectively implement the ... educational theory which” the state had approved. Flores v. Arizona, 172 F.Supp.2d 1225, 1239 (“Flores II ”) (D.Ariz.2000). Eight years later, Arizona has still not satisfied the terms of that judgment, nor complied with the bulk of the injunctions entered against it as a result of that ruling. Although Arizona and the Arizona Board of Education acknowledge that the state remains out of compliance and do not seek to vacate the judgment or the injunctive orders, the Arizona Superintendent of Public Instruction, the Speaker of the Arizona House of Representatives, and the President of the Arizona Senate moved for relief from judgment. We affirm the district court’s denial of relief. I. BACKGROUND It has been more than fifteen years since the initial complaint in this action was filed and eight years since the final judgment on liability issued. The post-judgment relief process has been arduous, with Arizona moving forward, to the considerable degree it has, largely in response to a consent decree resulting from this litigation and a series of post-judgment relief orders. Over this long period, local and national conditions have shifted in significant ways. The moving parties contend they have done so sufficiently as to warrant relief from judgment, even though Arizona has never complied with the specific terms of the present injunction against it. Explaining why that argument fails requires rehearsing the history of this case. We do so here, beginning with the complaint and Declaratory Judgment, moving through the intermediate post-judgment orders, and then discussing in detail the proceedings leading up to this case, which were triggered by a new state funding statute and resulted in an eight-day evidentiary hearing. A. The EEOA and the Complaint Nogales is a small city along the Mexican border in southern Arizona. In 1992, when this suit began, the population of the Nogales Unified School District (“NUSD”) was almost entirely Hispanic; that is still so. Most of its approximately 6,000 students come from homes where Spanish is the first language. These students are distributed among six elementary schools, two middle schools, one high school, and one alternative high school. The great majority of the district’s students are classified as English language learners (“ELL”) for at least some portion of their academic careers. In 2006, for instance, thirty percent of the students were in ELL programs and an additional sixty percent had been in such programs. Although Arizona’s theory of ELL instruction has changed over the years, the enormous importance of such programs to students and parents in Nogales has not. ELL students and parents in Nogales (we refer to them as “Flores,” after class representative Miriam Flores), were faced with serious inadequacies in ELL instruction and sued to correct them. The suit proceeded as a class action, with the class defined as “all minority ‘at risk’ and limited English proficient children now or hereafter, enrolled in Nogales Unified School District ..., as well as their parents and guardians.” Flores’ second amended complaint, filed November 29, 1996, primarily alleged that the “State has failed to provide financial and other resources necessary for adequate implementation of mandatory [ELL] programs by public school districts in Arizona,” because “[t]he cost of [ELL] instruction complying with federally prescribed state mandates far exceeds the only financial assistance the State theoretically provides school districts for such purposes.” As a result, Flores contended, Arizona, the state Superintendent, and the state Board of Education violated the EEOA. ' The relevant portion of the EEOA provides: No State shall deny equal educational opportunity to an individual on account of his or her race, color, sex, or national origin, by— ... (f) the failure by an educational agency to take appropriate action to overcome language barriers that impede equal participation by its students in its instructional programs. 20 U.S.C. § 1703. This provision of the EEOA was intended to remedy the linguistic discrimination identified by Lau v. Nichols, 414 U.S. 563, 94 S.Ct. 786, 39 L.Ed.2d 1 (1974), in which the Supreme Court held that failing to provide for the needs of non-English speaking students “is to make a mockery of public education,” rendering classroom experiences for these children “wholly incomprehensible and in no way meaningful.” Id. at 566, 94 S.Ct. 786; see also Castaneda v. Pickard, 648 F.2d 989, 1008 (5th Cir.1981) (noting that the EEOA codifies the “essential holding of Lau, ie., that schools are not free to ignore the need of limited English speaking children for language assistance to enable them to participate in the instructional program of the district.”). Flores alleged that such needs were not being met in Arizona. She charged Arizona with “administering] a school finance scheme that is just sufficient to let less distressed, predominantly Anglo districts impart State-mandated essential skills to their mainstream student bodies ... but that does not and will not enable NUSD or similarly situated districts to impart the same State-mandated essential skills to decisively minority enrollments requiring expanded compensatory programs, smaller class- sizes and further efforts of like nature in order to acquire them.” Flores’ complaint was premised on the EEOA analytic framework provided by the Fifth Circuit in Castaneda. See 648 F.2d at 1009-10; see also Gomez v. Illinois State Bd. of Educ., 811 F.2d 1030, 1041-42 (7th Cir.1987) (applying the Castaneda analysis). The Castaneda framework is three-fold:' First, courts must be satisfied that the “school system is purs[uing] a program informed by an educational theory recognized as sound by some experts in the field or, at least, deemed a legitimate experimental strategy.” Castaneda, 648 F.2d at 1009. Second, “the programs and practices actually used by a school system [must be] reasonably calculated to implement effectively the educational theory adopted 'by the school.” Id. at 1010. There must, in other words, be sufficient “practices, resources and personnel ... to transform the theory into reality.” Id. Third, even if theory is sound and resources are adequate, the program must be borne out by practical results. Id. Flores alleged, consistent with Castaneda step two, that Arizona had “failed to provide financial and other resources necessary for adequate implementation” of its ELL programs. B. The Declaratory Judgment and Arizona’s School Funding System After lengthy pre-trial proceedings and a bench trial, the district court on January 24, 2000, held that Arizona was in violation of the EEOA and granted declaratory judgment in Flores’ favor. See Flores II, 172 F.Supp.2d at 1239. Of the many issues raised in Flores’ complaint, only one EEOA issue was decided by the court: “[W]hether or not Defendants’ [sic] adequately fund and oversee the Lau program in NUSD.... ” Id. at 1226. The rest of the EEOA violations originally alleged, including failures adequately to evaluate and monitor ELL students, to provide tutoring and other forms of compensatory instruction, and to design successful ELL programming, were covered by a consent decree approved by the district court on July 31, 2000. See id. The particulars of Arizona’s school funding system are therefore at the heart of this case. We pause to describe them