MEMORANDUM OPINION
SULLIVAN, District Judge.
I. Introduction
On November 2, 1999, this Court approved a final Consent Decree to resolve an employment discrimination class action involving African American management employees at Amtrak. The Consent Decree provides for a compensation study to be performed. If the compensation study shows that there are disparities between the salaries of African American managers and white managers to the detriment of African American managers, the Consent Decree provides for salary adjustments to be made.
Currently, the compensation study has been completed and Amtrak has increased the salaries of some African American class members. The Class plaintiffs have filed a Motion for Enforcement of the Consent Decree. Amtrak claims that it has completed all that it agreed to do under the Consent Decree. In addition, Amtrak has filed an uncontested Motion requesting that the Court accept its response to plaintiffs’ motion under seal because it contains statistical analyses based on the financial history of persons who are not named plaintiffs.
II. Factual Background
Section IV.D. of the Consent Decree provides that Amtrak is to hire a third party consultant to study the compensation of Amtrak’s management employees. In addition, the Consent Decree states:
If the compensation study shows that there are disparities in salaries between African American Management employees and white employees in comparable positions, to the disadvantage of African American Management employees, Amtrak will increase the compensation of African American Management employees to eliminate racial disparities.
§ IV.D.2.
Amtrak hired Dr. Jessica Pollner to perform the compensation analysis. As permitted by the Consent Decree, Class Counsel met with Dr. Pollner as she was developing her analysis. Dr. Pollner used a multiple regression analysis to determine a “predicted salary” for each employee. Dr. Pollner then compared each employee’s predicted salary to his or her actual salary. She found that 56% (238 of 427) of African American employees were paid less than their expected salaries and 44% (189 of 427) were paid more than their expected salaries. Moreover, 5.4% of African Americans (28 of 427) were significantly underpaid at the 90% confidence level and 2.3% were significantly underpaid at
the 95% confidence level. In the end, Dr. Pollner found that white management employees were paid about $455,000 per year more than their African American counterparts.
Amtrak takes the position that the Consent Decree requires Amtrak to adjust only the compensation of those class members for whom the disparity between their salaries and their predicted salaries are statistically significant. Amtrak also takes the position that the Consent Decree requires it to adjust only the salaries of those few class members to the point where the discrepancies between their actual compensation and their predicted salaries are not statistically significant. Amtrak proposed to reduce the disparities by increasing the salary of 28 African American employees by an aggregate of $52,171 per year. This would not make their salaries equal to their predicted salary but it would make the difference between their actual salaries and their predicted salaries statistically insignificant.
Class plaintiffs objected to Amtrak’s position and hired their own expert, who came to contrary findings. Under plaintiffs’ expert’s analysis, 308 African American managers are underpaid a total of $1,289,000.
III. Discussion
A. All class members paid less than expected are entitled to adjustments sufficient to eliminate the overall disparity.
It is well-settled that the “scope of a consent decree must be discerned within its four corners.”
United States v. Armour & Co.,
402 U.S. 673, 682, 91 S.Ct. 1752, 29 L.Ed.2d 256 (1971). Moreover, “the construction of a consent decree is essentially a matter of contract law.”
Citizens For A Better Environment v. Gorsuch,
718 F.2d 1117, 1125 (D.C.Cir.1983). “Under general contract law, the plain and unambiguous meaning of an instrument is controlling, and the Court determines the intentions of the parties from the language used by the parties to express their agreement.”
See WMATA v. Mergentine Corp.,
626 F.2d 959, 961 (D.C.Cir.1980);
see Lucas v. U.S. Army Corps of Eng’rs,
789 F.Supp. 14, 16 (D.D.C.1992) (“Intent is construed by an objective standard and evidenced from the words of the contract itself.”).
1. Class-wide relief
Plaintiffs argue that this suit was brought as a class action to address a pattern and practice of discrimination. Patterns of discrimination are usually fashioned out of relatively small disparities suffered by a large number of class members rather than a relatively small number of large disparities. Plaintiffs argue that when a pay discrimination violation is established in an employment discrimination class action, courts fashion remedies that benefit all class members, not just those whose individual disparities are great. The Decree refers to “disparities,” “salaries,” “employees,” and “positions” in the plural. Plaintiffs claim that the plural reference is consistent with the class purpose of the lawsuit and class-wide nature of the resolution sought through the Consent Decree.
