Uscc comments to ofccp nprm on statistical evidence


January 29, 2020


Mr. Craig E. Leen
Room C-3325
Office of Federal Contract Compliance Programs
200 Constitution Avenue, NW
Washington, DC 20210

By Electronic Submission:

RE: RIN 1250-AA10 - Nondiscrimination Obligations of Federal Contractors and Subcontractors, 84 Fed. Reg. 71875 (December 30, 2019)

Dear Director Leen:

The U.S. Chamber of Commerce (“the Chamber”) submits the following comments in response to the Office of Federal Contract Compliance Programs (“OFCCPs” or “the Agency”) Notice of Proposed Rulemaking Rule regarding Nondiscrimination Obligations of Federal Contractors and Subcontractors (“NPRM” or “the proposed rule”) published in the Federal Register on December 30, 2019 (84 Fed. Reg. 71875, December 30, 2019). Among the Chamber’s members are many federal contractors subject to OFCCP’s jurisdiction who will be directly affected by the outcome of this rulemaking.


As an initial matter, the Chamber recognizes the Agency’s recent efforts, to promote transparency regarding its compliance evaluation procedures and processes. The Chamber also acknowledges that the Agency’s stated goal with regard to this NPRM is to provide some measure of regulatory certainty regarding the Predetermination Notice (“PDN”) and Notice of Violation (“NOV”) process and the statistical evidence on which it will rely when pursuing claims of systemic discrimination. All too often, OFCCP has relied on sub-regulatory guidance, rather than formal notice and comment rulemaking, to establish its governing processes and procedures. The Chamber shares the Agency’s interest in providing government contractors greater certainty and transparency with respect to its PDN and NOV process.

Notwithstanding the laudatory goals of the NPRM, the Chamber has significant concerns with the proposed rulemaking. The central thrust of the NPRM is the Agency’s treatment of statistical and non-statistical proof in its assessment of potential systemic discrimination in compensation and other personnel practices. While the Chamber supports certain elements of the NPRM, including its appropriate recognition that variances must be both statistically significant and practically significant, the Chamber has substantial concerns about much of the NPRM’s treatment of statistical and non-statistical evidence. The Agency should not adopt a formulaic approach to statistical evidence, particularly in the context of compensation. OFCCP should continue to recognize that its statistical models: (1) must be tailored to each contractor’s particular compensation or personnel practices; (2) must be based on groupings of similarly situated individuals (in accordance with Title VII ); and (3) must account for all factors that impact the dependent variable. Any final rule should continue to require that any finding of discrimination must be supported by evidence that the variance at issue is practically significant as well as statistically significant, and accompanied by non-statistical evidence.

The Chamber believes the NPRM has the following significant problems:

  • The NPRM’s proposed declaration that non-statistical evidence is necessary only when a statistical variance is less than three standard deviations is arbitrary and without support in Title VII case law. Case law analyzing the disparate treatment/pattern and practice framework requires the showing of non-statistical evidence in all except the most extreme cases, such as the “inexorable zero.” Accordingly, the Agency should refrain from establishing a statistical threshold above which it can rely solely on statistical evidence in support of a finding of discrimination. Such an assertion of policy, devoid of support in Title VII case law, is beyond the scope of the Agency’s authority.
  • The arbitrary statistical thresholds of “discrimination” the Agency attempts to establish will work to the unfair disadvantage of large employers where applicant pools or incumbent pools are large.

Further, the Agency has, in practice, improperly aggregated across different populations and multiple years to generate larger statistical variances so that they are statistically significant but not necessarily practically significant; such data manipulation creates false presentations of asserted discrimination. Issuing a PDN without non-statistical evidence of discrimination, based only on a statistical variance of at least three standard deviations, will only further encourage improper aggregations and exacerbate this pattern. Without any supporting evidence of discriminatory actions mere statistical variances cannot suffice as support for a PDN.

  • The NPRM does not expressly limit its application to matters involving OFCCP claims of disparate treatment/pattern and practice discrimination. Any final rule should state that it does not extend to matters arising under the disparate impact theory of discrimination.