here. Consistent with Article XI, § 1 of the Arizona Constitution, school funding in Arizona is designed to be essentially equal across districts, despite differences in local property values. See Roosevelt Elementary Sch. Dist. No. 66 v. Bishop, 179 Ariz. 233, 877 P.2d 806, 811-16 (“Roosevelt I”) (Ariz.1994). To ensure equality, the state calculates the maximum support level a district may spend, referred to as the revenue control limit, see ARIz.Rev.Stat. § 15-947, and then pays the difference between that level and the amount the district can raise by levying a state-mandated tax rate against property in its area, see Ariz.Rev.Stat. § 15-971. Thus, regardless of the local tax base, the amount each district can spend per pupil is roughly equalized statewide, by first setting a maximum spending level and then making up the gap between local tax revenues and that level with state funds. The revenue control limit, and hence the district support level, is calculated by adding together funds designated for transporting students to school, see Ariz.Rev. Stat. § 15-945, and “base support” funds, see Ariz.Rev.Stat. §§ 15-943, 944. It is these base funds with which we are primarily concerned. These funds are unrestricted — school districts may spend them essentially as needed — and do not appear to have been calibrated “to any minimum amount necessary for a basic education.” See Roosevelt I, 877 P.2d at 810. Nonetheless, they represent the core funding provided for such an education. Funding allocations are made by using a weighting system: A per pupil weighted amount of funding is calculated by multiplying a “support level weight” (which varies by district type and size and by the student’s grade level) by a statewide baseline amount. See Ariz.Rev.Stat. §§ 15-943(1)(a)-(2)(a) (weights), 15-901(B)(2) (baseline). For the 2006-2007 school year, for instance, the statewide baseline amount was $3,133.53. Ariz.Rev.Stat. § 15-901(B)(2)(a). The support level weight for a typical student in grade eight in most school districts is 1.158, Ariz.Rev.Stat. § 15-943(2)(a), so a district could spend $3,628.63 to support that student in 2006-2007. Not all students are “typical,” of course. Some have special needs which impose additional, incremental costs. ELL students, for example, naturally require additional support. Arizona’s funding formula provides for these incremental costs through an additional “Group B” weighting system which adds additional funding for various student groups, including ELL students. See Ariz.Rev.Stat. § 15-943(2)(b). In 2006-07, for instance, the Group B weight for ELL students was 0.115, so a district received an additional $360.36 to cover the ELL component of that student’s education, on top of the base level $3,628.63 provided for each pupil. The core assumption of Arizona’s funding formula, then, is that ELL students impose incremental costs on a district, above the base level funding allocation. So, while all of the funding (base level and Group B weights) is allocated to the district as a block grant, a district that spends more on ELL incremental costs than allocated is necessarily spending less on basic educational needs per pupil. Taking this framework as a given, the district court in its 2000 declaratory judgment opinion inquired whether Arizona’s funding specifically for ELL students — the Group B weights — actually covered the incremental costs of ELL programming. The court found that it did not. Flores II, 172 F.Supp.2d at 1239. This finding was well supported. Arizona had conducted a cost study in 1987-88 which determined that school districts were spending about $450 (in 1988 dollars, unadjusted for inflation) per ELL student. Id. at 1228. This study was seriously methodologically flawed; it did not, for instance, determine what districts should be spending. But even by the study’s flawed measure, Arizona’s ELL funding was inadequate on a statewide basis. Twelve years after the study, the district court found, Arizona’s ELL Group B weight, 0.60, provided only about $150 per student. Id. at 1238-39. That this support was inadequate to live up to Arizona’s EEOA obligations to its school districts and their students was supported by several examples of resource-linked ELL program deficiencies in NUSD, including: 1) too many students in a class room, 2) not enough class rooms, 3) not enough qualified teachers, including teachers to teach [English as a Second Language] and bilingual teachers to teach content area studies, 4) not enough teacher aids, 5) an inadequate tutoring program, and 6) insufficient teaching materials for both [English as a Second Language] and content area courses. Id. at 1239. In short, Arizona’s “minimum base level for funding Lau programs [was] arbitrary and capricious and [bore] no relation to the actual funding needed to ensure that [ELL] students in NUSD are achieving mastery of its specified ‘essential skills.’ ” Id. In particular, the court held, the ELL Group B weight appropriation was “not reasonably calculated to effectively implement” the ELL programs, and Arizona had, therefore, “failed to follow through with ... resources ... necessary to transform theory into reality,” as the Castaneda framework requires. Id.; see Castaneda, 648 F.2d at 1010. C. Post-Judgment Relief and Arizona’s ELL Programs The judgment was not appealed. Nonetheless, Arizona did not take action to eliminate the violations found in the Declaratory Judgment. Instead, the state legislature defeated several attempts to commission an adequate study of ELL costs that would enable it to set appropriate ELL funding levels. In October of 2000, ten months after the issuance of the Declaratory Judgment, the district court ordered Arizona to “prepare a cost study to establish the proper appropriation to effectively implement” ELL programs. Flores v. Arizona, 160 F.Supp.2d 1043, 1047 (“Flores III”) (D.Ariz.2000). Such a study was critical, the court indicated, because “as a matter of law the State’s minimum base level for funding La% programs bears no relation to the actual funding needed to ensure that [ELL] students are achieving mastery of the State’s specified ‘essential skills.’ ” Id. at 1044. This injunction was not appealed either. Shortly after the injunction issued, in November 2000, Arizona voters approved two propositions that altered Arizona’s school funding and ELL programs. Proposition 203 largely abolished bilingual education, replacing it with sheltered English immersion, a teaching method “in which nearly all classroom instruction is in En-glish but with the curriculum and presentation designed for children who are learning the language.” Ariz.Rev.Stat. § 15-751(5). Proposition 301 increased school funding generally in Arizona through a sales tax increase. See Ariz.Rev.Stat. § 15-901.01. Prop. 301 funding is available for many purposes but is not directed specifically towards ELL programs. See Ariz.Rev.Stat. § 42-5029(E)(1)-(10). Meanwhile, as ordered, Arizona commissioned a cost study (we refer to it as the “SjobergEvashnek study,” after the consulting group that largely directed it). The final study, however, proved to be a disappointment. It did not estimate the incremental costs of EEOA compliance, but instead reported on what several districts nationwide were spending. The study reported widely variable incremental costs for various ELL programs, ranging from about $185 to $3,000 per ELL student. It also found that NUSD was spending $331.60 in incremental costs per pupil, while noting that the NUSD program might not “fully address the unique and extraordinary needs of [NUSD’s ELL] population.” Some additional guidance came from a study conducted by the staff of several Democratic members of the legislature (the “Staffers’ study”), also released in 2001. Aggregating data from other states and from Arizona, the Staffers’ study estimated $1527 in incremental costs per student would be required for EEOA compliant programs — a number, it noted, which was generally consistent with the experience of other states. Arizona did not act on either study. After Flores again moved for post-judgment relief, the court, on June 25, 2001, restated its holding that “as a matter of law the State’s minimum base level for funding Lau programs” was not rationally linked to “funding needed to ensure that [ELL] students are achieving mastery of the State’s specified ‘essential skills.’ ” Flores v. Arizona, No. CV-92-596, 2001 WL 1028369 at *1 (“Flores IV”) (D.Ariz.2001). It ordered Arizona to provide funding that “shall bear a rational relationship to the actual funding needed” by no later than January 31, 2002. Id. at *2. This injunction, too, was not appealed. The legislature responded by enacting HB 2010 on December 19, 2001. That law increased the ELL Group B weight to 0.115, which generated, at that time, about $339.61 per student, roughly the amount that NUSD was recorded as spending in the SjobergEvashnek study. HB 2010 also provided funds for a more comprehensive cost study than had been done previously. Contending that HB 2010 was not adequate, Flores again moved for post-judgment relief. Initially, the district court agreed with Flores. It held, in an order issued on April 8, 2002, that the $339.61 in base level funding was not adequate because it was based on NUSD’s actual spending (as of 2001), which generated “the very program that this Court’s Declaratory Judgment held to be deficient.” Flores v. Arizona, No. CV-92-596 at 3 (“Flores V”) (D.Ariz. 2002). A few months later, however, the court reconsidered, on Arizona’s motion, and held that HB 2010’s funding levels were adequate “as an interim measure pending further study and review,” including the cost study funded by HB 2010. Flores v. Arizona, No. CV-92-596 at 1-2 (“Flores VI”) (D. Ariz. June 12, 2002). The order was not appealed, and three years passed. Again, the legislature gathered some data and, again, it did not act upon the data it received. As part of HB 2010, the legislature commissioned a detailed cost study from the National Conference of State Legislatures (the “NCSL study”). NCSL submitted a first draft in August 2004 and a second draft in February 2005, upon which we rely. The study used two approaches: A survey sent to Arizona school districts to determine then-current ELL program spending and two expert panels, one of state and one of national experts, convened to estimate the costs of EEOA-compliant ELL programs. The school district survey was hampered by a small response rate — only seven of the sixteen surveyed districts provided data. NCSL noted the problem but analyzed the available data, which showed an average expenditure of $669.35 per pupil in incremental costs. NCSL cautioned that this data was limited and of questionable quality, noting that the cost figure might be “somewhat understated.” Expert estimates of appropriate (rather than actual) spending were significantly higher. The national expert panel recommended a range of spending, based on the degree of students’ need for help and grade level from $1,026 per pupil for low-need students at the high school level to $2,571 for high-need elementary school students. The state expert panel recommended spending $1,785 in per pupil incremental costs in grades K-2 and $1,447 in grades 3-12. Again, Arizona did not act to bring its system into line with either set of cost data. Again, Flores moved for post-judgment relief. On January 28, 2005, the court set an April 30, 2005 deadline for Arizona to “appropriately and constitutionally fund[ ] the state’s ELL programs taking into account the Court’s previous orders.” Flores v. Arizona, No. CV-92-596 at 5 (“Flores VII ”) (D.Ariz.2005). D. The First Contempt Order and HB 2064 Arizona failed to act before the deadline. Flores moved for sanctions. On December 15, 2005, deploring the fact that “[thousands of children who have now been impacted by the State’s inadequate funding of ELL programs had yet to begin school when Plaintiffs filed this case,” the court held that Arizona was in civil contempt. Flores v. Arizona, 405 F.Supp.2d 1112, 1113, 1119 (“Flores VIII”) (D.Ariz.2005). It set a deadline of fifteen days after the start of the 2006 legislative session for compliance with its order, and imposed a schedule of fines that would begin to accrue if Arizona did not act and that were to be distributed to Arizona schools to support ELL students. Id. at 1120-21; see also Order re: Distribution of Fines, No. CV-92-596 (D.Ariz. March 16, 2006). Arizona did not enact compliant legislation and accrued over $20 million in fines. Eventually, in the spring of 2006, Arizona Governor Janet Napolitano allowed HB 2064, the legislature’s effort to create a permanent compliant funding system, to become law without her signature. HB 2064 is consistent with Arizona’s “base costs plus Group B weights” approach to education funding, but includes some additional ELL funds designed to cover incremental costs above Group B weight levels, although there are no guaranteed appropriations for these funds. Along with this funding structure, the statute provides for further statewide standardization of ELL programs, presumably in part to make it possible for the state more easily to monitor and assess the impact and use of allocated funds and the need for additional funds. To drive the standardization process, HB 2064 creates the “Arizona English language learners task force” within the state Department of Education. See Ariz.Rev. Stat. § 15-756.01. The task force is “to develop and adopt research based models of structured English immersion programs for use by school districts and charter schools.” Id. at § 15-756.01(C). Each school district and charter school is to adopt one of the models or request a waiver. See Ariz.Rev.Stat. § 15-756.02. To aid in implementing the models, HB 2064 also creates an “Office of English language acquisition services,” which is to develop guidelines for monitoring ELL students and programs, create teacher training programs, and provide other forms of technical assistance. See Ariz.Rev.Stat. §§ 15-756.07, 15-756.08. Finally, the state Board of Education is charged with developing a structured English immersion “endorsement” program for use in training and certifying Arizona’s teachers. See Ariz.Rev.Stat. § 15-756.09. The models, and ELL education generally, are funded by three mechanisms under HB 2064, representing significant potential increases in ELL funding but coming with significant limitations: First, HB 2064 raises the Group B weight for ELL students from 0.115 to 0.140, which corresponds to an increase from about $340 per student to about $450 per student. See HB 2064 § 6, to be codified at Ariz.Rev.Stat. § 15-943(2)(b). There is, however, a significant cut-back as well: Unlike earlier Group B weight funding, under HB 2064 Group B “funding for the same ELL pupil shall not be provided for more than two fiscal years.” HB 2064 § 6. Second, HB 2064 establishes the “Arizona structured English immersion fund,” which provides supplemental funds above the Group B weights to implement the models. See Ariz.Rev.Stat. § 15-756.04. Those funds come, however, with several restrictions. Like the Group B weights, they are limited to two years of availability for any one student. Ariz.Rev.Stat. § 15-756.01(J). And only “incremental costs” of the programs are funded, which are defined as those “in addition to the normal costs of conducting programs for English proficient students,” id. at § 15-756.01(L), and which do not include “compensatory instruction.” “Compensatory instruction” consists of “programs in addition to normal classroom instruction ... [which] are limited to improving the english proficiency of current english language learners and pupils who were english language learners and who have been reclassified as english proficient within the previous two years.” Ariz. Rev. Stat at § 15—756.11(G). In addition, the amount of money provided by the structured English immersion fund is offset by the following funding sources: 1. All federal title III monies and any other federal monies designated solely for the educational needs of english language learners. 2. The portion of title I and title II A monies determined by the english language learner population as a percentage of the qualified population. 3. The portion of impact aid monies determined by the english language learner population as a percentage of the qualified population. A school district or charter school shall only apply unexpended impact aid monies to en-glish language learner programs after it has applied its impact aid monies for other allowable uses as permitted by state law. 4. The portion of [monies the district may levy to further desegregation efforts pursuant to Aeiz.Rev.Stat.] § 15-910 ... determined by the english language learner population as a percentage of the qualified population. 5. The ELL [Group B] weight.... Ariz.Rev.Stat. § 15—756.01(I)(1)—(5). At the same time, HB 2064 provides that “[m]onies from the fund [are to be used] ... to supplement existing programs for english language learners .... [and] shall not be used to supplant available monies used to pay for the normal costs of conducting programs for english proficient students.” Ariz.Rev.Stat. § 15-756.04(F). Importantly, there is no guaranteed legislative appropriation for the structured English immersion fund. Instead, after school districts and charter schools submit budget requests, the Department of Education is to verify and “collect all ... [the] requests and submit them to the [Arizona] legislature for funding.” Ariz.Rev.Stat. § 15-756.03(0). The legislature is not required to appropriate any monies to fund such requests. Third, HB 2064 establishes the “statewide compensatory instruction fund” to be used for compensatory instruction only. See Ariz.Rev.Stat. § 15-756.11. These funds, too, come with a caveat: They “shall not be used to supplant any federal, state or local monies, including desegregation monies ..., used for compensatory instruction that [had been] budgeted for english language learners as of February 23, 2006.” Id. at § 15-756.11(E). In other words, the state funds provided must not replace any funds already in use for the same purpose. The total amount of funding is also limited. Although the legislature appropriated $10 million for this program for the 2006-07 fiscal year, no further appropriations are required by the statute. HB 2064 § 9. HB 2064 was for the most part immediately effective. School districts and charter schools, however, could not adopt the models until they were developed, which they now have been. Also, monies for the structured English immersion fund do not have to be appropriated and to this point have not been. Most importantly, the provisions associated with the Group B weight increase, require court approval. In this regard, HB 2064 provides: [The Group B weight increase and appropriation to fund it, along with the two-year cut-off] do not become effective unless the United States District [C]ourt for the [District of Arizona in the case of [Flores v. State of Arizona ] issues an order that, by this act, the state has taken appropriate action to establish a program that addresses the orders in the case and, at least on an interim basis, the court will permit this act to be fully implemented to determine whether the resulting ELL plans and available funding to implement the plans bear a rational relationship to the cost of implementing appropriate language acquisition programs. HB 2064 § 15(A). Although Governor Napolitano allowed HB 2064 to become law, she did so without vouching for its compliance with the 2005 court order or with the Declaratory Judgment. To the contrary, she announced that she was “convinced that getting this bill into court now is the most expeditious way ultimately to bring the state into compliance with federal law,” and requested that the Attorney General move the court for expedited consideration of the law. E. The First Relief from Judgment Ruling and the Remand By this point, six years after the Declaratory Judgment, the positions of the parties had shifted, with Arizona and the state Board of Education abandoning their defense of the suit and largely siding with Flores. As a result, when, on March 3, 2006, the state Attorney General moved on behalf of Arizona for the court to consider whether HB 2064 satisfied its order, he was arguing against Arizona’s own law. The Superintendent, however, maintained his position and was joined by the Speaker of the Arizona House of Representatives and the President of the Arizona Senate as intervenors (we refer to them as the “Legislative Intervenors”). Later that March, these parties moved to purge contempt and, in the alternative, for relief from judgment under Rule 60(b)(5). Their joint motion was based largely on HB 2064, which, they contended, “creates a plan for adequate funding of programs for [ELL students],” and brings Arizona into compliance. The district court disagreed, ruling on April 25, 2006, that HB 2064 does not comply with its orders or with the Declaratory Judgment because the “Act does not [bear] any rational relationship to the cost of providing an ELL program ... and it has added new hurdles to the mix.” Flores v. Arizona, No. CV-92-596 at 3 (“Flores IX”) (D.Ariz.2006). The court held the Group B weight increase insufficient, and the two-year cut-off on most funds irrational. Id. at 7-8. The two new funds, moreover, were structured in ways that violated federal law. Id. at 4-7. The Superintendent and Legislative Interve-nors appealed both that order and the December 2005 contempt order. On August 23, 2006, in an unpublished memorandum disposition, this court vacated both orders as well as the obligation to pay fines. Flores v. Rzeslawski, 204 Fed.Appx. 580 ("Flores X”) (9th Cir.2006). The court noted that “the landscape of educational funding has changed significantly” since 2000 and remanded for the district court to hold “an evidentiary hearing ... regarding whether changed circumstances required modification of the original court order or otherwise had a bearing on the appropriate remedy.” Id. at 582. The court made clear that it reached none of the other issues in the case. Id. F. The