This is the latest appeal growing out of the nearly two-decades old, racial discrimination in employment lawsuit involving the Alabama Department of Transportation (ALDOT) and the State Personnel Department (SPD). Those two state agencies were sued in 1985 by what became two plaintiff classes of black employees and prospective employees. A partial settlement was reached and a consent decree was entered in 1994, but instead of ending the case the decree became a platform for additional litigation.

A more detailed history of the case can be found in our six other published opinions involving it. See Reynolds v. McInnes, 338 F.3d 1201, No. 98-6164, 2003 WL 21692337 (11th Cir. July 22, 2003) ("Reynolds V"); Reynolds v. Butts, 312 F.3d 1247 (11th Cir.2002) ("Reynolds IV"); Davis v. Butts, 290 F.3d 1297 (11th Cir.2002); Reynolds v. Roberts, 251 F.3d 1350 (11th Cir.2001) ("Reynolds III"), cert. denied, 534 U.S. 1161, 122 S. Ct. 1171, 152 L. Ed. 2d 115 (2002); Reynolds v. Roberts, 207 F.3d 1288 (11th Cir.2000) ("Reynolds II"), cert. denied, 533 U.S. 941, 121 S. Ct. 2576, 150 L. Ed. 2d 739 (2001); Reynolds v. Roberts, 202 F.3d 1303 (11th Cir.2000) ("Reynolds I"). We won't repeat any more of that history than is necessary for an understanding of the issue before us in this appeal.

The issue is whether the district court abused its discretion when it modified one part of one article of the 1994 consent decree at the defendants' request and over the objection of the plaintiffs. Insisting that it did, the plaintiffs have appealed. Concluding that it did not, we affirm.

The consent decree, aimed at ending racial discrimination in ALDOT's employment practices, has twenty-one articles. This appeal is about Article Two, which governs the development and use of "minimum qualifications" (MQs), which are part of the selection procedure for hiring and promoting employees in ALDOT jobs. A job seeker wanting to sit for an employment examination must meet the MQs first. MQs are designed to screen for skills needed at entry into a new position, and can screen for, among other things, "knowledge, skills and abilities" (KSAs) relevant to a position. The job examinations themselves measure KSAs.

The provision of Article Two that was modified by the district court is ¶ 1, which is called "the no-overlap provision." This is what the paragraph says:

Minimum qualifications will not be utilized on examination announcements or to preclude an applicant from examination unless the minimum qualification bears a manifest relationship to skills, knowledges, or abilities necessary to the performance of the job at entry without a brief orientation period and such skills, knowledges or abilities are not addressed in the examination process.

The key point for our purposes is that the "skills, knowledges, or abilities" to which the MQs must bear a manifest relationship cannot be ones "addressed in the examination process." Before getting into the reasons that the no-overlap provision is problematic, we need to explain some more about the job selection procedures process that is required by the decree.

Other portions of Article Two incorporate the federal Uniform Guidelines on Employee Selection Procedures, 29 C.F.R. §§ 1607.1-1607.18 ("Uniform Guidelines"), and require SPD to meet the Uniform Guidelines in developing MQs. Article Two, ¶ 3 allows for the use of MQs that have been determined to be "valid" within the meaning of the Uniform Guidelines. To establish that a job selection procedure is "valid" an employer must demonstrate that the selection procedure is appropriately job-related. Under the Uniform Guidelines, a job selection procedure may be validated through the use of one of three different types of studies: content validity, criterion-related validity, and construct validity. Id. § 1607.5(A).

Of those three types of validation methods, the parties chose content validation for MQs. They wrote that choice into Article Two, ¶ 2, which specifies use of "a content validation procedure to determine the appropriate minimum qualifications" for ALDOT jobs. A content validity study measures whether and to what extent the selection procedure is "representative of important aspects of performance on the job for which the candidates are to be evaluated." Id. § 1607.5(B). It measures knowledge, skills, or abilities that are "necessary prerequisite[s]" for the "performance of critical or important work behavior(s)" for the job. Id. § 1607.14(C)(4). The Guidelines provide that a selection procedure may be used "if it represents a critical work behavior (i.e., a behavior which is necessary for performance of the job) or work behaviors which constitute most of the important parts of the job." 29 C.F.R. § 1607.14(C)(8). Under a "criticality" approach to content validation, an approach the plaintiffs advance as one that would be valid under the Uniform Guidelines and comply with the decree, MQs could be based on a small number of "critical" KSAs instead of most of the KSAs.

