dissenting.
The decision below was based on the extensive evidentiary record that was developed over nine days of trial. The trial judge heard testimony from more than twenty witnesses, including two expert statisticians, and received hundreds of documents into evidence. In stripping Plaintiffs of their victory, the majority ignores this evidentiary record and relies instead on a series of largely unexplained conclusions. Because the district court committed no error of law, and the factual findings underpinning its decision are not clearly erroneous, we should affirm. The majority has no legal basis to do otherwise, and therefore I respectfully dissent.
The majority reverses the district court’s finding of liability against Defendant Metropolitan Government of Nashville and Davidson County, Tennessee (“Metro”) on the basis that “Plaintiffs failed to establish a prima facie case of disparate impact liability.” (Maj. Op. at 738.) To make out a prima facie case of disparate impact in violation of Title VII, a plaintiff must “(1) identif[y] a specific employment practice to be challenged; and (2) through relevant statistical analysis prove[ ] that the challenged practice has an adverse impact on a protected group.” Dunlap v. Tenn. Valley Auth., 519 F.3d 626 (6th Cir.2008) (citations omitted); see also Lewis v. City of Chicago, — U.S. -, 130 S.Ct. 2191, 2197, 176 L.Ed.2d 967 (2010). The district court, sitting as finder of fact, found that Plaintiffs had satisfied their prima facie burden, and as explained below, its conclusion should be upheld as to each element of the prima facie case.
A. Identification of Specific Employment Practices
The district court did not err in finding that Plaintiffs’ burden of identifying the specific employment practices that are *743challenged “is established by a preponderance of the evidence,” through proof of the following: “tailoring job qualifications for promotions, lateral transfers, selective interview processes for promotions, out-of-class assignments and subjective decision-making standards for promotions and discriminatory compensation practices.” (Dist. Ct. Op. at 58-59); see also Phillips v. Gates, 329 Fed.Appx. 577, 580 (6th Cir.2009) (finding sufficient, as to the first element, a challenge to the “practices and procedures regarding employee promotions”); Scales v. J.C. Bradford & Co., 925 F.2d 901 (6th Cir.1991) (sustaining challenge to promotion practices, specifically: (1) failure to advertise openings; (2) favoritism to friends and associates of supervisors; and (3) use of subjective criteria).
The specific employment practices identified by Plaintiffs as discriminatory are well documented in the record. The district court found that Defendant’s posted minimum job qualifications “are frequently tailored or altered from the original job descriptions to fit the person whom [] management desires to fill the specific position.” (Dist. Ct. Op. at 6.) “As a factual matter”, the court explained, “the proof also clearly establishes that [Metro’s Water Services Department (“MWS”) ] distorts educational requirements, seniority, and experience in its promotion decisions of its white employees.” (Id, at 65.) In one instance, Metro eliminated a bachelor’s degree requirement for a director position after a qualified black employee applied, and awarded the position to a white applicant without a degree, even though the previous director had both bachelor and master’s degrees. (Id. at 12-13.)
With regard to lateral transfers, the district court found that Metro frequently permits white employees to transfer internally, thereby circumventing the competitive employment application process. (Id.) To the extent that the competitive application process was utilized, the district court found that Defendant’s department managers possess discretion over whom to interview for promotions, and once the relevant employees are identified, the usual practice is that each employee’s supervisor will serve on the interview panel. This results in a highly subjective — and problematic — process. See Wal-Mart Stores, Inc. v. Dukes, — U.S. -, 131 S.Ct. 2541, 2554, 180 L.Ed.2d 374 (2011) (“[A]n employer’s undisciplined system of subjective decisionmaking [can have] precisely the same effects as a system pervaded by impermissible intentional discrimination.” (internal citation and quotation marks omitted; alteration in original)). The problems of this subjective process are exacerbated by the frequent variations in the size and composition of interview panels, and the heavy (and sometimes controlling) weight accorded the results of an oral interview.
Additionally, the district court found that Metro has a policy whereby an employee may be assigned to fill a temporary vacancy in a higher position — a so-called “out of class assignment” — and that after 100 days, the employee would be promoted into the higher position. (Dist. Ct. Op. at 16-19.) White employees, including six individuals identified at trial, would receive the promotion after 100 days, whereas black employees often would not. (Id. at 17.) Additionally, black employees working “out of class” are not always paid at the higher “out of class” rate, even though white employees were. In one situation, as proved at trial, a black employee assumed the duties of a public information officer on an “out-of-class” basis, but was told he was not qualified to fill the position on a permanent basis, even though the position was later filled by a biologist without any public relations background, at a *744higher salary than the temporary employee. (Id. at 19.) Situations like these are consistent with black employees, as the district court found, being “concentrated in lower job classifications at lesser compensation levels.” (Id. at 60.)