Evidentiary Hearing On remand, the district court held an eight-day hearing in January 2007. Testimony at the hearing focused on three broad areas: (1) HB 2064 and other changes in Arizona’s ELL programming and funding; (2) conditions in NUSD; and (3) conditions in other school districts statewide. As the evidence adduced in the hearing is at the core of our inquiry today, we summarize it in some detail here. 1. Statewide Changes As is required by the Flores consent decree, Arizona has significantly improved its ELL infrastructure. It has also increased overall school funding and, to a lesser degree, ELL program-specific funding. Nonetheless, Arizona’s 134,000 ELL students continue to lag behind statewide average test results for all students. The new administrative structure created by HB 2064 augments Arizona’s efforts to further improve and standardize ELL programs statewide. The state Department of Education holds regular seminars and training sessions and has developed monitoring protocols for school districts. State-developed proficiency standards for ELL students are used to evaluate programs and to identify problems. Arizona has also made efforts to standardize the testing used to classify children as ELL or English proficient. Perhaps most importantly, after August 31, 2006, all classroom teachers, supervisors, principals, and superintendents are required to obtain an endorsement in ELL teaching methods. See Ariz.Rev.Stat. § 15-756.09; Ariz. Admin. Code R7-2-613(J). Along with these ELL-specific structural changes, Arizona has increased overall school funding. On an inflation-adjusted statewide basis, including all sources of funding, support for education has increased from $3,139 per pupil in 2000 to an estimated $3,570 per pupil in 2006. Adding in all county and local sources, funding has gone from $5,677 per pupil in 2000 to an estimated $6,412 per pupil in 2006. Finally, federal funding has increased. In 2000, the federal government provided an additional $526 per pupil; in 2006, it provided an estimated $953. Arizona’s ELL-specific funding increases are associated with HB 2010 and HB 2064. At the hearing, two important points concerning HB 2064 emerged: First, none of the defense witnesses were able to establish what data, if any, the Group B weights are based upon. The weights do not appear to have been set with regard to any specific program costs, known or estimated. Second, Flores’ expert on federal educational funding statutes, Thomas Fagan, testified that if the federal government concluded that HB 2064 was in violation of federal statutes, it could take significant enforcement actions, including a cut-off of some or all federal education funds. Despite Arizona’s expanded ELL infrastructure and budget, ELL students still achieve below — and often far below — state average passage rates on Arizona’s “AIMS” academic achievement test and fall below the minimum passage rates Arizona must meet to reach the “annual measurable objectives” required by the No Child Left Behind Act of 2001 (“NCLB”), Pub.L. No.107-110, 115 Stat. 1425, as test results from the 2004-05 and 2005-06 school years show. See 20 U.S.C. § 6842 (federal standards). These test results must be viewed with two significant caveats: First, because AIMS testing was not carried out in 2000, we do not know whether the performance of ELL students has improved relative to that time. Second, Arizona has changed the AIMS test itself, including altering the passing score, and has twice changed the system it uses for reclassifying ELL students — once at the beginning of the 2004-OS school year, and again at the start of the 2006-07 school year. Results for those students classified as ELL across those years are therefore not clearly comparable. Due to these limitations, and to a general lack of longitudinal data on individual ELL students, we do not have data that conclusively demonstrates whether ELL programs ultimately succeed — that is, whether children pass through them rapidly and ultimately perform as well as non-ELL students. So, while the test results we next discuss are certainly troubling, it is important to understand the limits of their analytic reach. With these caveats, we turn to the data. While Arizona students generally exceeded NCLB-mandated passage rates in math and reading, passing the AIMS test at rates of between 60% and 70%, ELL students were far behind. For example, among third graders, who pass the exam at a higher rate than older ELL students, only 50% passed the math exam in 2005 and just under 40% passed the reading exam in the same year. The situation grows worse at higher grades — in 2005 only 33% of ELL tenth graders passed math and only 20% passed in 2006. In reading, only 30% passed in 2005 and that number fell to barely more than 10% in 2006. In neither year did ELL reading scores in any grade meet federal standards, and in 2006 ELL students’ math test results for all grades also fell below the federal line at every grade level. Nor do students necessarily leave ELL status rapidly. While all witnesses agreed that some students may swiftly become proficient in English, they also agreed that many will need ELL instruction for more than two years, and that some will still need help after three years of training. In short, despite considerable efforts, and some improvements in outcomes, Arizona, as a state, does not appear to have turned the corner on ELL education performance. 2. Conditions in Nogales The statewide achievement gap between ELL and non-ELL students is also present in NUSD, although the district has made significant strides since 2000. The improvements it has made, however, appear to be due largely to successful management rather than to adequate state funding for ELL programs. The record indicates that with adequate incremental ELL funding, the improvements could have been greater and this gap could have been further narrowed. a. Policy Changes and Test Results Much of the improvement at NUSD stems from the successful management of Kelt Cooper, the district’s Superintendent from 2000 to 2005. His successor, Dr. Guilermo Zamudio, has largely sustained Cooper’s management policies. Despite the efforts of both superintendents, however, ELL students in Nogales continue to face serious challenges. Cooper, who took the helm at about the time the Declaratory Judgment issued, adopted policies that ameliorated or eliminated many of the most glaring inadequacies discussed by the district court at that time. Cooper was able to reduce class sizes by enforcing restrictions on the district’s open enrollment policy, significantly improving student/teacher ratios. In addition to lowering class sizes, Cooper improved teacher quality by changing the district’s policies regarding the hiring of experienced teachers and by refusing to pay unqualified teachers who had been certified on an emergency basis as if they had the proper experience. He also fired many teachers’ aides, many of whom were, he found, largely unqualified. Cooper also worked to institute district-wide student performance monitoring. Additionally, he pioneered a uniform system of textbook and curriculum planning, and he largely eliminated what had been a severe shortage of instructional materials by better accounting for and preserving available materials and by acquiring new materials when needed. Using careful financial management and applying for “all funds available,” Cooper was able to achieve his reforms with limited resources. His budget did, however, rise between two and four percent annually, partly because the district twice passed an override, that is, a county-wide tax measure. Nonetheless, Cooper’s successor, Dr. Zamudio, indicates that there are still significant resource constraints. Despite Cooper’s recruitment efforts, Nogales still must rely on some long-term substitutes, rather than upon permanent teachers. Other teachers have been “emergency-certified” and so have not been trained according to Arizona’s standards. NUSD’s starting base pay of $28,500 per year, which is below the statewide average, makes it difficult to recruit the fully-qualified teachers that NUSD needs. The recruitment challenges mean that Dr. Za-mudio has been unable to reduce student/teacher ratios further, to 15:1, which, both he and a defense expert testified, would significantly enhance English learning success. He would also like to be able to hire trained teacher’s aides, who would be helpful in ELL programs, for the lower grades but has been unable to do so because of resource limitations. The limits of NUSD’s progress, even as it has improved its ELL programs, are apparent in the AIMS test results and reclassification test results introduced at the hearing, which show the same problems that appear in the statewide data. The test results, we caution, have the same limitations that we have discussed at the state level, and cannot be read unequivocally to demonstrate that ELL programs are failing in NUSD because we lack longitudinal data on the performance of student cohorts who began as ELL students and who may or may not have been reclassified. It is fair to say, however, that the indications provided by the test data largely do not point in positive directions. For instance, while it is clear that NUSD has reclassified many students, it is not clear how much of this success can be attributed to genuine academic progress and how much to changes in its classification methodology. During the 1999-2000 school year, there were 5,104 ELL students. The next year saw similar numbers, but between the 2001-02 and the 2003-04 school years, the number of such students hovered around 3800. Then, when Arizona implemented a new testing protocol (which was later concluded to be reclassifying too many students and replaced), the ELL population declined to about 3200 students in the 2004-05 school year and to 2474 the next year. In 2006-07 a new, reputedly more accurate and more difficult to pass test came into use, and Dr. Zamudio testified that he expects ELL student numbers to increase again. Because of these shifts in reclassification methodology, the meaning of “ELL” changed over time. Comparison of AIMS test data for ELL students and non-ELL students over time are therefore not reliable. But, as some patterns persist, we can point to several general trends based on AIMS test results for the 2003-04, 2004-05, and 2005-06 school years. First, within NUSD, ELL students in lower grades are doing substantially better than ELL students in higher grades. In 2005-06, for instance, while only 27% of ELL third graders failed math, 76% of ELL tenth graders failed. ELL third graders failed reading 37% of the time; 78% of ELL tenth graders failed. And 35% of ELL third graders failed writing, while 76% of ELL tenth graders failed. Second, ELL students in NUSD generally do worse than the state average score for all students (ELL and non-ELL). Only students in the lower grades sometimes reached or beat the average. Older students are falling far behind. For tenth graders in 2005-06, for instance, only 36% failed math statewide, but 76% of ELL tenth graders in NUSD did. For reading, the tenth grade state failure rate was 29%, while the tenth grade ELL failure rate in NUSD was 78%. And for writing, the state tenth grade failure rate was 36% and the NUSD tenth grade ELL failure rate was 76%. Third, within NUSD, ELL students are still falling behind the district average for all students. Again, the gap grows more pronounced in higher grades. In 2005-06 in third grade, ELL students failed math at a 27% rate, reading at a 37% rate, and writing at a 35% rate, rates that were not too much worse than the district averages of 22%, 29%, and 30%, respectively. By tenth grade, however, ELL students failed math at a 76% rate, reading at a 78% rate, and writing at a 76% rate, far worse than the district averages of 50%, 42%, and 39%, respectively. The picture, then, is of relative success at lower grades (although not equal to that of English speaking students within NUSD), and increasing failures for older students, a significant majority of whom are failing the state’s basic achievement tests. Federal performance classifications under NCLB generally correspond with this portrait: In 2004-05, one of Nogales’ two middle schools failed to meet its adequate yearly progress goals under NCLB, as did the alternative high school. In 2005-06, the same middle school was under a restructuring plan, and corrective action was being taken in the alternative high school. The main high school had been given a warning, as had a lower-performing elementary school. In a survey of the test results of ELL students statewide, NUSD students’ performance was also consistent with this general snapshot. This survey, commissioned by Arizona’s Department of Education, reported the 2005 test results of students who were classified as ELL in 2003 and ranked schools by the average AIMS scores of these students. Some, but not all, of NUSD’s younger ELL students did very well. Four NUSD elementary schools ranked in the top ten schools in the state by this measure, with passage rates of over 70%. Two other elementary schools fared more poorly, ranking 41st and 152nd, with passage rates of about 60% and about 50%, respectively. Middle school performance was lower — one tied for 165th place, also with about a 50% passage rate, and the other ranked 409th, with only about 40% of its ELL students passing the test. The high schools are at the bottom of the heap. The main high school ranks 575th, with a 28.37% overall passage rate and Santa Cruz Alternative High School is four from the bottom at 625th place and a 8.62% passage rate. Finally, there is at least one clear bright spot: For all grades in 2005-06, reclassified ELL students were doing about as well as native English speakers, which is a notable accomplishment, and which suggests that the ELL programs may well benefit those who successfully transition out of them. But, as the data on such scores does not track individual students, showing when they passed through ELL programs and how long it took them to do so, and because the reclassification methodology continues to shift, this bright spot does not offset the otherwise troubling ELL test data. On the data available, it is possible that some high achievers may rapidly be leaving ELL programs while other students continue to struggle, never achieving at the same levels as non-ELL students. Indeed, Dr. Zamudio testified that reclassification in NUSD takes, on average, four to five years and the district court so found. The encouraging success of reclassified students is therefore of limited significance with regard to the overall impact of NUSD’s ELL program. b. Funding NUSD estimates its per pupil incremental costs for ELL programming for