Opinion
Civil Action No. 11-cv-00256-MSK-KMT
07-08-2016
MARIAN G. KERNER; and ROMONA J. LOPEZ, on behalf of themselves and all others similarly situated, Plaintiffs, v. CITY AND COUNTY OF DENVER, Defendant.
OPINION AND ORDER ON DAMAGES
THIS MATTER comes before the Court pursuant to the parties briefing (# 224, 225) addressing the appropriate calculations for purposes of awarding damages in this case.
The Court assumes the reader's familiarity with the proceedings to date. In summary, the Plaintiffs, a class of black and Hispanic applicants for various positions with the City and County of Denver ("Denver"), alleged that Denver's use of a written employment screening test, the "AccuPlacer," had a disparate impact on minority applicants in violation of Title VII of the Civil Rights Act, 42 U.S.C. § 2000e.
Following an eight-day bench trial in April 2016, the Court found in favor of the Plaintiffs, but observed that both parties' damage experts had provided constantly-shifting models and calculations at trial, making an immediate award of damages impossible. Addressing various points of dispute that had arisen between the parties with regard to damages, the Court set out, in detail, a series of findings that would govern an award of damages, and directed the parties to have their experts make calculations consistent with the directives and based on the data in the trial record. The Court expected that, given the detailed and specific instructions it gave, the parties would be able to agree upon a final damage amount that could be embodied in a judgment.
The parties' briefing makes clear that this expectation was optimistic. The parties have identified six separate areas in which they were unable to agree how to implement the Court's instructions. The Court has now reviewed the parties' briefing on these issues, the instructions that it gave to the parties, and the trial record. It now proceeds to resolve those disputes and enter judgment. In doing so, the Court notes that its analysis is guided by Docket # 225-1, a chart prepared by the Plaintiffs that shows Dr. Mullin's "baseline" damage calculation of $1,164,523. That chart identifies the disputed issues between the parties and identifies the dollar value attributed to the Plaintiffs' positions on the various issues.
A. Calculation of shortfall
The Court directed that the first step of the damages calculation entail a calculation of "shortfall" which measures the disparate impact the test had on minority test takers. It is a calculation of "the percent of black and Hispanic applicants in the initial pool minus the percent of black and Hispanic applicants remaining in the pool after the AccuPlacer test was administered, with that figure multiplied by the number of jobs in the classification for which hires were actually made." The Court directed that this calculation be made separately for each of the 20 different job classifications at issue in the case. It is clear that the parties disagree as to how to perform this shortfall calculation, although, as explained below, it is not clear whether this disagreement has any tangible consequences.
At trial, there were 21 job classifications at issue. The Court found that there was no disparate impact with regard to one classification, thus the measure for calculation of damage is limited to 20 classifications.
Facially, Denver's explanation for its shortfall calculations appears to contradict the instructions given by the Court. Exhibit B to Denver's brief is a set of assumptions and calculations by its damages expert, Dr. Mullin. Page 2 of that exhibit summarizes Dr. Mullin's assumptions and the additional calculations required by the Court. That list of assumptions notes that, for purposes of trial, Dr. Mullin did not calculate damages for 12 of the 20 job classifications. It explains that, to calculate damages for the additional 12 classifications as required by the Court's ruling, Dr. Mullin calculated the damages for the additional classes by "simple scale up of 8 job classes based on shortfall proportions and additional scale up based on weighted difference in average starting salaries between 8 significant job classes and remaining classes with a shortfall[,] weighted by shortfall." Denver's brief does not elaborate on this methodology, and the description is largely inscrutable to the Court. It does not appear that Dr. Mullin performed the classification-by-classification shortfall analysis as the Court directed.
The Plaintiffs' expert, Dr. Bardwell, appears to have complied with the Court's instructions. In addition to the 8 classifications for which the where the parties identified significant shortfalls at trial - Dr. Bardwell concluded that 7 additional classifications showed small shortfalls reflecting a range of .06 to .1 of a hire.