With regard to the proposed codification of the PDN process, the Chamber appreciates OFCCP’s efforts to codify its sub-regulatory guidance on PDNs, but believes the Agency should go a step further in the interests of efficiency and transparency. The Agency should require a pre-PDN conference with the contractor to address all alleged violations and permit the contractor an opportunity to respond before the PDN issues. Postponing this exchange until after the PDN issues, and requiring contractor responses within 15 days, is an unworkable and inadvisable approach. Moreover, a 15-day period to respond if a PDN is issued is unreasonably short and should be extended to 45 days to provide for meaningful review, reflection, and an opportunity to understand and resolve a PDN.

Finally, the Agency should expressly endorse use of the Early Resolution Process (ERP) and Early Resolution Conciliation Agreements (ERCAs) in its final rule and codify the process. The NPRM and proposed regulatory language fail to adequately codify ERCAs as a meaningful path for resolving audits.


The Agency Should Not Set a Statistical Threshold at Which it Will Proceed to a PDN or NOV Without Supporting Non-Statistical (Anecdotal) Evidence of Discrimination.

In the NPRM, OFCCP provides that non-statistical evidence of discrimination is not necessary when a statistical disparity is greater than three standard deviations. In support of its bright-line “numbers evidence without more” test, OFCCP claims its position is “in keeping” with Title VII case law. On this issue, OFCCP is wrong.

While in rare instances, such as the existence of an “inexorable zero”1 of successful minority candidates or incumbents, there may be a sufficient basis to establish a prima facie case of discrimination, without non-statistical evidence, there is simply no support for moving forward with a PDN or NOV based solely on an arbitrary statistical threshold of more than three standard deviations. OFCCP’s position finds no support in existing case law and demonstrates a deep misapplication of the meaning of statistical significance as it relates to determining whether employment discrimination has occurred.

As explained by Labor Economist Dr. J. Michael DuMond, a disparity designated as “statistically significant” suggests only that the observed disparity is unlikely to have been the result of a random occurrence. (Declaration of Dr. DuMond (“DuMond Decl.”), ¶ 14.)2 However, a statistically significant disparity does not automatically mean that the reason for the identified disparity is discrimination. Id. ¶ 14, 17. Dr. DuMond’s declaration is backed by a treatise OFCCP cites on its own public website, which explains that “significant differences may be evidence that something besides random error is at work. They are not evidence that this something is legally or practically important.”3

  1. A statistical situation characterized by the absence, or near absence, of any examples to counter the inference of discrimination.
  2. Dr. DuMond’s declaration is attached as Exhibit 1.
  3. D. Kaye & D. Freedman, Washington, D.C.: National Academy of Sciences Press, Reference Guide on Statistics (2011), available at (last visited Jan. 24, 2020); Office of Federal Contractor Compliance, Practical Significance in EEO Analysis, Frequently Asked Questions, available at (last visited Jan. 24, 2020).

The statements regarding the limits of what a statistically significant finding means (and what it does not mean) are supported by Title VII case law. See Carpenter v. Boeing Co., 456 F.3d 1183, 1195 (10th Cir. 2006) (citing Hazelwood Sch. Dist. v. United States, 433 U.S. 299, 308 n.14 (1977)) (reasoning that statistical significance only rules out “chance” as an explanation for observed disparities in an employer’s workforce and allows a court to infer that “something” must be causing these statistical anomalies). However, statistical significance does not, on its own, establish that “the something” is the challenged practice under consideration. Foxworth v. Police, 228 Fed. App’x 151, 156 (3rd Cir. 2007) (agreeing with district court that statistics “showed a disparity” but were “insufficient as a matter of law to prove that the automatic disqualification policy caused it”); Carpenter, 456 F.3d at 1202 (noting that while “something in the overtime process consistently results in males obtaining more overtime” the “large number of standard deviations tells us nothing about what that ‘something’ is[.]”).

Rather, in intentional discrimination cases, the Supreme Court and other Courts have relied on evidence of statistical disparity, but have also required more as a way of bringing the “cold numbers convincingly to life” through anecdotal and other evidence. Int’l Bhd. Of Teamsters v. U.S., 431 U.S. 324, 329-30, 342-43 (1977). To hold otherwise would result in employers being held potentially liable, not for the alleged discriminatory practice, but for “the myriad of innocent causes that may lead to statistical imbalances in the composition of their workforces.” Watson v. Fort Worth Bank and Trust, 487 U.S. 977, 992 (1988); Wards Cove Packing Company, Inc. v. Atonio, 490 U.S. 642, 657 (1989).