Plaintiffs maintain that the Consent Decree requires Amtrak to make salary adjustments necessary “to eliminate
any
racial disparities” not just statistically significant disparities. § IV.D.2. (emphasis added). Plaintiffs contend that courts determine whether an employer may be liable in a discriminatory pay case by measuring whether there are statistically significant disparities in compensation. If statistical significance is found, courts do not limit relief to elimination of the disparity. Rather, courts adjust compensa
tion to the predicted compensation level based on the non-discriminatory variables in the analysis.
Defendant argues that statistical significance is a commonly used method for separating irrelevant factors, or random differences, from differences that can be attributed to racial differences.
See Rudebusch v. Hughes,
313 F.3d 506, 527 (9th Cir.2002). Defendants maintain that the courts and literature agree that an interval around predicted value is the most appropriate way to assess a disparity and that the most commonly accepted interval is 0.05 or two standard deviations (the 95% significance level.)
Id.
Defendant claims that it met its obligations under the Consent Decree when it followed Dr. Pollner’s recommendation and adjusted the salary of 28 class members who were underpaid at the 90% confidence level to within the 90% confidence level. Defendant asserts that this eliminated any statistically-identified racial disparity.
In evaluating the four corners of this Consent Decree, the Court finds that Amtrak’s decision to provide relief to only those employees who are underpaid to a statistically significant degree and only to the point that the disparity is no longer statistically significant is not supported by the plain meaning of the Consent Decree. Most strikingly, the Court has found no mention of statistical significance in the Consent Decree. While Amtrak claims that statistical significance is a commonly accepted method for separating irrelevant factors, the fact remains that there is no contemplation of statistical significance in the Consent Decree. In writing this Consent Decree, the parties vigorously negotiated the terms. To allow a party to benefit from stringent terms not explicitly included in the Consent Decree, even if they are commonly used in other cases, would be improper.
Here, the Consent Decree states: “If the compensation study shows that there are disparities in salaries between African American Management employees and white employees in comparable positions, to the disadvantage of African American Management employees, Amtrak will increase the compensation of African American Management employees to eliminate racial disparities.” The plain meaning of that provision provides that any African American employee who makes less than her white counterpart should have her salary increased so that the two employees receive the
same
salary, not merely a salary that is no longer statistically significantly different. As the Consent Decree speculates, the natural way to accomplish this is to determine the predicted salary for a certain position using the data for white employees, and then adjust the salary of any African American employee who does not receive at least the same salary.
Thus, to the extent that this has not been done, the Court GRANTS IN PART the Motion for Enforcement of the Consent Decree.
B. The Court will not Substitute the Analysis of Plaintiffs’ Expert for that of Defendant’s Expert
The Consent Decree provides:
Within 60 days of the revisions to the job descriptions, as set forth in Section IV(E) of this Decree, Amtrak shall hire an outside consultant to perform a study ... Class Counsel shall have the opportunity to meet with the consultant before the consultant completes his or her report to discuss the factors the consultant intends to take into account in ana
lyzing the appropriateness of salaries, how the consultant intends to take those factors into account, and how the consultant will identify comparable positions.
Def. Mot.App. A. Consent Decree § rV.D.l. Defendant hired Dr. Jessica Pollner, and, as provided in the Consent Decree, Class Counsel met with Dr. Poll-ner. Dr. Pollner considered Class Counsel’s positions on several issues. Plaintiffs then hired Dr. Richard Stanley to perform an analysis of the issue. Dr. Stanley used a similar multiple regression analysis but saw three “flaws” in Dr. Pollner’s analysis. Plaintiffs ask that the Court instruct Amtrak to use Dr. Stanley’s analysis.
Although the parties could have negotiated to each have their own expert perform an analysis and submit the analyses to the Court, that was not what the parties bargained to do. For the Court to overlook this bargained-for-exchange would be to make the terms of the Consent Decree meaningless. Thus, plaintiffs did not reserve the right to have their own expert’s opinion considered, and the Court refuses to substitute Dr. Stanley’s analysis for that of Dr. Pollner.
While the Court is unwilling to substitute the analysis of plaintiffs expert for that of the expert specified in the Consent Decree, the Court will consider the issues plaintiffs raised with Dr. Pollner’s analysis.
1. Plaintiffs claim that Dr. Pollner used tivo “tainted” variables.
In her analysis, Dr. Pollner controlled for several factors, including EEO category, tenure, tenure squared, job title, SBU, region, an employee’s initial salary, and an employee’s salary band and zone. Plaintiffs believes that Dr. Pollner improperly controlled for two “tainted variables”: an employee’s initial salary and an employee’s salary band and zone.
“A ‘tainted variable’ is one whose value is affected by discrimination and has the effect of concealing disparities due to discrimination.”