It was the Court's intention that the calculation of damages would occur on a classification-by-classification basis, and thus, an individualized determination of each job classification's particular shortfall would be an essential component of the calculation. Although it is somewhat unclear as to whether Denver did or did not perform those individualized calculations, the Plaintiffs concede that the difference between Denver's calculations on this point and their own reflects a difference in damages of a mere $ 2,182. Accordingly, the Court adopts the Plaintiffs' method of calculating shortfall. Denver's initial calculation of damages -with all of the disputes resolved in its favor - yields the baseline figure of $ 1,164.523. The Court begins from that calculation, and then adds $ 2,182 to that amount to reflect Dr. Bardwell's more correct calculation of shortfall.
B. Damages start date
Calculating the appropriate hire date for a given job classification is necessary for two purposes: first, it measures when damages begin to accrue, and second, because the Court has declared that damage accrual will end after a specific period of time, the hire date implicitly establishes the endpoint for damage calculations. In short, the hire date bookends all damage calculations. The parties disagree about how to calculate the hire date for each given classification.
At trial, Dr. Bardwell explained that he "made an assumption of a 30-day delay between the date of application and the date of hire." The Court understands this testimony to mean that Dr. Bardwell examined the date on which an applicant sought a job, and, assuming the applicant received the job, began calculating that applicant's earnings 30 days later. That determination is a simple one for job classifications - and again, it was the Court's intention that there would be separate damages calculations for each job classification -- in which only one hire was made during the relevant period: the hire date for damages purposes is the date on which the only hire was made. For classifications that made multiple hires over the entire time period that the test was in use, the selection of a particular "hire date" for all applicants in that class is more complicated. Perhaps the most reasonable means of calculating a hire date in that situation is to calculate some sort of weighted average hiring date that falls in between actual dates of hire.
For example, if Denver filled two positions on February 1 of the year, one position on July 1 of the year, and one position on August 1, the weighted average hire date might measure from January 1 of that year and be: (2 x 31 days) + (180 days) + (211 days), for a total of 453 days, divide that sum by four for an average of 113 days, and thus set the average hire date at April 23.
According to Exhibit 2 to Denver's brief, Dr. Mullin selected the hire date for his calculations by deriving the "average of start dates of 298 hires by job class." The "by job class" reference here is unclear: it is not clear whether Dr. Mullin derived a single average start date for the entire class, or whether he derived 20 average hire dates, one for each classification. However, the calculation of "average" start dates by Dr. Mullin appears to conform to the Court's expectations as to how the parties would compute hire dates for each job classification.
Dr. Bardwell instead chose a different meaning for the term "hire date," selecting the last date on which any of the positions at issue in this case were filled, and apparently using that as the hire date for all classifications. Using this late date - August 18, 2008 - benefits the Plaintiffs by delaying the date for the ending of damage accrual to as late as possible. But doing so distorts the model the Court intended to be used. For example, according to Trial Exhibit 7, three openings in job classification for Aviation Customer Service Agent were filled during the relevant period, all on July 9, 2007. For this job classification, the hiring date for damages calculation purposes would unambiguously begin on July 9, 2007, and conclude 8.2 years later (somewhere in the neighborhood of September 24, 2015). Artificially selecting a hiring date of August 18, 2008 for this classification for damages purposes would delay the accrual of damages for these positions for over a year and, similarly, extend the ending date for such damages by an equal amount. There is no logical basis for doing so. Dr. Bardwell's basis for selecting a single hiring date for all classes seems to be based on his interpretation of an instruction from the Court that damages would cease at the end of 8.2 years "from the last hiring date at issue here." Admittedly, the "here" at the end of that statement may be confusing, but the Court's clear intention was that each job classification would entail a separate damage calculation, making the selection of a unitary hiring date that would apply to all damage calculations to be impractical.
At the same time, selecting this late date would also have the effect of delaying the beginning of damage accrual, to the Plaintiffs' detriment. It is not clear to the Court whether Dr. Bardwell's model is internally inconsistent, using August 18, 2008 as the hiring date for both the beginning and end of damages calculations, or whether Dr. Bardwell has chosen a different "hiring date" for purposes of commencing damage accrual.
Accordingly, based on its understanding of the parties' relative positions on this point, the Court adopts the mechanism selected by Dr. Mullin for calculating a hiring date. Thus, it does not further adjust Dr. Mullin's baseline damage calculation in this regard.
C. Actual paid benefits
In calculating damages, the Court directed that the parties determine the median value of fringe benefits that were actually paid by Denver to the successful hires, as a percentage of salary. It was the Court's understanding that the parties generally agreed that this benefits figure typically reflected approximately 26% of an employee's salary.