In order to construct content valid MQs and a valid examination under the Uniform Guidelines, a detailed job analysis must be performed for each job classification and must focus on the work behaviors necessary for successful performance of the job and also on the tasks associated with those behaviors. Id. § 1607.14(C)(2). Each job analysis conducted by SPD for ALDOT included interviews with incumbents in the jobs or their supervisors, upon whom were bestowed the laudatory label "subject matter experts." Article Two, ¶ 2 also provides that the MQs and the validation procedure "will be subject to challenge by plaintiffs and no new minimum qualifications will be implemented without approval by the plaintiffs or the Court."

Having finished our description of the job selection procedures in general, we focus again on the Article Two, ¶ 1, the no-overlap provision. It is important to say at the outset what all the experts for both sides agreed upon at the hearing in this case: the no-overlap provision is novel in the field of employment testing. There has been a whole lot of employment testing in this country over the years, but there is nothing in the record indicating that the no-overlap provision, variously described as "novel" and "unusual," has ever been used anywhere before. Nonetheless, for several years after the consent decree was adopted, a joint panel of experts — two selected by the plaintiffs and two selected by the defendants — worked to develop MQs and job examinations for ALDOT jobs that would comply with the no-overlap provision.

The plaintiffs' and defendants' experts initially agreed to develop "task-based" MQs, which would evaluate a job candidate's experience in performing the discrete tasks required at entry into a new position. They reasoned that measuring experience with certain tasks, instead of typical KSAs, would provide the most valid MQs while avoiding overlap between the KSAs used in the MQs and those measured by the examinations. However, problems resulted from use of the task-based MQs, including a higher rejection rate of applicants than expected and a high rate of adverse impact, meaning that the pass rate of black applicants was significantly lower than that of white applicants.

SPD then tried a different approach, using another MQs screen as an alternative to the existing one that had not worked. The new MQs screen was a more traditional one which included an educational requirement and an experience requirement. That approach did not work either. None of the parties was satisfied with the resulting MQs that were in place in 2001.

By the summer of 2001, SPD and its experts decided to strike out on their own to develop new MQs for some of the job classifications, beginning with the job of Civil Engineer Manager. The experts used by SPD looked at job analysis reports that had already been developed for the jobs, selected subject matter experts, conducted MQs development sessions (where a group of subject matter experts were brought together to help determine the types of MQs to use), and they looked for some consensus about potential MQs. Then, the experts linked the MQs to the KSAs in order to ensure that they were actually related to each other as required under Article Two, ¶ 1 of the consent decree. Despite all their efforts, the experts were unable to design any method for developing valid MQs that would comply with the no-overlap provision.

On December 13, 2001, the defendants moved to modify the consent decree by removing the no-overlap provision of Article Two, ¶ 1, on the ground that the provision is contrary to accepted professional practice in the development of selection procedures, and it cannot reasonably be satisfied. By that point, the defendants had developed MQs for two jobs — Civil Engineer Manager and Senior Right of Way Specialist — and the job examination for Civil Engineer Manager had been administered to applicants who had gotten past the MQs screen. With the parties' consent, the district court appointed a special master to consider the issue and make recommendations to the court.

After conducting a three-day hearing, the special master made detailed findings. The gist of them is that the defendants had made reasonable, good faith efforts to comply with the no-overlap provision, but those efforts had been unsuccessful because the no-overlap provision "is unworkable across all job classification[s]." The special master recommended to the district court that it find the defendants had satisfied the requirements for modification of a consent decree as set out by the Supreme Court in Rufo v. Inmates of Suffolk County Jail, 502 U.S. 367, 112 S. Ct. 748, 116 L. Ed. 2d 867 (1992).