Butler v. Home Depot,
1997 WL 605754, *10, 1997 Dist. LEXIS 16296, at *37 (N.D.Cal. Aug. 29, 1997).
Defendant asserts that, as stated in Dr. Pollner’s declaration, “The methodology utilized in the Final Report follows a generally accepted standard among statisticians for studies of disparities in compensation.” Def. Mot.App. B., Ex. 2, ¶ 5. Defendant points out that plaintiffs never allege that Dr. Pollner’s report failed to meet the standards “generally accepted ... by compensation specialists,” the criteria specified in the Consent Decree.
A. Initial Salary
Plaintiffs claim that the use of initial salary is improper because it fails to acknowledge the discrimination in pay at the starting salary. Plaintiffs assert that controlling for this variable would result in the systematic deflation of predicted salaries for African American management employees.
Plaintiffs maintain that even if initial salary were not a tainted variable, its use would be suspect without some type of reasonable adjustment for the passage of time.
In response, defendants argue that initial salary is not a
per se
tainted variable. Initial salary is often influenced by factors not included in the compensation study, including education and prior experience. Dr. Pollner noted that “Initial salary is also a strong determinant of current compensation” and “initial salary cannot be overlooked when modeling current compensation.” Def. Mot.App. B, ¶ 11. She further concluded that “Dr. Stanley’s hypothesis that the initial salary difference is the result of discriminatory behavior is not supported by the data.”
Id.
at ¶ 12.
It appears to the Court that the Consent Decree anticipated this issue. Section IV. D.l of the Consent Decree states that the compensation study “shall consider, if appropriate, the experience, prior salary ... of each employee.” The Court finds it unlikely that if the parties considered prior salary or initial salary to be a
per se
“tainted variable,” this reference would have been included in the Consent Decree.
B. Band and Zone
All Amtrak employees fall into a band on the compensation scale. The higher an employee’s band, the higher the employee’s salary. Plaintiffs contend that controlling for “band and zone” was inappropriate.
During the first two years of the Consent Decree, Amtrak reduced the number of job titles it used and reassigned job titles to new bands and zones. Currently, most of the employees with the same job title are in the same band. Plaintiffs claim that controlling for band and zone for these employees is superfluous. However, plaintiffs claim that where job titles are assigned to multiple bands, it may be because there is a difference in responsibility or it may be a product of racial discrimination.
Dr. Pollner believes that, in this case, band and zone is a statistically appropriate variable. Def. Mot.App. B, ¶ 7. Dr. Poll-ner explains that band and zone is not a tainted variable because most employees with the same job title are assigned to the same band and this variable is a significant predictor of salary. Id. at ¶ 8. Further, defendant contends that job titles in different bands and zones reflect a clear difference in job duties, not discrimination. Finally, defendant notes that because the Consent Decree requires comparisons of employees in “comparable positions” and comparable positions are defined in terms of band and zone, it is necessary to include band and zone as a factor in the compensation study. Def Mot-App. A, Consent Decree. Dr. Pollner concluded that “Because it is necessary to compare similarly-situated employees, it is my opinion that the inclusion of band/zone is critical.” Def. MotApp. B, ¶ 10.
While plaintiffs attempt to argue that initial salary and band and zone are not “appropriate” in this case, the Court finds that the four corners of the Consent Decree delegated considerable discretion to the outside consultant in conducting a compensation analysis using “generally accepted factors relied on by compensation specialists.” Def. Mot-App. A. Consent Decree § IV.D.l. Although plaintiffs disagree with some of Dr. Pollner’s decisions regarding the analysis, they have not alleged that her analysis incorporates variables that are outside the generally accepted factors.
Class Counsel may well be correct that Initial Salary and Band and Zone have been found to be inappropriate in other contexts. However, here, the consultant considered Class Counsel’s position and decided against it. Because the parties contracted to put the discretion in a third party’s hands, the Court will not upset Dr. Pollner’s decision to use initial salary and band and zone absent an allegation that a variable used was wholly outside of what is generally acceptable.
Thus, the Court DENIES IN PART the Motion for Enforcement of the Consent Decree.
2. Dr. Pollner failed to account for race in calculating the difference between predicted and actual salaries for each individual.
When Dr. Pollner determined that African American employees were paid less than their white counterparts, she controlled for race. However, when Dr.
Pollner determined the amount by which each individual was paid more or less than predicted, she did not take race into account. Plaintiffs claim that her omission of race as a variable in the second calculation was in error.
By omitting race as a variable, Dr. Poll-ner did not, as the plain language of the Consent Decree requires, compare the salaries of African American managers with white managers in the same position. Instead, she compared the salaries of African American management employees with the salaries of
all
management employees, including other African Americans.