Although the parties appear to disagree with each other over what these calculations yield, to the tune of more than $ 350,000, the briefing does not adequately apprise the Court of the particulars of the parties' dispute. Both parties acknowledge the general formula described by the Court, both parties agree as to the source data that are to be used to make that calculation (Trial Exhibits 8 or 126, which are identical), and both parties insist that they followed the correct formula. The Plaintiffs contend that Dr. Mullin "reduces the benefits received by employees approximately 15 to 20 percent in the first year of his calculations" as compared to the trial exhibits. Denver alleges that Dr. Bardwell is using "new rates that were not offered at trial."
Examining the parties' damage calculation spreadsheets, the Court agrees with the Plaintiffs. By means of example, the Court looks to each expert's calculations regarding the job classification "311 Customer Service Agent". Dr. Mullin's spreadsheet for this category (# 224-2 at 6) lists the value of benefits in this classification at $ 10,325 for 2007; Dr. Bardwell's spreadsheet for this category (# 225-4 at 6) lists the value of benefits at $ 12,850 for 2007. Trial exhibits 8 and 126 both list the value of benefits for "311 Customer Service Agent" as $ 12,850. This is precisely the figure Dr. Bardwell used, and thus, the Court will assume that Dr. Bardwell's calculation is correct. Whatever rationale that Dr. Mullin may have used in reducing the benefit values below those shown in Trial Exhibits 8 and 126, that reduction was not directed by the Court or justified by Dr. Mullin in the instant briefing. Accordingly, the Court adopts Dr. Bardwell's calculations and increases the baseline damages calculated by Dr. Mullin by an additional $ 355,329, the apparent difference between the parties' overall calculations on this point.
Column headings in these trial exhibits seem to suggest that the benefit values they reflect are 2008 figures, but neither party's 2008 benefit value calculations match the numbers in the exhibits. Thus, the Court will assume that the trial exhibits reflect 2007 values.
D. Unemployment period
Both experts included a figure in their initial damage calculations that accounted for "unemployment" in some capacity. Dr. Bardwell never particularly described this element of his calculations, and Dr. Mullin addressed it only briefly, explaining that:
I factored in a short time period with regard to the potential for a duration of unemployment. So, as indicated previously, I calculated an average start date, to start the damages at, and then I built in a 20 week potential unemployment. I don't know that that's
necessarily, because there is no evidence that everybody was unemployed, but I put it in there way, I think it was 20 weeks.. . . [T]here is no necessarily direct evidence offer indication that all of the individuals at issue were unemployed when they applied with the City. Obviously, you can apply for a job even if you still have a job. But I know Dr. Bardwell used 30 days, I used a little bit longer period, so to extent, he was being defendant-friendly, and I was being plaintiff-friendly. I didn't want to use nothing, because 30 days didn't seem to be one of those issues that you really want to get into an argument about with another expert on the other side. So I pulled in a duration of unemployment. I typically do it with terminations more than I would do it with a missed hire. But there is the possibility that some of the individuals at issue are unemployed so I wanted to account for that at least to some extent.Dr. Mullin's explanation notwithstanding, it is unclear to the Court exactly what accommodations were being made for "potential unemployment." As best the Court can determine, it would seem that both experts expected that rejected applicants would need a period of time to find alternative - that is, mitigating - employment, and thus, each expert delayed some period of time in the damage calculation before beginning to factor in mitigation effects. This element was factored into the parties' initial damage models during the trial, but, as noted, the parties eventually refashioned those models by the end of trial and it is not clear to the Court whether this unemployment period calculation remained present in the parties' final models.
The disagreement between the parties at this stage suggests that it did, and now the Plaintiffs ask the Court to adopt Dr. Mullin's "plaintiff-friendly" estimation of 20 weeks of unemployment, while Denver asks the Court to adopt Dr. Bardwell's "defendant-friendly" estimation of 30 days, a difference amounting to some $ 195,000. In essence, each party now seeks to conclusively enshrine in the final judgment the benefit of the doubt that the opponent's expert extended to them in initial calculations.