The district court adopted the special master's recommendation and granted the defendants' motion in part. The court ruled that "the defendants may develop and use minimum qualifications (MQs) for those classifications for which examinations have already been developed." The effect of the ruling is to modify the consent decree by making the no-overlap provision of Article Two, ¶ 1 inapplicable to the fifteen job classifications for which the defendants had developed examinations at the time of the ruling. The district court explained that the defendants had made a good faith effort to comply with the provision without success, and "it is simply time to proceed with the development of the MQs without the [no-overlap] provision."

With respect to the other job classifications, the ones for which the defendants had not developed examinations at the time of the ruling, the district court referred the matter back to the special master. It directed him to determine whether for those remaining classifications it is still possible to comply with the no-overlap provision without unreasonable delay in the development and administering of the examinations. As recently as the oral argument of this case in June of this year, those proceedings were still ongoing before the special master.

For modification purposes, a consent decree is not treated as a contract, but as a judicial act akin to an injunction. Rufo, 502 U.S. at 378, 112 S. Ct. at 757; United States v. Swift & Co., 286 U.S. 106, 114-15, 52 S. Ct. 460, 462, 76 L. Ed. 999 (1932); Jacksonville Branch, NAACP v. Duval County Sch. Bd., 978 F.2d 1574, 1578 (11th Cir.1992). Partly because of that, we review a district court's decision granting a party's request for modification of a consent decree only for an abuse of discretion. Sierra Club v. Meiburg, 296 F.3d 1021, 1032 (11th Cir.2002); Jacksonville Branch, 978 F.2d at 1578. Findings of fact relating to the modification are reviewed only for clear error. Jacksonville Branch, 978 F.2d at 1578.

Rule 60(b) of the Federal Rules of Civil Procedure allows a district court to modify a consent decree when "it is no longer equitable that the judgment should have prospective application." Fed. R. Civ. Pro. 60(b)(5); Rufo, 502 U.S. at 378-79, 112 S. Ct. at 757. The standard for modification of a consent decree in an institutional reform case is a flexible one, because flexibility "is often essential to achieving the goals" of institutional reform litigation. Rufo, 502 U.S. at 379, 381, 112 S. Ct. at 757-58, 758. Flexibility in modifying decrees like this one is in the public interest, because "such decrees reach beyond the parties involved directly in the suit and impact on the public's right to the sound and efficient operation of its institutions." Id. at 381, 112 S. Ct. at 758-59 (internal marks and citation omitted). The need for modification of a consent decree and the importance of flexibility in regard to it can be especially apparent in an institutional reform case like this one where efforts to implement the decree have been bogged down for years, see generally Reynolds v. McInnes, 338 F.3d 1201, No. 98-6164, 2003 WL 21692337 (11th Cir. July 22, 2003) ("Reynolds V").

"[A] court faced with a motion to modify a consent decree in institutional reform litigation must begin by determining the `basic purpose' of the decree." United States v. City of Miami, 2 F.3d 1497, 1504 (11th Cir.1993). If the provision that a party seeks to modify "is central to the decree, or ... `the most important element' of the decree, then the modification is likely to violate the basic purpose of the decree and, therefore, will be forbidden." Id. at 1504-05. However, if the provision for which modification is sought "merely sets out one of several means of accomplishing the purpose of the decree or one of several means of measuring compliance with the decree's objective, then the requested modification is not necessarily prohibited." Id. at 1505.

If modification is permissible under those criteria, the court must decide whether to exercise its discretion to modify the decree. The Supreme Court has articulated a two-part test for determining when a consent decree should be modified. First, "a party seeking modification of a consent decree bears the burden of establishing that a significant change in circumstances warrants revision of the decree." Rufo, 502 U.S. at 383, 112 S. Ct. at 760. A party "may meet its initial burden by showing ... a significant change either in factual conditions or in law." Id. at 384, 112 S. Ct. at 760. Second, the proposed modifications must be "suitably tailored" to address the new factual or legal environment. Id. The plaintiffs do not contest that the second Rufo requirement has been met, only the first one.

Rufo's first requirement, that there be changed factual or legal circumstances, is interpreted flexibly, and different sorts of factual changes may qualify as changed circumstances permitting modification. Ensley Branch, NAACP v. Seibels, 31 F.3d 1548, 1563 (11th Cir.1994). The Rufo Court observed that modification based on factual change may be appropriate in any of at least three situations: (1) "when changed factual conditions [have made] compliance with the decree substantially more onerous"; (2) "when a decree proves to be unworkable because of unforeseen obstacles"; and (3) "when enforcement of the decree without modification would be detrimental to the public interest." 502 U.S. at 384-85, 112 S.Ct. at 760 (citations omitted).

