Discrimination cases are essentially “games” of making relevant, suitable, or valid comparisons between suitable, valid, or appropriate comparator(s). While anti-discrimination laws do not define discrimination in a comparative sense, comparators have had clear appeal as an evidentiary or heuristic aid for gauging whether discrimination has occurred.
Comparators and comparisons make visible the occurrence of comparatively adverse treatment or impact, by showing that not all employees or applicants have been fired, disciplined, rejected, or otherwise unfavorably treated or adversely impacted. When statistics are introduced in the litigation, the relative proportion of employees or applicants who were fired, disciplined, or rejected from each group becomes the basis that the court uses to draw inferences of discriminatory adverse action. Comparison of variously treated, or impacted employee(s), or applicant(s) helps isolate whether the protected trait is the reason or the cause for the adverse action.
Because of their utility in enabling inferences of adverse action based on a protected trait, comparators and comparisons have emerged as the predominant methodological device for evaluating discrimination claims.
I have noticed a curious trend. Federal courts and the Office of Federal Contract Compliance Programs (“OFCCP” hereinafter) have been moving in opposite directions in their approaches to comparators and comparisons. The stringent and tight standards developed by federal courts are in contrast to a loose and utilitarian approach taken by OFCCP. Instead of using a principled approach that relies on federal jurisprudence, the OFCCP’s approach appears to be based on the end justifying the means – for which the end is assessed in terms of maximum victims and damages assessed against contractors. Some of the conciliation agreements entered by the contractors and the OFCCP include “findings” of a contractor discriminating against almost every sub-group present in the contractor applicant flow. How could that be?
This article will not focus on comparators in the sense of “similarly situated” in all material respects. Meeting the similarly situated standard set by the federal court requires that the totality of circumstances must be nearly identical (tight fit). It will not address to what extent the nature of the theory of discrimination behind the claim (e.g. individual cases of disparate treatment, patterns-or-practices of disparate treatment, or disparate impact) affects or impacts the choice of comparator and the type of comparisons. Further, due to the complexity of addressing or considering the question of comparators, comparisons, and discrimination in compensation, this analysis does not apply to discrimination in compensation.
This article focuses on comparators and comparisons of the immutable traits generally referenced in statutes and regulations or “created” by the court, the OFCCP, or the Equal Employment Opportunity Commission (EEOC) (e.g. Women, Men, Asian, Black, Hispanic, etc.). I will critically examine the OFCCP’s approach to comparators and comparisons that federal contractors are now very familiar with, expose the assumptions embedded in them, and suggest that flawed comparisons and inappropriate comparators do not warrant the agency reliance on them as illuminators of discrimination. We characterize the OFCCP’s non-contextual approach as the computerized manufacturing of comparators and comparisons.
Young v. UPS was a Multiple Comparators Case
We will begin our discussion with the recent Supreme Court decision of Young v. UPS, a case with no apparent connection to the OFCCP. Justice Breyer’s majority opinion, Justice Alito’s thoughtful concurrence, Justice Scalia’s colorful dissent, and Justice Kennedy’s tempered dissent could not produce a workably clear guidance on whether employers covered by the Pregnancy Discrimination Act of Title VII may be required to make reasonable accommodations for work restrictions caused by pregnancy and related conditions. The majority opinion, perhaps sensing the limitations of its reasoning, cautioned that the new pregnancy discrimination test crafted in Young is narrow and “limited to the Pregnancy Discrimination Act context.”
Under these circumstances, one may ask how could this case be signaling anything? It would seem to be even less signaling to the OFCCP, which had nothing to do with the case (except in a narrow relation to the proposed sex discrimination regulations). It is my contention, however, that the era of computerized manufacturing of comparators, comparison and discrimination is about to enter a rocky phase.
At the heart of the court’s deliberations and decision is the identification of the appropriate comparator(s) to the pregnant female employees. The justices had a genuine and difficult struggle with the task of identifying the appropriate comparators to pregnant employees. This was understandable since the justices were faced with a plaintiff with two protected traits (gender and pregnancy) and a complex situation of multiple comparators. In fact, the record presented them with more than three groups of comparators to the group of pregnant employees. Apparently, Ms. Young and other UPS workers were subject to a collective bargaining agreement, which provided for reasonable accommodations for groups in categories: (1) disabilities within the meaning of the Americans with Disabilities Act; (2) on-the-job injuries; and (3) employees who were unable to drive because they had lost their certifications under U.S. Department of Transportation regulations.