The Court rejects both contentions. The trial testimony about this calculation was limited (the quoted text above accounts for the practical entirety of the discussion about it), one-sided (Dr. Bardwell never discussed his application of the calculation and his reasons therefor), and the purpose and rationale for the calculation remains unclear to the Court even at this time. When the Court was devising a damage model to apply, it did not intend to include any calculation reflecting an unemployment period. Indeed, the very premise of the calculation is that most or all of the rejected applicants were unemployed at the time they sought a position with Denver, and that they remained so for several weeks after being rejected. The Court discussed this assumption at some length in its oral ruling, finding the evidence on this point to be inconclusive - that is, that there was evidence that some applicants (if not an outright majority) were still employed at a previous job when they were rejected by Denver. In such circumstances, a calculation that assumes those applicants to be unemployed for a period as long as 20 weeks would be inappropriate.
The damage model directed by the Court anticipated that damage accrual would begin on the 30th day following the calculated hire date, and, for purposes of simplicity, it assumed that mitigation reduction would begin on this date as well. In this sense, to the extent that Dr. Mullin used a 30-day unemployment period calculation when deriving the baseline damages calculation referred to above, that calculation most closely matches the Court's intentions. The Plaintiffs can hardly claim to be prejudiced by a conclusion that mitigation of damages would begin 30 days after a hiring date, as this is the apparent assumption that their own expert, Dr. Bardwell, initially proposed. Accordingly, the Court declines to make any addition to the baseline calculation on this point.
As the Court instructed, wage and benefit accrual and mitigation reductions would be "prorate[d] for the first year of the calculation based on a 30 day after hiring date calculation." --------
E. Annual attrition date
The damage model envisioned by the Court required an initial calculation of lost salary and benefits, reduced by mitigation earnings. That sum was to be adjusted annually upward (to reflect cost of living and inflation adjustments) and downward (to reflect expected attrition in the employees who left employment with Denver). The parties disagree as to when the annual adjustment should be made. The Plaintiffs contend that each annual adjustment should be made on the anniversary of the "hiring date" (that is, a "rolling year"); Denver contends that the adjustment should be made on the "mid-point" date of each calendar year (that is, July 1), with the first adjustment being made on July 1, 2008.
In reviewing its ruling, the Court recognizes that its directives on this point were unclear. Indeed, upon reflection, it appears that the Court did not fully contemplate the issue in the first instance. The Court generally adopted Dr. Mullin's mitigation approach, which used the "mid-point" date, but directed that first-year damages be "prorate[d]," a concept that is only logical when using an approach based on calendar years, not rolling years. Upon reflection, it may be that the Plaintiffs' rolling year approach yields more reasonable results because it depends on the hiring date selected. A rolling year approach ensures that a full year has passed before a year's worth of attrition is calculated. In contrast, Dr. Mullin's "mid-point" approach applies an entire year's worth of employee attrition to a hire date that occurred only 40 days earlier.
At this point in time, the Court is persuaded that the rolling year approach urged by the Plaintiffs is the appropriate means of applying annual adjustments. Because the Court intended to have each job classification analyzed separately, the hiring dates for each classification could differ. Take three examples: as noted above, the hiring date in the "Aviation Customer Service Agent" classification might be as early as July 2007. The hiring date for the classification of "Vehicle Impound Clerk" would be late December 2007 (two hires made December 10, 2007; one hire on January 22, 2008). The hiring date for the classification of "Fingerprint Identification Clerk" would appear to be August 18, 2008 (based on the single hire in that classification). Using Denver's approach, the Court would apply a year's worth of attrition to the "Aviation Customer Service Agent" classification on July 1, 2008, exactly one year after that classification's hire date, a perfectly sound result. But it would also apply a full year's worth of attrition on that date to incumbents in the "Vehicle Impound Clerk" classification, even though those incumbents had only been on the job for about six months at that point. And it would apply a full year's worth of attrition to the incumbent in the "Fingerprint Identification Clerk" position on July 1, 2008, even though that position had not even been filled by that date. This result would be illogical.