We have explained that modification also may be appropriate when "significant time has passed and the objectives of the original agreement have not been met" despite the defendants' efforts, or when a continuation of the decree would be inequitable. Jacksonville Branch, 978 F.2d at 1582 (citing Rufo, 502 U.S. at 384, 112 S. Ct. at 760; Newman v. Graddick, 740 F.2d 1513, 1520 (11th Cir.1984)). See also Heath v. DeCourcy, 992 F.2d 630, 634 (6th Cir.1993) (Institutional reform consent decrees "are subject to a lesser standard of modification, requiring the lower court to identify a defect or deficiency in its original decree which impedes achieving its goal, either because experience has proven it less effective, disadvantageous, or because circumstances and conditions have changed which warrant fine-tuning the decree.") (internal quotations and citation omitted). We have also said that "[a] district court has broad discretion to modify an existing injunctive order when factual circumstances have changed or new ones have arisen since the order was issued, as long as notice and an opportunity to be heard are provided before the modification is made." Riccard v. Prudential Ins. Co., 307 F.3d 1277, 1298 (11th Cir.2002).

With those principles in mind, we turn now to the plaintiffs' contentions in this case. The first one is that modification of the no-overlap provision was impermissible because that provision is one of the consent decree's "basic purpose[s]." City of Miami, 2 F.3d at 1504-05. The short answer is that it is not.

The consent decree itself provides that its "intent and purpose" is "to undo the effects of the past practices which have been the subject of this case and [d]ecree and to prevent further practices which may perpetuate such efforts or otherwise discriminate against the plaintiffs or the class they represent." The decree accomplishes this purpose through twenty-one articles that require changes in the way the defendants hire, promote, classify, and pay ALDOT employees. Reynolds II, 207 F.3d at 1293. The no-overlap provision is only one part of one of those twenty-one articles, and it is by no means central to the decree.

The no-overlap provision is not the only provision that addresses the development of non-discriminatory MQs. It is part of Article Two, which also provides for non-discriminatory MQs in two other ways: by requiring (also in ¶ 1) the MQs to bear a "manifest relationship" to the KSAs necessary to perform a particular job at entry, and by requiring (in ¶¶ 2 and 3) the procedure for determining MQs be valid within the meaning of the Uniform Guidelines. Not only that, but beyond Article Two there are other provisions in the decree that seek to prevent discriminatory procedures at the hiring stage. For example, Article Three, ¶ 4(a) requires that SPD use only selection criteria and procedures that have been validated in accordance with the Uniform Guidelines. And Article Eight, ¶ 2 requires ALDOT to develop and use non-discriminatory interview questions.

The plaintiffs contend that because the no-overlap provision is the only provision aimed at eliminating the overlap of KSAs between MQs and examinations, it is the only provision that will save from early elimination some members of the plaintiff class. But so what? It seems to us that even if the provision saves some applicants from early elimination by removal of the overlapping KSAs, those applicants will still be eliminated by the measurement of the same KSAs in the job examination. Even if we are mistaken about that, the plaintiffs' argument fails anyway. Just because a provision of a consent decree is the only one that addresses a narrow and particular concern of the parties does not mean that the provision can never be modified. If that were true, the only parts of a consent decree that could be modified are those that are redundant, and they could be modified only to remove the redundancy, which would not change the operation or effect of the decree at all.

In summary, the no-overlap provision is not "central to the decree, or ... `the most important element' of the decree," and modification of it is not "likely to violate the basic purpose of the decree." City of Miami, 2 F.3d at 1504. It is not even the most important part of Article Two of the decree. The provision is but "one of several means of accomplishing the purpose of the decree," and is, in fact, but one of several means of ensuring the development of non-discriminatory MQs. Id. at 1505. That means the requested modification is not necessarily prohibited. Id. So, we turn now to whether the defendants showed a significant change in factual conditions, as required by Rufo. 502 U.S. at 383, 112 S. Ct. at 760.