the 2005-06 school year at $1570.42, over three times the Group B level weight funding Arizona provides for ELL programs under either HB 2010 or HB 2064. NUSD has been able to finance the improvements it has made by seeking grants, diverting state base level funding away from other purposes, and passing county-level budget overrides. Of the $1570.42 per pupil that NUSD indicates it spends on ELL programming, over half this amount, $824.25, is drawn from the district’s maintenance and operations account (which includes both the ELL Group B funds and base level support funding). As this amount is well above the Group B weights, it indicates that NUSD is dipping into funds that would otherwise be used to fund basic educational purposes. Federal funds add about $484.05 per student, but most of these funds are earmarked for at-risk, low-income students, rather than ELL students (although the two groups overlap). Another $43.43 per student comes from county override funds. The remaining $218.69 per student comes from a variety of state, federal, and local funding sources and grants. NUSD’s budget has increased over the years, but ELL-specific funds have been a relatively small part of that increase. A schematic break-down of the budget over the relevant period, which does not appear to include the relatively small amount of funding from the county, was entered into evidence before the district court. We reproduce a somewhat modified version of that table here. The break-down shows all funds available to ELL students, including funds that are not ELL-specific. ELL-specific funding sources are labeled as such, and the total amount of ELL-specific funding is separately recorded: 1999- 2000- 2001- 2002- 2003- 2004- 2005- 2006-_2000 2001 2002 2003 2004 2005 2006 2007 Base level state funds_ $2,592 $2,618 $2,721 $2,788 $2,858 $2,929 $3,039 $3,173 Group B ELL funds_ $156 $157 $163 $321 $329 $337 $349 $444 Other state ELL funds_ $0 $0 $0 $126 $83 $64 $0 $74 Federal Title I funds_ $439 $448 $467 $449 $487 $638 $603 $597 Federal Title II funds_ $58 $63 $74 $101 $109 $91 $92 $87 Federal Title HI (ELL) funds_ $0 $0 $0 $67 $89 $114 $118 21 Other state and federal grants_ $58 $56 $59 $47 $207 $214 $205 $109 Total of all funds $3,3 02 $3,342 $3,484 $3,899 $4,162 $4,387 $4,406 $4,605 funds_ Total ELL funds $156 $157 $163 $514 $501 $515 $467 $639 In sum, as this table suggests, ELL-specific funds have never covered all ELL programming costs. NUSD can pay the incremental cost of its ELL programs only by diverting non-ELL-specific funds for that purpose, and the district was able to do that at the time of the Declaratory Judgment, as now. These ELL-specific funds, however, have increased, due largely to the rise in Group B weights occasioned by HB 2010 in 2002-03 and will increase again due to HB 2064 if the new Group B weight goes into effect. There remains, however, a fundamental mismatch between the ELL costs NUSD requires and the funds provided for that specific purpose. c. Conditions in Other Districts NUSD’s reported incremental costs for ELL programming, and its need to divert other funds to cover them, are generally consistent with the experiences of districts across Arizona. District officials testifying at the hearing reported incremental costs ranging from $1,077 per student to $4,072 per student. All of these districts are funding their programs by diverting funds from other sources, often by using additional taxes that they are allowed to levy for use in desegregation programs. Group B weights, alone, are not generally sufficient to fund ELL programs. Testimony from officials in other districts confirms that reclassification regularly takes longer than two years for many students. The average reclassification time in the Tucson Unified School District, for instance, is 4.6 years. In the Murphy Elementary School District, nearly two-thirds of ELL students took more than four years to be reclassified. Similarly, in the Scottsdale Unified School District, the majority of ELL students take more than two years to be reclassified, and in the Glendale Union High School District, 46% of ELL students at one school and 15% of ELL students at another had been in ELL programs for more than two years. No district submitted evidence showing that all of its ELL students had been reclassified within two years. G. The Ruling on Remand After the evidentiary hearing at which the facts concerning the current status of ELL students and ELL funding were presented, the district court again denied relief from judgment. Flores v. Arizona, 480 F.Supp.2d 1157, 1167 (‘‘Flores XI”) (D.Ariz.2007). The court held that the improvements at NUSD do not establish that Arizona is fulfilling its duty to fund ELL programming rationally. Id. at 1160, 1166. Most of these improvements, the court found, are due to NUSD’s own management improvements, not to reliable or sufficient funding. Id. at 1160. And the improvements are limited: NUSD’s ELL high school students, in particular, are still falling well behind its non-ELL students, and even the successes will be “fleeting at best” unless Arizona meets its funding obligations. Id. Thus, although NUSD is “doing substantially better,” id., and the state has developed a significantly improved infrastructure for ELL programming, “mere amelioration of those specific conditions” cited as examples of the funding shortage in the Declaratory Judgment, is “inadequate” to justify relief: “Rather, compliance would require a funding system that rationally relates funding available to the actual costs of all elements of ELL instruction.” Id. at 1165. The district court went on to determine that the Superintendent and the Legislative Intervenors had not demonstrated that such a system was in place. The court decided that HB 2064 does not sufficiently address the inadequacies of Arizona’s ELL funding system and, in fact, introduces new problems. In so holding, the district court reasoned as follows: HB 2064’s increase in Group B weights is inadequate, as “the per-student incremental cost of providing ELL instruction is greater than either the current Group B weight of $365 [set by HB 2010] or the increased weight of $444 that would be provided if [the e]ourt approved HB 2064, both in NUSD and in other districts.” Id. at 1162. HB 2064’s two-year cut-off of that funding would suddenly, and irrationally, further underfund school districts. Id. at 1166. The HB 2064 grant programs, which might have linked funding rationally to costs, are no better. The Arizona structured English immersion fund suffers from the same irrational two-year cut-off, id. at 1163, and both it and the compensatory instruction fund violate provisions of federal law that bar taking federal funds into account in making state funding decisions and bar supplanting existing state funding with federal monies. Id. at 1166. The district court concluded that, without a rational funding system for ELL incremental costs, Arizona remains out of compliance with the EEOA, despite some successes in NUSD: On January 24, 2000, this Court held that the State’s minimum funding level for ELL programs was arbitrary and capricious and bore no rational relation to the actual funding needed to insure that ELL students could achieve mastery of the State’s academic standards .... More than 7 years later, circumstances in this regard remain the same. The Moving Parties have not shown compliance