In fact, the cases cited by OFCCP do not provide the sweeping support for OFCCP’s reliance on mere statistical evidence to establish discrimination. For instance, in Adams v. Ameritech Servs., Inc., the court did not suggest that statistical evidence could establish discrimination on its own, but instead observed that while “statistical evidence can be very useful to prove discrimination… it will likely not be sufficient in itself” and reached its conclusion by relying on the fact that additional, non-statistical evidence had been introduced. 231 F.3d 414, 423-24, 427 (7th Cir. 2000) (“If [the statistical analysis] on its own is all the plaintiffs’ had introduced, we would agree … that the record would have supported summary judgement against them. It was not, however, and in our view, the other items of evidence if believed by the jury could have done the rest of the job”).

OFCCP’s reference to Malave v. Potter—a disparate impact, not disparate treatment, case—does not lend support for its position that it may proceed without non-statistical evidence. 320 F.3d 321 (2nd Cir. 2003). As noted in the NPRM, the Court did remand the case and instruct the district court to “determine whether the plaintiff can show a statistically significant disparity of two standard deviations.” Id. at 327. But it said more. It also said that the District Court must decide whether plaintiff’s “statistical analysis and other evidence, is of a kind and degree sufficient to show” the promotion practices was the cause of the disparity. Id. (emphasis added). It also emphasized that “no bright line rules exist to guide courts in deciding whether [a] plaintiff['s] statistics raise an inference of discrimination,” and offered some “overarching principles [that] inform the issue.” Id. at 325-326. Among these principles was a caution that any statistics relied upon “must be of a kind and degree sufficient to reveal a causal relationship between the challenged practice and the disparity” when combined with other evidence. Id.

Rather than lending support for OFCCP’s position, the body of Title VII case law makes clear that context matters and that there is no bright line threshold that can be applied under which non-statistical evidence is no longer necessary. Palmer v. Shultz, 815 F.2d 84, 92 (D.C. Cir. 1987) (noting that Supreme Court has not established “an exact legal threshold at which statistical evidence, standing alone, establishes an inference of discrimination”). Rather, the Chamber urges the Agency to follow Title VII principles and refrain from setting arbitrary thresholds that apply to all findings. 4 See also, Chin v. Port Auth. of New York & New Jersey, 685 F.3d 135, 153 (2d Cir. 2012) (“no minimum statistical threshold requiring a mandatory finding [of] … a violation of Title VII. Courts should take a ‘case-by-case approach’ in judging the significance or substantiality of disparities, one that considers not only statistics but also all the surrounding facts and circumstances.”); Watson, 487 U.S. at 996 n. 3, 108 S.Ct. 2777 (approving the case-by-case approach because “statistics ‘come in infinite variety and ... their usefulness depends on all of the surrounding facts and circumstances.’”) (quoting Teamsters v. United States, 431 U.S. 324, 340, 97 S.Ct. 1843, 52 L.Ed.2d 396 (1977)). These cases demonstrate the caution expressed by courts against bright line rules at which statistical evidence, standing alone, establishes an inference of discrimination.

a. Nonstatistical Evidence is Required in All But Extreme Cases Which Generally Requires More than Simply a Finding of More than Three Standard Deviations.

Although the Chamber strongly opposes a bright line test that applies a “no non-statistical evidence requirement” if there is a finding of three standard deviations or more, it acknowledges that there are rare exceptions that do not require non-statistical evidence. For instance, the Supreme Court has recognized that in certain extreme cases, exemplified by the presence of the “inexorable zero”—i.e., the total or near total absence of members of a protected group—it may be appropriate to reach an inference of discrimination, without additional non-statistical evidence. The basis for the Supreme Court’s reasoning was that no amount of “fine tuning” could have overcome the “glaring absence” of minority drivers. Teamsters v. United States, 431 U.S. at 342.