The plaintiffs' second contention is that, even if modification of the no-overlap provision is not prohibited outright, the defendants did not prove changed circumstances as required by Rufo, 502 U.S. at 383-84, 112 S. Ct. at 760. At the modification hearing, the defendants did not submit evidence of any circumstances that specifically motivated inclusion of the no-overlap provision in the consent decree when it was adopted in 1994. The plaintiffs argue that without evidence of the specific circumstances that gave rise to the provision, there can be no determination that those circumstances have changed.1 The plaintiffs contend that evidence of the circumstances as they existed at the time the parties agreed to the decree is always required under Rufo in order to determine if the circumstances have changed. Given the array of circumstances which both the Supreme Court and this Court have said may satisfy the Rufo changed circumstances requirement, such a narrow reading of that requirement is neither necessary nor grounded in case law. See Rufo, 502 U.S. at 384-85, 112 S. Ct. at 760; Ensley Branch, 31 F.3d at 1563; Jacksonville Branch, 978 F.2d at 1582; Newman, 740 F.2d at 1520.

When a party seeks modification of a consent decree provision on the ground that the provision is unworkable because of unforeseen obstacles and cannot reasonably be satisfied, Rufo, 502 U.S. at 384, 112 S. Ct. at 760, it is not necessary that the court have before it evidence of the precise circumstances that existed when the parties agreed to the provision. It is enough that when they agreed to the provision in question the parties thought it would be workable — which we hope will always be the case — but good faith effort and hard-earned experience has proven that it is not.2 In other words, the fact that a provision of a consent decree has proven to be unworkable is itself a "significant change in circumstances." Id. at 383, 112 S. Ct. at 760. Remember that the standard for modification of consent decrees in institutional reform litigation is, and should be, a flexible one. Id. at 379, 381, 112 S. Ct. at 757-58, 758

Turning to the evidence in this case, the defendants presented evidence during the special master's hearing that they attempted in good faith to implement the no-overlap provision but that it is unworkable. The defendants showed that their experts collaborated with the plaintiffs' experts in developing the task-based MQs, a collaboration which demonstrates good faith. Both sides' experts testified that the no-overlap provision was novel or unusual in the field of employment testing, and one of the plaintiffs' own experts testified that he had never seen a provision like this one applied in developing a selection procedure. There was no evidence before the district court that it had ever been used anywhere else. The experts also testified that a joint panel of experts had attempted to develop examinations that would measure as many of the KSAs as possible because the more KSAs an examination measures the more likely it will be found valid under a content validation approach. One of the defendants' experts testifying at the hearing said that it would be "very hard" to measure most of the KSAs in the MQs and then measure most of the KSAs in the examinations and comply with the no-overlap provision of the consent decree. He also testified that he could not envision a way to satisfy both the requirements of content validity and the no-overlap provision and stated that the provision "almost precludes best professional practices."

The plaintiffs also presented evidence at the hearing. Their two experts were of the opinion that the no-overlap provision is workable and does not need modification. They said that even with the no-overlap provision, valid MQs that would pass muster under the Uniform Guidelines could be developed through the use of a "criticality" approach, which the defendants had not attempted. Both of the plaintiffs' experts took issue with the characterization that the task-based MQs project had failed, and said instead that the joint effort was aborted by the defendants before it was finished. But what the plaintiffs' experts did not say stands out. They did not deny that the no-overlap provision prevents important KSAs that are used in the MQs from being used in the examinations, thereby decreasing the likelihood that the examinations would be judged valid. Nor did they explain why if the no-overlap provision is a good idea it has never been used by anyone before.

Examining all the evidence, the special master credited the defendants' experts and found that the defendants' efforts had been reasonable and in good faith, but ultimately unsuccessful. He also found that "[t]he evidence shows that plaintiffs and defendants share responsibility for the choices that have brought about the current situation, and that neither party in fact foresaw that these jointly-made decisions would lead to the defendants' (and intervenors') current conclusion that the no-overlap provision is unworkable." The special master agreed with the defendants that the no-overlap provision is not workable.