with this Court’s decree, much less changed circumstances that would warrant modification or dissolution of this Court’s order. Id. at 1167. The district court gave Arizona until the end of the then-current legislative session to comply. Id. The Legislative Intervenors and the Superintendent timely appealed. II. ANALYSIS A. Standard of Review and Jurisdiction Rule 60(b)(5) provides that “[o]n motion and just terms, the court may relieve a party ... from a final judgment ... [when] the judgment has been satisfied, released or discharged; it is based on an earlier judgment that has been reversed or vacated; or applying it prospectively is no longer equitable.” The first two circumstances do not obtain here: Although the moving parties argue that Arizona has achieved the purposes of the judgment through actions other than those required by the Declaratory Judgment and the ensuing orders, Arizona has in fact not “satisfied” the prior judgment, as it has not complied with its terms. And there is no pertinent earlier judgment that has been reversed or vacated. Instead, the moving parties are, in practical terms, contending that it “is no longer equitable” to apply the judgment and the present injunction prospectively. It is appropriate to grant a Rule 60(b)(5) motion for this reason when “the party seeking relief from an injunction or consent decree [meets its initial burden by showing] ‘a significant change either in factual conditions or in law.’ ” Agostini v. Felton, 521 U.S. 203, 215, 117 S.Ct. 1997, 138 L.Ed.2d 391 (1997) (quoting Rufo v. Inmates of Suffolk County Jail, 502 U.S. 367, 384, 112 S.Ct. 748, 116 L.Ed.2d 867 (1992)); see also Bellevue Manor Assoc. v. United States, 165 F.3d 1249, 1255 (9th Cir.1999) (“Rufo sets forth a general, flexible standard for all petitions brought under the equity provision of Rule 60(b)(5).”). Once that showing is made, modification may be warranted “when changed factual conditions make compliance with [a consent decree or an injunction] substantially more onerous,” when a decree or injunction “proves to be unworkable because of unforeseen obstacles[,] or when enforcement of the decree without modification would be detrimental to the public interest.” Rufo, 502 U.S. at 384, 112 S.Ct. 748 (citations omitted); see also SEC v. Coldicutt, 258 F.3d 939, 941-42 (9th Cir.2001) (using this standard). Finally, modification is necessary if a consent decree or injunction “become[s] impermissible under federal law.” Rufo, 502 U.S. at 388, 112 S.Ct. 748; see also Agostini, 521 U.S. at 215, 117 S.Ct. 1997 (same). We review district courts’ rulings on motions for relief from judgment under Rule 60(b) for abuse of discretion. United States v. Asarco Inc., 430 F.3d 972, 978 (9th Cir.2005). “We may not reverse a district court’s exercise of its discretion unless we have a definite and firm conviction that the district court committed a clear error of judgment in the conclusion it reached upon weighing the relevant factors.” Coldicutt, 258 F.3d at 941. In this context, “[a] district court abuses its discretion if it does not apply the correct law or if it rests its decision on a clearly erroneous finding of material fact.” Asarco, 430 F.3d at 978. We emphasize that our task in this context is a limited one: We do not sit to retry the case, nor to reexamine unappealed legal determinations, and do not have jurisdiction to do so. “[T]he timely filing of a notice of appeal in a civil case is a jurisdictional requirement,” Bowles v. Russell, — U.S. -, 127 S.Ct. 2360, 2366, 168 L.Ed.2d 96 (2007), and “[because the time for appeal ha[s] passed in this case, the interest in finality must be given great weight.” Ashford v. Steuart, 657 F.2d 1053, 1055 (9th Cir.1981). We have cautioned against the use of provisions of Rule 60(b) to circumvent the strong public interest in timeliness and finality served by Rule 4(a). See In re Stein, 197 F.3d 421, 425-26 (9th Cir.1999). Simply put, motions for relief from judgment may not be used to remedy a failure to contest in the first instance the legal rulings underlying the judgment itself. Our review is somewhat closer in the context of institutional injunctions against states “due to federalism concerns” implicated by such injunctions. Hook v. Arizona Dep’t of Corr., 107 F.3d 1397, 1402 (9th Cir.1997). In such contexts, “[w]e scrutinize the injunction closely to make sure that the remedy protects the plaintiffs’ federal constitutional and statutory rights but does not require more of state officials than is necessary to assure their compliance with federal law.” Clark v. Coye, 60 F.3d 600, 604 (9th Cir.1995); see also Hook, 107 F.3d at 1402. Nonetheless, “we will defer to the district court so long as any injunctive relief it provides remains within these parameters,” Clark, 60 F.3d at 604, and the burden of proof remains on the moving party. Hook, 107 F.3d at 1402; Rufo, 502 U.S. at 383-84, 112 S.Ct. 748. Additionally, federalism concerns are substantially lessened here, as the state of Arizona and the state Board of Education wish the injunction to remain in place. The Superintendent and Legislative Intervenors also question the district court’s compliance with our mandate on remand and the propriety of several evidentiary rulings. “This court reviews de novo a district court’s compliance with the mandate of an appellate court.” United States v. Perez, 475 F.3d 1110, 1112 (9th Cir.2007) (citations and alterations omitted). With exceptions not pertinent here, “we generally review the district court’s ruling on a motion in limine only for an abuse of discretion.” United States v. Ross, 206 F.3d 896, 898 (9th Cir.2000). Finally, the parties dispute each others’ standing. The Legislative Intervenors, first, contend that Flores lacks standing to argue that HB 2064 does not represent a change in circumstances sufficient to warrant relief. The Legislative Intervenors are wrong: HB 2064 holds itself out as a remedy for Flores and was presented to the court by Arizona and by the Superintendent and Legislative Inter-venors as such. Whether or not Flores would have had standing to challenge HB 2064 in the first instance, she certainly has standing to argue that a purported remedy will not satisfy a judgment in her favor. Second, Flores challenges the standing of the Superintendent because he does not make the funding decisions at the core of this case. As to that basis for disputing the Superintendent’s standing, Flores is wrong. The Superintendent does not, it is true, write school funding laws, but he does administer the state’s education system, including the HB 2064 grant system, see Ariz.Rev.Stat. § 15-251, and obviously has a vital interest in whether or not funding for the programs he oversees is EEOA compliant. He is a named defendant in the case; the Declaratory Judgment held him to be in violation of the EEOA, Flores II, 172 F.Supp.2d at 1239-40, and the current injunction runs against him. That judgment against him may not be essential to relief does not mean that he lacks standing to contest an order running against him. The Superintendent’s standing is, however, for a different reason, fairly tenuous. He is “[s]ubjeet to supervision by the state board of education” with regard to distributing monies to local schools, and must “[e]xecute, under the direction of the state board of education, the policies which have been decided upon by the state [B]oard.” Ariz.Rev.Stat