However, even in those cases, courts have often relied on additional non-statistical evidence. See Johnson v. Transp. Agency, 480 U.S. 616, 665 n.4 (1987) (Scalia, J., dissenting) (“[E]ven an absolute zero is not ‘inexorable.’ While it may inexorably provide ‘firm basis’ for belief in the mind of an outside observer [that discrimination occurred], it cannot conclusively establish such a belief on the employer’s part, since he may be aware of the particular reasons that account for the zero.”)

Additionally, not only do most of the inexorable zero cases also rely on non-statistical anecdotal evidence, those extreme cases involve statistical significance that far exceed three standard deviations. See Teamsters, 431 U.S. at 342 (even in the face of evidence that no Black employees were ever employed on a regular basis, it still relied on anecdotal evidence);

See also, D. Baldus & J. Cole, supra, § 9.4, at 188–89 (Supp.1987) (courts should use tests of statistical significance only as “an aid to interpretation” and not as a “rule of law”); Tagatz v. Marquette Univ., 861 F.2d 1040, 1044 (7th Cir.1988) (“Correlation is not causation.”); Chin v. Port Auth. of New York & New Jersey, 685 F.3d 135, 153 (2d Cir. 2012) (“no minimum statistical threshold requiring a mandatory finding [of] … a violation of Title VII. Courts should take a ‘case-by-case approach’ in judging the significance or substantiality of disparities, one that considers not only statistics but also all the surrounding facts and circumstances.”).

Calloway v. Westinghouse Elec. Corp., 642 F.Supp. 663 (M.D. Geo. 1986) (finding there were no black managers or supervisors over a 14 year period, but still relied on anecdotal evidence); EEOC v. Boeing Co., 577 F.3d 1044 (9th Cir. 2009) (applying inexorable zero where plaintiff was the only woman in her skill code who was laid off while every male employee identified for termination in all three RIFs ultimately remained, but also relied on anecdotal witness testimony); EEOC v. O&G Spring and Wire Forms Specialty Co., 38 F.3d 872 (7th Cir. 1994) (applying inexorable zero where there were no black hires over a 6 year period, but still relied on anecdotal evidence); Grant v. Bethlehem Steel Corp., 635 F.2d 1007, 1016–17 (2d Cir. 1980) (applying inexorable zero where defendant rejected applications by all three appellants for foremen’s jobs, but still relied on anecdotal evidence that less qualified white applicants were appointed to such positions).

These cases demonstrate that while Courts do recognize that in some instances when the number of individuals in a group is zero, or near zero, that statistical information alone may be sufficient to support an inference of discrimination, there is often non-statistical evidence that supports the data and brings the “cold numbers convincingly to life.” Teamsters, 431 U.S. at 329-30, 342-43 (1977).

b. OFCCP’s Proposed “Greater Than Three Standard Deviation” Framework, By Design, Will Disproportionally Lead to More Findings of Discrimination.

Employers with large workforces in particular, will be disproportionally impacted with findings of discrimination if OFCCP proceeds with the NPRM as drafted. This is so because such employers generally have larger applicant or incumbent pools, and even minor discrepancies can be considered statistically significant when the difference is measured across a large number of observations. (DuMond Decl. ¶ 13.) Relatedly, any aggregation of data can lead to results that are statistically significant, again simply because of larger pool sizes. (DuMond Decl. ¶ 10). Dr. DuMond offers a table to illustrate the point: small average pay disparities for females (under 1%) in two groups are not statistically significant when considered separately, but reach statistical significance based solely on the larger sample size:

Analysis group chart

(DuMond Decl. ¶¶ 12-13.) As the table shows, simply combining the pools and increasing the number of observations analyzed, caused a significant jump in the level of significance, such that the variance went from not statistically significant to over three standard deviations. Yet under each grouping (A, B or A&B), the adjusted pay difference remains relatively the same. Thus, the driver of the statistically significant finding has nothing to do with alleged discriminatory decision-making, rather, it is being driven by applying the same percentage variance to a larger population.5 And in this example, the level of variance surpasses OFCCP’s proposed threshold of more than three standard deviations, such that OFCCP would proceed to an PDN with no additional supporting non-statistical evidence. (Id. ¶ 11.) It is nonsensical to allow mere aggregation or large sample sizes to be the driver of moving from statistical non-significance to significance.