After an independent review of the record, the district court found that the defendants had "made a diligent and good-faith effort to comply with the no-overlap provision without success," and concluded that "the defendants are not under an obligation to negate all possibilities, no matter how time-consuming or unending in endeavor or farfetched or impractical, for compliance with this provision." We agree completely. Complex consent decrees in institutional reform cases, like much in life, are a series of experiments, and when part of a decree has been proven unworkable the law does not condemn the institution to the Sisyphean task of endlessly rolling a stone up the hill only to have it roll back down. The consent decree's basic goal of ending racial discrimination in the Alabama Department of Transportation's employment practices can be achieved better and sooner if the Department spends its time and efforts on the provisions of the decree that can be made to work without the expenditure of an unreasonable and disproportionate amount of time and resources.

The plaintiffs' last attack on the district court's order is that it is, in their view, self-contradictory. The court modified the decree by ending the no-overlap provision only as to the fifteen job classifications for which examinations had been developed at the time of the court's order. As to the other job classifications, the ones for which examinations had not been developed yet, the court referred the matter back to the special master with directions for him to make a recommendation on whether it was possible for the defendants to comply without unreasonable delay with the no-overlap provision as to those job classifications. The point of the plaintiffs' argument is that if the no-overlap provision is workable as to the job classifications for which no examination had been developed, it is also workable as to those for which job examinations have been developed.

There is some tension in the district court's order insofar as it distinguishes between the job classifications based upon whether an examination had been developed at the time of the order. That tension exists to the extent (which is considerable) that the problem with the no-overlap provision is an inherent one instead of one that depends on the nature of the job classification or the timing of the examination development process. If the no-over lap provision is unworkable as to the classifications for which examinations already have been developed, how can it be workable as to classifications for which examinations have not yet been developed? There are a couple of answers to that question, neither of which is favorable to the plaintiffs.

The first answer is that the district court, unlike the special master, did not find that the no-overlap provision could never work. Instead, the court said that even if there is still some possibility of developing for the completed job examinations MQs that would comply the no-overlap provision, that possibility is sufficiently small that it would be unreasonable to require that the defendants continue their attempts to do so. Having made a good faith effort, the defendants have done what is reasonable, the court explained, so it is time to move on with the job classifications for which examinations have been developed. The classifications for which no examination have been developed are different, it might be argued, because the defendants have made no effort to develops MQs for them that comply with the no-overlap provision.

The second and more fundamental answer to the question posed by the plaintiffs' argument that even if the district court's order is inconsistent because it treats the two categories of job classifications differently, we are not required to resolve that inconsistency in the plaintiffs' favor any more than we are required to resolve an inconsistent jury verdict in favor of a criminal defendant. Where a defendant logically is either guilty of both counts in an indictment or not guilty of either, but the judge or jury acquits on one and convicts on the other, the defendant is not entitled to have the conviction set aside simply because the verdict is inconsistent. Harris v. Rivera, 454 U.S. 339, 345, 102 S. Ct. 460, 464, 70 L. Ed. 2d 530 (1981) (per curiam); United States v. Schlaen, 300 F.3d 1313, 1317 (11th Cir.2002). The reason is that the verdict inconsistency itself establishes only that either the acquittal or the conviction is inaccurate; it does not tell us which one is the problem. United States v. Powell, 469 U.S. 57, 65, 105 S. Ct. 471, 477, 83 L. Ed. 2d 461 (1984). What we do is examine the conviction and, if it is not invalid on some other ground, we affirm it, which means the acquittal was the problem. See Harris, 454 U.S. at 345, 102 S. Ct. at 464; Schlaen, 300 F.3d at 1317. In other words, just because the criminal defendant got one gift from the jury does not compel us to give him another one.

The same type of reasoning should apply here. Assuming that the district court's order is inconsistent as to the two categories of job classifications, that inconsistency indicates at most only that the district court got it right as to one category and wrong as to the other. We have examined the order insofar as it concerns the category of job classifications in which examinations have already been completed and found it to be correct, or at least not an abuse of discretion. That is all we need decide to dispose of this appeal, because the defendants did not cross-appeal the part of the order referring the other category of job classifications back to the special master.

The district court did not abuse its discretion in modifying the consent decree to remove or suspend the no-overlap provision contained in Article Two, ¶ 1, insofar as it concerns job classifications for which examinations had been completed at the time of the order.

AFFIRMED.