Plaintiffs claim that Dr. Pollner should have run a regression analysis on only the non-African American employees to determine values for each independent variable. Dr. Pollner could then have compared those values to the variables of African American employees to determine predicted salary. The methodology suggested by plaintiffs would have resulted in predicted salaries that are $130,000 greater than what Dr. Pollner’s analysis provided. By including African Americans in the comparison group, who are paid less than then-white counterparts, Dr. Pollner lowered the predicted salaries for African Americans and reduced the disparities between actual salary and predicted salary. Pis. Mot Ex. C. ¶ 17(b).
In response, defendant argues that Dr. Pollner controlled for factors related to base salary across all management employees in order to identify similarly situated employees and then conducted a regression analysis of pay disparities between African American employees and Caucasian employees. Def. Mot.App. B ¶ 6. Defendant notes that, as explained in Dr. Pollner’s Declaration, the problem with plaintiffs’ approach is that it assumed that all the statistical variables for African American and Caucasian employees are identical.
Id.
at ¶ 19. Defendant attempts to dispute plaintiffs’ argument by alleging the Dr. Pollner used a “statistically acceptable and commonly used approach.”
The Court finds that the plain meaning of the Consent Decree requires that a comparison be made between African American employees and their white counterparts. To the extent that a comparison was made between African American employees and all employees, the comparison must be corrected to reflect the plain meaning of the Consent Decree. While other methods of comparison may be commonly used, a comparison between African American employees and their white counterparts is the method of comparison that the parties agreed to in the Consent Decree. To the extent that this has not been done, the Court GRANTS IN PART the Motion to Enforce the Consent Decree.
C. The compensation adjustment need not reflect that some class members are paid more than predicted, but should reflect the delay in payment and Amtrak’s partial adjustment to the 28 class members.
Plaintiffs aver that 308 African American employees are underpaid a total of $1,289,000. The remaining 119 African American managers are paid $842,000 more than their predicted salaries. Thus, the total compensation disparity is $2,131,000.
Plaintiffs note that the adjustment should take into account the fact that 28% of class members are earning more than their predicted salary. Plain
tiffs posit that since there will be no reductions in salary, for African Americans as a group not to be “overpaid,” the total amount of adjustment should be $1,289,000 (or $842,000 less than the total disparity of $2,131,000.)
Plaintiffs argue that the fairest way of taking “overpayments” into account is to reduce the compensation adjustment of each underpaid African American employee by the percentage of overpayment to other African American employees. Thus, according to plaintiffs, the salary of each underpaid African American employee should be increased only by a percentage of which the salary fell short of the predicted amount.
Plaintiffs note that the adjustments were due October 1, 2002. Plaintiffs assert that Amtrak should be required to make a lump sum payment for each employee paid less as of February 15, 2002. The lump sum should include an appropriate interest adjustment. Plaintiffs note that the lump sum payments to the 28 individuals whose salaries have been increased should be credited by the amount they have been paid.
The Court finds that the calculation method suggested by Class Counsel was not contemplated by the Consent Decree and will not be accepted by the Court. Moreover, this method leaves many African American employees with less than their “predicted salary” and fails to “eliminate racial disparities.”
The Court finds that lump sum payments and an interest adjustment are appropriate. The Court will direct that credit be given for any adjustments already made to the salaries of some class members.
IV. Conclusion
In sum, the four corners of this Consent Decree require that an independent expert of defendant’s choice perform a compensation analysis. While plaintiffs were afforded an opportunity to consult with the expert, the Consent Decree gave the expert considerable discretion in determining the appropriate variables for the study. The expert exercised her discretion in choosing to use initial salary and salary band and zone. The Court will not upset her decision.
However, the Court will direct that the plain meaning of the Consent Decree be followed. Each African American employee who is currently paid less than her “predicted salary,” as determined by comparing her current salary to the current salary of white managers in the same position, shall have her salary adjusted to make it equal to her “predicted salary,” not merely to the point where the difference is no longer statistically significant. Thus, plaintiffs’ Motion to Enforce the Consent Decree will be GRANTED IN PART and DENIED IN PART.
Finally, defendant’s uncontested Motion to File its Response Under Seal will be GRANTED.
An appropriate Order accompanies this Memorandum Opinion.
ORDER
Pursuant to the Memorandum Opinion entered this same day, it is by the Court hereby
ORDERED that plaintiffs’ Motion to Enforce the Consent Decree is GRANTED IN PART and DENIED IN PART; and it is
FURTHER ORDERED that defendant’s uncontested Motion to File its Response Under Seal is GRANTED.