By adopting the Plaintiffs' rolling year approach, the first attrition adjustment for the classification of "Vehicle Impound Clerk" would occur in December 2008, after those incumbents had been on staff for a full year. The first attrition adjustment to the "Fingerprint Identification Clerk" classification would occur on August 18, 2009, after a full year in that position, and so on. Not only does this approach honor the Court's intention that each classification's damages be calculated separately and according to its own particular details, it also avoids illogical results, such as the notion that a quarter of hires in a given position would quit within a few days or weeks of receiving that very job. Accordingly, the Court adopts the Plaintiffs' position that attrition should be calculated on a rolling year basis. Docket # 255-1 indicates that this change adds $ 152,773 to the baseline damage calculation.
F. Cliff attrition date
The last dispute between the parties concerns when "cliff attrition" - that is, the date on which all remaining damages calculations stops. It is the date on which one reasonably can expect that a hypothetical person excluded from a Denver city job by virtue of failing the Accuplacer test would have fully mitigated his/her loss in the private sector.
At trial, Dr. Mullin proposed that this date should be fixed at five years after the relevant hiring date. Dr. Bardwell's proposal on this point changed during trial. At first, Dr. Bardwell contended that damages should continue to accrue for an additional 10 years after the date of trial (18 years from a hiring date). By the end of trial, however, Dr. Bardwell had reduced his cliff attrition calculation to 8.2 years from a hiring date. The Court ultimately adopted Dr. Bardwell's 8.2 year figure. It explained:
During trial, however, Dr. Bardwell abandoned this approach, testifying on rebuttal that he had settled on a "cliff attrition" period of 8.2 years. His explanation for the 8.2 year period is "based on City's data for attrition without assuming full mitigation at 5 years." At trial, he explained that this figure "is computed using the same actual hire data that Dr. Mullin used for his attrition analysis, except that I don't assume that it's cut off at five years, and, instead, use standard technique to estimate the average work - the average term of employment at the City's position as 8.2 years. So it's a standard attrition analysis, but not just be truncating the data at five years." Dr. Bardwell explained that he used a statistical technique called the Cox Hazard Model to do this calculation, and that he chose a point that reflected "90 percent probability," - which corresponded to 8.2 years on the time axis - although he did not explain the significance of that choice or the reasons why he selected it. The Court assumes that Dr. Bardwell's calculation reflects a point in time where 90% of hires will have departed from City employment via attrition, such that the damage calculation can effectively conclude at that point in time. Beyond explaining his 8.2 year figure, Dr. Bardwell did not discuss his new attrition calculations. . . . (Emphasis added.)
Although the Court adopted Dr. Bardwell's 8.2 year cliff attrition figure, the Plaintiffs now argue that the Court should instead extend that calculation to 21 years, nearly tripling the baseline damage calculations. The Plaintiffs base this argument on the Court's passing comment that assumed that Dr. Bardwell's otherwise-unexplained "90 percent probability" comment was referring to 90% of the hypothetical hires having completed their tenures with Denver by 8.2 years. The Plaintiffs explain that, according to Dr. Bardwell's calculations, it would take 21 years for 90% of hypothetical hires to end their careers.
The Plaintiffs' request here is essentially a request to re-open the trial record, to offer additional evidence, and then to ask the Court to reconsider its prior ruling. In other words, the Plaintiffs ask for leave to allow Dr. Bardwell to clarify and further explain that the "90 percent probability" comment does not equate to the completion of 90% of mitigation, to abandon his previous position that cliff attrition is appropriate at 8.2 years, to assert that the proper measure of cliff attrition should be the point at which 90% of hypothetical hires' tenures end, and to suggest that this date is 21 years after hiring begins. The Court denies this request.
The Court observes that Dr. Bardwell proposed two distinct - and starkly different - cliff attrition figures during trial in this case, and is now proposing a third. An expert's willingness to perpetually re-shape his opinions suggests a lack of intellectual rigor and the absence of solid foundations supporting any of those opinions. Moreover, it calls into question the expert's overall credibility. The Court selected Dr. Bardwell's 8.2 year cliff attrition figure over Dr. Mullin's 5 year figure in part because it seemed less arbitrary and more data-driven than Dr. Mullin's, and in part because the Court (apparently mistakenly) believed that Dr. Bardwell was suggesting that his figure accounted for 90% of the class' damages. In this regard, the Plaintiffs actually benefitted from the Court's resolution of the considerable ambiguities in Dr. Bardwell's testimony, as the Court resolved those ambiguities in a way that served to bolster Dr. Bardwell's credibility. Were the Court to permit the Plaintiffs were to reopen the record to clarify Dr. Bardwell's testimony about the "90% figure" and to posit a third cliff attrition figure, the Court might very well be inclined to revisit the entire "5 years vs. 8.2 years (vs. 21 years)" cliff attrition question, and it might very well conclude upon reconsideration that Dr. Mullin's unchanged figure was now the most credible approach.