OFCCP can still accomplish its stated desire of greater transparency and consistency without setting the arbitrary thresholds contemplated here. To achieve this desired result, OFCCP should instead consider clearly stating that: (1) the Agency will mirror a contractor’s compensation system or other personnel practice when developing its statistical models; (2) the Agency will compare only individuals who are similarly situated (pursuant to Title VII standards); and (3) the appropriate statistical methodology to be applied in each matter will depend on a number of factors, including the size of the comparison groups at issue, the personnel practice at issue, the variables impacting the personnel practice at issue, and the availability of data to perform appropriate statistical analyses.

Accordingly, the Chamber strongly urges OFCCP to withdraw its position that it will move forward with a PDN, based solely on a statistical finding of more than three standard deviations.

As Dr. Dumond notes in his Declaration, aggregating across subgroups also carries the risk of ignoring the fact that variables in a regression may impact the various subgroups differently, such that aggregation is improper. (DuMond Decl. ¶ 8.) As Dr. DuMond notes, economists and statisticians can rely on a statistical test (i.e., a Chow test) to determine if the pay of subsets of employees is determined not only by the same factors (e.g., experience, education), but also whether the relative weights of those factors are similar among the subsets. When the Chow test fails, the conclusion that follows is that it is improper to aggregate the subgroups into a combined analysis.

The Agency Properly States that Variances Must Be “Practically Significant” to Warrant A PDN or NOV, but OFCCP Should Provide Further Explanation of this Key Concept.

The NPRM states that OFCCP shall issue a PDN only after “first considering” whether “the unexplained [statistical] disparity is both practically and statistically significant.” 84 Fed. Reg. at 71885. The preamble explains further that to “determine whether the evidence of discrimination is sufficient to warrant a PDN, OFCCP considers whether an employment or compensation disparity identified during the compliance evaluation is both practically and statistically significant.” Id. at 71876. The preamble then explains in footnote six that “practical significance refers to whether an observed disparity in employment opportunities or outcomes reflects meaningful harm to the disfavored group.” Id. at 71876, n.6. (emphasis added) - Citing OFCCP FAQs at

The Chamber endorses OFCCP’s decision to incorporate the key concept of “practical significance” into its regulations. For several reasons, the Chamber urges OFCCP, in its final rule, to state affirmatively that any alleged disparity must be practically significant for OFCCP to move to a PDN with regard to the employment practice at issue.

Practical significance is not a novel concept. It is expressly referenced in the Uniform Guidelines on Employee Selection Procedures (UGESP). 41 C.F.R. Part 60-3 (1978). Section 60-3.4D discusses the “four-fifths rule” and states that “[s]maller differences in selection rate may nevertheless constitute adverse impact, where they are significant in both statistical and practical terms.” 41 C.F.R. Section 60-3.4. Indeed, labor economists agree that the four-fifths rule, itself, is a practical significance concept. See “Measuring Practical Significance in Adverse Impact Analyses”, Frederick L. Oswald, Eric M. Dunleavy, and Amy Shaw (Adverse Impact Analysis: Understanding Data, Statistics, and Risk. Routledge 2017).

Practical significance has likewise been recognized by the courts and by OFCCP. See, e.g., Waisome v. Port Authority, 948 F.2d 1370, 1376 (2d Cir. 1991) (finding no discrimination where variance in passing scores on written test was statistically-significant, but selection rates passed the four-fifths test and the disparity, though statistically significant, “was of limited magnitude”); Apsley v. Boeing, 691 F.3d 1184 (10th Cir. 2012) (finding no discrimination where variance in selection rates for reduction in force was statistically significant, but the shortfall was small in relation to the total number of employees in the analysis, such that selection rates were “practically nonsignificant”); Contreras v. City of Los Angeles, 656 F.2d 1267 (9th Cir. 1981) (no discrimination notwithstanding statistically-significant difference in selection rates where difference was approximately 2%).