The majority does not explain why the specific employment practices proved by Plaintiffs at trial, and identified by the district court, failed to satisfy Plaintiffs’ burden of identification. The majority identifies no factual finding by the district court that is clearly erroneous, and cites no legal authority to support its conclusion. This sort of truncated analysis is particularly troubling in light of the nature and significance of this case — a civil rights class action against a major public employer — and the availability of extensive evidence in the record.
B. Disparate Impact
The district court additionally did not err in finding that Plaintiffs carried their burden to establish that the challenged practices have an adverse impact on a protected group. The district court held that “[t]he specific employment practices alone and in combination have had the effect of denying and delaying promotions to black employees [at Metro], as set forth by Dr. [Michael] Moomaw’s testimony and analyses and Plaintiffs’ other proof.” (Id. at 59.) The majority offers no reason why this conclusion was clearly erroneous.
As to the statistical proof, the district court concluded based on a binomial distribution analysis that the rate of promotions of black employees, across nearly every job category, was three to four standard deviations lower than would be expected in the absence of discrimination. See Vogel v. City of Cincinnati, 959 F.2d 594, 600 (6th Cir.1992); see also Alexander v. Local 496, Laborers’ Int’l Union of N. Am., 177 F.3d 394, 419 (6th Cir.1999) (Batchelder, J., concurring in part and dissenting in part) (reasoning that district court did not clearly err in finding disparate impact “[g]iven the extreme statistical disparity” proved at trial). This analysis reflected the report of Dr. Moomaw, who recategor-ized the MWS workforce and analyzed employment data across four dimensions, finding “stark and significant differences in representation between white and black employees that extend to all categories of MWS’ positions.” (Dist. Ct. Op. at 33); see also Phillips, 329 Fed.Appx. at 581 (stating that “ ‘sufficiently substantial’ statistical disparities raise an inference of disparate impact”).
The district court determined that these “widespread statistical imbalances” were a result of the employment practices challenged by Plaintiffs. (Dist. Ct. Op. at 60.) Rather than determining the effect of each challenged practice, which it found were “not capable of separation for analysis,” the district court analyzed Metro’s promotion practices as “one promotion practice.” See 42 U.S.C. § 2000e-2(k)(l)(B)(i) (stating that if a “complaining party can demonstrate to the court that the elements of a respondent’s decisionmaking process are not capable of separation for analysis, the decisionmaking process may be analyzed as one employment practice”). The majority disagrees with the district court’s decision to treat the promotional process as one practice, because, as it states, Plaintiffs “never attempted to demonstrate that the elements of that process are incapable of separation of analysis.” (Maj. Op. at 740.) But the majority offers no legal authority or citation to the record, or elucidates why the district court’s finding in that regard is clearly erroneous.
Considering the process as a whole, the district court made the following factual findings as to the cause of the statistical abnormalities, none of which the majority contends is clearly erroneous:
*745MWS’s practice of tailoring job descriptions involves removing job requirements for higher positions to favor MWS’s white employees. MWS’s altered educational requirements for positions that have the actual effect of increasing the number of white employees promoted.
Lateral transfers have the effect of denying MWS’s black employees from promotions to higher level positions. MWS used selective oral interviews with open discussion of scoring among panel members, and the results of the oral interview represented from 50% to 80% or 100% of the selection decision, thus undermining the applicant’s seniority and experience.
... MWS’s misuse of out-of-class assignments resulted in the denial of promotions to black employees that were given to MWS’s white employees. MWS also delayed higher compensation paid to black employees who worked six months to two years on an “out-of-class” basis. These black employees were not promoted as white MWS employees were and had to file grievances and objections to receive the higher pay required by the “out-of-class” policy for such work....
MWS’s compensation practice ... resulted in compensation levels of black employees who are also concentrated in the lowest grade levels within the same pay ranges for white employees. The statistics also show that black employees at MWS are also concentrated in lower job classifications at lesser compensation levels....
(Dist. Ct. Op. at 59-60.)
The majority overturns the district court’s finding because, as it explains, “Plaintiffs simply did not present relevant statistical data that MWS’s promotion practices caused an adverse, disparate impact on its black employees.” (Maj. Op. at 741.) According to the majority, “Plaintiffs’ evidence was merely a description of the racial demographics of MWS’s workforce.” (Id. at 8.) The majority explains that Plaintiffs’ statistical evidence “falls short” because, “instead of comparing the employees who actually applied for or were eligible for promotions with those who received them, Plaintiffs compared the promotion of black eiñployees in high-paying positions with the proportion of black employees within the entire MWS workforce.” (Id. at 9.) Even if the latter comparison was problematic, the majority continues, Plaintiffs were “still required to construct a generally qualified applicant pool” from which to make a comparison. (Id.)