The Plaintiffs, having already secured the cliff attrition figure they proposed at trial, may not now urge that an even longer period should be applied. Accordingly, the Court declines the request to deviate from the 8.2 year figure it previously adopted.
G. Final issues raised by Plaintiffs
The final portion of the Plaintiffs' brief requests that the Court make "an additional" award of $ 100,000 to Ms. Kerner and $ 50,000 to Ms. Lopez, to compensate them for their efforts as class representatives, and to request smaller amounts for an unspecified number of additional class members who participated in discovery in this case. It is not clear whether the Plaintiffs are requesting that these sums be segregated from the pool of damages awarded to the class and be diverted to Ms. Kerner, Ms. Lopez, and the others, or whether the Plaintiffs are suggesting that the Court should further increase the damages awarded against Denver to include these sums.
The latter proposition is, frankly, preposterous. The Plaintiffs point to no law that would support awarding additional damages, over and above those reflecting actual injuries sustained by the class members, against an employer in a disparate impact case for the purpose of compensating a class representative for "steadfast devotion" to the case. See Hadix v. Johnson, 322 F.3d 895, 897-98 (6th Cir. 2003) (rejecting a request for an award to class representative that would require "defendants to pay the incentive award [as] an additional expenditure").
The former proposition, on the other hand, finds some support in the law. In class action cases where the resultant judgment creates a "common fund" of damages to be shared by the class members, courts have sometimes approved "incentive awards" that steer a portion of that common fund to the class representatives or other class members to recover an additional share of those funds as compensation for enduring the expenses and burdens of litigation. Id.; see also UFCW Local 880 v. Newmont Mining Corp., 352 Fed.Appx. 232, 235-36 (10th Cir. 2009). However, because incentive awards to class representatives involve taking money from the common fund available to all class members, they are subject to particular scrutiny by the Court for fairness and reasonableness. Hadix, 322 F.3d at 897.
Because any request to pay an incentive award to Ms. Kerner, Ms. Lopez, or others will have to be made as part of a request to approve a distribution plan to the class, that request, along with all other aspects of a proposed distribution plan, must find support (or at least no persuasive opposition) by members of the class that are affected. Such proposal requires notice to all class members, who must have had an opportunity to lodge objections. Thus, the Court will not approve any incentive awards at this time.
However, the Court does note, in passing, that the $ 100,000 and $ 50,000 awards contemplated by the Plaintiffs here are extraordinarily large and would be very likely to be rejected. Hadix recounts several examples of incentive awards that were found to be reasonable, including: $50,000 to each of six class representatives in a case in which the common fund was in excess of $ 56 million (i.e. the incentive award was approximately .5% of the total fund); awards of between $ 35,000 and $ 55,000 to five class representatives in a case where the common fund was more than $ 18 million (i.e. awards valued at approximately 1.1% of the total fund); and a $50,000 incentive on a $ 5.25 million fund (approximately 1% of the fund). 322 F.3d at 898. Here, the more than $ 150,000 in incentive awards contemplated would account for nearly 10% of the total judgment against Denver, and after litigation costs are deducted from that fund, the percentage is likely to swell even more - a remarkably high amount to be diverted away from all class members in order to benefit only the class representatives. The Court advises the Plaintiffs that it is extremely unlikely to approve of incentive awards that, in total, significantly exceed 1% of the funds that are actually distributed to class members.
H. Conclusion
For the foregoing reasons, the Court adopts Dr. Mullin's baseline damages calculation of $ 1,164,523, and adds the following sums: $ 2,182 (reflecting the correct shortfall calculation), $355,329 (reflecting actual benefit levels), and $ 152,773 (reflecting a rolling year for mitigation purposes). This yields a total damage award of $ 1,674,807. The Court will enter judgment in that amount contemporaneously with this Order.
Dated this 8th day of July, 2016.
BY THE COURT:
/s/_________
Marcia S. Krieger
Chief United States District Judge