The concept of practical significance and its relevance to OFCCP determinations can be readily illustrated by just a few examples. The table below shows the criticality of practical significance whenever the incumbent pool or applicant pool at issue is large. For this example, we show that even a 1% difference in selection rates can be statistically significant when the sample sizes reach a certain threshold. Without consideration of practical significance (i.e., the difference in selection rates is just 1%), OFCCP could assert that the statistically significant result when the sample size reaches 8,400 evinces discrimination; it does not. Moving to a PDN based solely on statistical significance in this example would be wholly unwarranted.

Ofccp chart 2

The next table shows a similar result in the compensation context, where very small differences in average pay to women and men yield a variance that is statistically significant, but not practically significant. For instance, in Case 1 below, a practically insignificant difference in pay – just 0.5% – yields a variance of 2.04 standard deviations when the pool of incumbents compared reaches just over 1,500.

Ofccp chart 3

In none of the cases identified in the second table should OFCCP advance claims of discrimination in pay, given the practical insignificance of the pay differences. Without examining the data through that lens, and if OFCCP were to rely solely on statistical significance, the contractor in each of the three scenarios would be faced with a PDN alleging discrimination in pay.

These tables also reflect the reality that, unless the Agency is required to show practical significance in each instance in which it pursues a PDN, large contractors will be disadvantaged. See DuMond Declaration at ¶13. Large contractors with thousands or tens of thousands of applicants for jobs in a single job group each Affirmative Action Plan year, or with tens of thousands of employees in their AAPs for purposes of compensation analyses, will be much more likely than small contractors to have some areas of statistically-significant differences. In

  1. Notably, under OFCCP’s proposed rule, the Agency could move to a PDN on this variance – 3.10 standard deviations – without non-statistical evidence of discrimination. This example underscores the frailty of that approach, as well.


those cases, requiring a showing of practical significance is critical to ensuring that large contractors are not disadvantaged or unfairly identified for enforcement actions.

As such, the appropriate measure of practical significance will vary from audit to audit and from one employment practice to another. Because of this variety, how OFCCP applies practical significance should be one of the issues addressed in negotiations between the Agency and the employer.7 As discussed more thoroughly below, the ideal opportunity for this would be a pre-PDN conference.

The Proposed Rule Has No Application Over Disparate Impact Claims And Should Be Explicitly Limited to Disparate Treatment Cases.

The NPRM is ambiguous on the question of whether the proposed treatment of statistical evidence will apply only to Title VII claims of disparate treatment (i.e., intentional discrimination), or if disparate impact claims, which encompass unintentional discrimination, fall within the scope of the proposed rule. E.E.O.C. v. Joe’s Stone Crab, Inc., 220 F.3d 1263, 1278 (11th Cir. 2000) (“As we have noted, the central difference between disparate treatment and disparate impact claims is that disparate treatment requires a showing of discriminatory intent and disparate impact does not.”)

For instance, the resolution procedures that would be adopted under the NPRM provide that a PDN would be issued only after a finding of practical and statistical significance and, where relevant, the non-statistical evidence demonstrates “an intent to discriminate.” 84 Fed. Reg. No. 249, 71875 at 71885 (Dec. 30, 2019). (emphasis added) Similarly, in the background section of the NPRM, the OFCCP provides that “non-statistical evidence, such as a cohort analysis, demonstrates an intent to discriminate.” Id. at 718777. (emphasis added)

Disparate impact claims require showings of both statistical evidence that a neutral policy or practice has a substantial adverse impact on a protected group and the demonstration of a causal relationship between the employment policy or practice at issue and the statistical disparity. See Wards Cove Packing Co. v. Antonio, 490 U.S. 642, 657 (1989) (plaintiffs must show that the challenged practice had a “significantly disparate impact on employment opportunities for whites and nonwhites.”); Watson v. Fort Worth Bank & Trust, 487 U.S. 977, 995 (1988) (disparities must be “sufficiently substantial that they raise such an inference of causation.”).

The NPRM makes no mention of the types of proofs implicated in disparate impact cases, namely the required explicit identification of a neutral policy or practice, nor does it describe the process by which employers can demonstrate the business necessity for the challenged practice.