As an initial matter, the majority’s insistence on a specific form of statistical evidence has no basis in our case law. We have never “limited a plaintiffs choices in Title VII cases involving statistical analysis in any way,” Isabel v. City of Memphis, 404 F.3d 404, 412 (6th Cir.2005), and, as the Supreme Court recognizes, statistics “come in infinite variety and ... their usefulness depends upon all of the surrounding facts and circumstances.” Int’l Bhd. of Teamsters v. United States, 431 U.S. 324, 339-40, 97 S.Ct. 1843, 52 L.Ed.2d 396 (1977). The statistical evidence relied upon by the district court was relevant, and to the extent it was less than perfect, “such flaws relate to the weight of the [evidence] which is a matter for the trier of fact.” Phillips v. Cohen, 400 F.3d 388, 401-02 (6th Cir.2005). The majority offers no legal authority to support its argument that the district court’s evaluation of the statistical evidence was improper as a matter of law. See Johnson v. U.S. Dep’t of Health and Human Servs., 30 F.3d 45, 48 (6th Cir.1994) (holding that the district court’s view of the sufficiency of statistical evidence is reviewed for clear error).
*746Moreover, comparing rates of actual promotion would be unhelpful in this case because, as the district court found, that comparison would not capture the reality of the promotions process at MWS. First, it would understate the promotion rate of white employees. Based on the trial record, the district court found that through “lateral transfers” and “out of class” assignments, MWS would frequently promote white employees outside of the normal application process. Second, it would overstate the promotion rate of black employees because, as the district court found, MWS, and its promotion process more generally, discouraged black employees from applying for promotions; this was accomplished by informing potential applications that the position was “not an appropriate fit;” permitting subjective evaluations by hiring managers; and altering job requirements to fit preselected candidates. See Kreuzer v. Brown, 128 F.3d 359, 364 n. 2 (6th Cir.1997) (citing Harless v. Duck, 619 F.2d 611, 617-18 (6th Cir.1980)).
The majority suggests that Plaintiffs’ failure to compare actual promotion rates could be excused if Plaintiffs had “constructed] a generally qualified applicant pool” and compared the promotion rates within that pool. (Maj. Op. at 742.) Such a statement, however, reflects the majority’s fundamental misunderstanding of this case. As the district court found, “the proof ... clearly establishes that MWS distorts educational requirements, seniority, and experience in its promotion decisions of its white employees,” and it is therefore not possible to determine the actual qualifications for many positions. (Dist. Ct. Op. at 65.)
Additionally, even if the actual qualifications for each position could be determined, any failure to control for this variable is not fatal under the circumstances of this case, given the extent to which Metro obfuscated and apparently manipulated the promotion process. See Phillips, 400 F.3d at 400-02 (reversing district court’s dismissal of disparate impact challenge to promotion process, and reasoning that the plaintiffs failure to control for employees’ qualifications in statistical data did not render its statistics legally insufficient). The majority’s reasoning to the contrary runs counter to well-established case law in this Circuit. See id. (citing Scales, 925 F.2d at 906 (finding gender discrimination on the basis that it took women longer than men to be promoted to the first managerial level in the company)).
Finally, the majority ignores the non-statistical evidence adduced at trial, thereby ignoring the extensive testimony by individual plaintiffs as to their “personal experiences with promotion decisions at MWS.” (Dist. Ct. Op. at 19-25); see also Int'l Bhd. of Teamsters, 431 U.S. at 340, 97 S.Ct. 1843 (“The individuals who testified about their personal experiences with the company brought the cold numbers convincingly to life.”). Our cases have recognized that “expert statistical evidence in disparate impact cases is not to be considered in a vacuum, as the only evidence permitting plaintiffs to meet their prima facie test; it must be considered in light of all the evidence in the record.” Phillips, 400 F.3d at 401 (holding that non-statistical evidence of disparate of impact “compensate[ed] to some degree for plaintiffs’ failure to demonstrate conclusively that they are promoted at lower rates than white employees.” (internal citation and quotation marks omitted)); see also Wal-Mart Stores, 131 S.Ct. at 2556 (recognizing the relevance of testimonial evidence apart from statistical evidence in disparate impact cases). This point is lost on the majority.
Accordingly, because the district court committed no error of law, and the find*747ings of fact underlying its decision are not clearly erroneous, we should affirm. The majority refuses to do so, and I respectfully dissent.