The absence of this type of language as it relates to disparate impact claims, creates the impression that the Agency intends to extend the application of the NPRM only to intentional conduct, i.e. disparate treatment cases. Accordingly, OFCCP should clarify that the NPRM applies only to intentional discrimination claims.

As discussed at length in the literature, there are a number of potential measures of practical significance, including impact ratios (the four-fifths rule), phi coefficient, odds ratios, absolute differences in selection rates, Cohen h-statistic, and shortfall statistics.

The Agency Should Add a New Step in the Case Resolution Procedure that Will Precede a PDN: the Pre-PDN Conference.

In September 2017, the Chamber issued “Office of Federal Contract Compliance Programs: Right Mission, Wrong Tactics, Recommendations for Reform,” a white paper detailing the Chamber’s concerns about OFCCP’s enforcement efforts and its relationship with the contractor community. The Chamber highlighted the Agency’s lack of transparency, its perceived antagonism towards contractors, its drawn-out, overly invasive compliance review practices, and its aggressive conciliation posture.

The Chamber is encouraged by a change in OFCCP’s tone. In recent years, the Agency has repeatedly expressed a desire to work with contractors to identify and resolve issues in a timely manner. In furtherance of these goals, the Agency has issued a number of guidance documents, including Directive 2018-01 Use of Predetermination Notices (PDN), Directive 2018-08 Transparency in OFCCP Compliance Activities, and Directive 2019-01 Compliance Review Procedures. Contractors have also seen changes in some of the regional offices, where compliance officers have struck a more collaborative tone in audits, and have sought to close or resolve compliance reviews in a more efficient manner. While these changes in policy and practice have been well received by the contracting community, there is more that can and should be done to improve the Agency’s effectiveness and efficiency, and OFCCP should take the opportunity in its proposed regulatory changes to do just that.

Directive 2018-01 (Use of Predetermination Notices (PDN)) mandates that OFCCP issue a PDN and provide the contractor at least fifteen (15) days to rebut OFCCP’s proposed findings of discrimination prior to issuing an NOV. The Directive also requires that both the regional Office of the Solicitor and OFCCP’s national office review and approve all PDNs before they are issued. The NPRM codifies in the regulations this requirement to issue a PDN prior to issuing an NOV, stating “for discrimination violations, OFCCP may issue the notice of violation following issuance of a predetermination notice if the contractor does not respond or provide a sufficient response within 15 calendar days of receipt of the notice, unless OFCCP has extended the predetermination notice response time for good cause shown.”

The Chamber appreciates that OFCCP is seeking to codify the PDN requirement in the regulations. However, this action does not go far enough to address OFCCP’s stated desire to improve the efficiency of the compliance review process and the collaboration between contractors and the Agency. Under the proposed structure, a contractor has no opportunity to address potential areas of discrimination until the Agency (1) has made preliminary findings, (2) has those findings reviewed and approved by the Regional Office of the Solicitor, and then (3) has the findings reviewed and approved by the National Office of OFCCP. As such, the Agency is required to expend a significant amount of time and effort preparing, reviewing, and approving the preliminary findings before the contractor is even made aware of the potential issues and given an opportunity to respond. This is inefficient and inconsistent with the Agency’s stated interest in collaboration.

Consistent with the Agency’s desire to seek early resolution of potential findings of non-compliance or discrimination, the Chamber proposes that OFCCP add to the regulations a new step in the compliance review process that would precede issuance of a PDN. This step would

involve a meeting between OFCCP and the contractor to discuss, in detail, all potential issues of non-compliance or discrimination. OFCCP would share its statistical and non-statistical evidence in that meeting, and how it has applied the requirement for practical significance, and the contractor would be given an opportunity to respond and provide additional information, before OFCCP issues a PDN.

If the contractor is not able to address the potential issues to OFCCP’s satisfaction, the parties can attempt to resolve the matter at this stage using the expedited conciliation option contemplated by the NPRM, discussed further below. If the parties are unable to reach a resolution, the Agency can proceed with the PDN process. Adding this additional step in the process will enable OFCCP and contractors to collaborate on an informal basis to reach understanding/resolution before OFCCP expends the time and effort of drafting a PDN and obtaining the necessary approvals to issue.

Finally, the Chamber believes contractors should be provided a period longer than 15 days to respond to a PDN. In nearly every case, and in complex cases in particular, 15 days is a period too short to permit a meaningful response. Rather than requiring contractors to routinely seek extensions of time to respond, OFCCP should modify this time period to 45 days and continue to permit requests for additional time.

OFCCP Should Formally Incorporate the Early Resolution Procedures and Early Resolution Conciliation Agreements in its Substantive Regulations.

One of the more effective actions taken by OFCCP in recent years to improve the efficiency and effectiveness of its enforcement efforts was its issuance of Directive 2019-02 – Early Resolution Procedures (“ERP”). The Agency’s ERP Directive offers contractors and the Agency a path to resolve purported violations identified by the agency during the audit through conciliation prior to the issuance of a NOV. It also permits the parties to agree on a broader resolution to address any similar alleged violations at a subset, or at all, of the contractor’s other establishments. The ERP Directive established the Early Resolution Conciliation Agreement (ERCA). Pursuant to an ERCA, a contractor agrees to remedy the Agency’s alleged violations at the establishment under review and agrees to review all, or a subset, of its other establishments for similar violations and implement corrective action as appropriate. The contractor is required to submit progress reports to the Agency for five years regarding its remediation of the alleged violations at all covered establishments. In recognition of the contractor’s willingness to engage in these proactive corrective measures, OFCCP agrees to not audit the establishments encompassed by the ERCA for seven years (the five-year ERCA reporting period plus two years after the end of the reporting period).

The Agency first began utilizing the ERP and ERCAs in 2019, and made effective use of ERCAs at the end of Fiscal Year 2019 to resolve compliance evaluations that had been open for years and for which OFCCP sought significant back pay amounts. Multiple companies, large and small alike, entered into ERCAs that resolved numerous long-standing audits; some resulted in multi-million dollar settlements. Each contractor that entered into an ERCA committed to an array of remediation and reporting obligations, and was given a seven-year reprieve from audit to effectuate these changes.

The NPRM references the ERP and ERCAs but does not sufficiently codify in the regulation the essential elements of these resolution mechanisms. The NPRM adds the following language to the regulations: “Expedited Conciliation Option. A contractor may waive the procedures set forth in paragraphs (a) [pre-determination notice] and/or (b) [notice of violations] of this section to enter directly into a conciliation agreement.” 84 Fed. Reg. at 71885. This proposed language does not explain the process by which the Agency and contractor can negotiate such an early resolution and it does not articulate the opportunity for the Agency and the contractor to reach a resolution that encompasses more than just the establishment(s) currently under review. And, of key importance, nowhere do the proposed regulations codify the seven-year audit moratorium that is at the heart of the ERP and ERCA process.

The NPRM codifies the PDN and NOV processes in some significant detail, enumerating the number of days a contractor has to respond to a PDN and mandating that the Agency cannot issue an NOV until it has gone through the PDN process. The final regulations should address the ERP and ERCAs in a similar manner, detailing the approach and codifying the option of a multi-establishment or corporate-wide resolution and a seven-year audit moratorium.


The proposed regulation has some commendable provisions. However, there are several very troubling components to the NPRM. OFCCP’s intent to rely solely on statistical evidence when it produces a variance of more than three standard deviations is an invitation to manipulate the data to find discrimination in the absence of any other factors such as non-statistical evidence. This provision should be reconsidered and withdrawn. Further, while stating that statistical variances must be “practically significant” as well as statistically significant, there is much uncertainty about what “practically significant” means. Finally, OFCCP should clarify that the processes identified in this regulation will not be used to support cases based on a “disparate impact” theory, and will be restricted to cases based on “disparate treatment.”


Marc Freedman

Vice President, Workplace Policy

Employment Policy Division

Outside Counsel

Camille Olson

Annette Tyman

Kris D. Meade

Rebecca Springer

Seyfarth Shaw LLP

975 F Street, N.W.

Washington, DC 20004

Crowell & Moring LLP

1001 Pennsylvania Avenue, N.W.

Washington, DC 20004

Uscc comments to ofccp nprm on statistical evidence