In Young, the Court Unanimously Rejected the Comparison of Pregnant Workers to the Most Favorably Treated Group of Employees
In its struggle to identify the appropriate comparator to pregnant workers, the court (the majority, the concurrence, and the dissents) unanimously rejected Young’s argument in favor of comparing pregnant workers to the most favorably treated comparator. Clearly, the court unanimously rejected what it called the “most-favored-nation” status approach to constructing the appropriate comparison for pregnant female employees in need of reasonable accommodations for work restrictions caused by pregnancy and related conditions . This is what makes Young different from many other run of the mill sex discrimination cases. Many of those cases involved the comparison of one favored group (men) and one disfavored group (females). Consequently, there was no “most-favored-nation” since the “most-favored-nation” and the “favored nation” were the same group (generally males). Now fast-forwarding to OFCCP…
OFCCP Discovers the Computerized Model for Manufacturing Comparators, Comparisons, and Discrimination
For years now, federal contractors have been very familiar with OFFCP’s unique approach to auditing for, and finding of discrimination. While the EEOC (in theory) is limited to the allegations made in the charge of discrimination (the four corners of the charge), when constructing the appropriate comparator that the EEOC will use to develop its case. The OFCCP, on the other hand, runs literally thousands of comparative analyses (comparisons) on the federal contractor affirmative action plan under audit, using a mathematical model that encompasses all possible sub-groups of comparators arranged in all possible combinations.
Subject to the 2% threshold of any sub-group or aggregated group in the applicant flow or the contractor workforce, the spectrum of comparators and comparisons of selection rates (56 in all) that the OFCCP ran, pre VF Jeans (“VF” hereinafter) ruling, included:
- The 4 “Traditional Comparisons”
- Male v. Female & Female v. Male;
- Minority v. White & White v. Minority (Disallowed under the maximalist interpretation of VF)
- The 12 “All Other Comparisons”
- Black v. All Other & All Other v. Black (Disallowed by VF)
- Hispanic v. All Other & All Other v. Hispanic (Disallowed by VF)
- Asian v. All Other & All Other v. Asian (Disallowed by VF)
- Native American v. All Other & All Other v. Native American (Disallowed by VF)
- Pacific Islander v. All Other & All Other v. Pacific Islander (Disallowed by VF)
- Two-or-More Races v. All Other & All Other v. Two-or-More Races (Disallowed by VF)
- The 42 “Sub-Minority Comparisons”
- White v. Black & Black v. White
- White v. Hispanic & Hispanic v. White
- White v. Asian & Asian v. White;
- White v. American Indian & American Indian v. White
- White v. Pacific Islander & Pacific Islander v. White
- White v. Two-or-More Races & Two-or-More Races v. White
- Black v. Hispanic & Hispanic v. Black
- Black v. Asian & Asian v. Black
- Black v. American Indian & American Indian v. Black
- Black v. Pacific Islander & Pacific Islander v. Black
- Black v. Two-or-More Races & Two-or-More Races v. Black
- Hispanic v. Asian & Asian v. Hispanic
- Hispanic v. American Indian & American Indian v. Hispanic
- Hispanic v. Pacific Islander & Pacific Islander v. Hispanic
- Hispanic v. Two-or-More Races & Two-or-More Races v. Hispanic
- Asian v. American Indian & American Indian v. Asian
- Asian v. Pacific Islander & Pacific Islander v. Asian
- Asian v. Two-or-More Races & Two-or-More Races v. Asian
- American Indian v. Pacific Islander & Pacific Islander v. American Indian
- American Indian v. Two-or-More Races & Two-or-More Races v. American Indian
- Pacific Islander v. Two-or-More Races & Two-or-More Races v. Pacific Islander
- The “Favored Group Comparisons”. Comparing the group with highest selection rate to all other groups individually and in all other possible aggregations is counted in the 56 type of comparisons described above. For OFCCP, this has become the preferred “go to” comparison.
In its golden years, this approach included 56 binary comparisons. If a federal contractor’s Affirmative Action Plan (AAP) under audit had 30 job groups, and if 3 types of selections (hiring, promotion and termination) were being audited, the OFCCP will run 56 x 30 x 3 = 5040 comparisons of selection rates.
Not only will the risk of finding a significant difference in selection rate (a trigger) increase with the number of comparisons the OFCCP conducts, but when we consider how federal contractors, or companies in general, manage their selection and how disparities in selection rates between groups come to life, it would be very difficult for a federal contractor to avoid a trigger. Several reasons can cause difference in selection rates, including, but not limited to: chance, measurement problems, record-keeping problems, difference in group distribution size, subgroup differences, true population differences, etc.
Unfortunately, the federal contractors with the most effective recruitment programs, in attracting a diverse pool of applicants, are also the most vulnerable under the OFCCP computerized model of manufacturing comparators and discrimination! (We will leave this discussion for another day.)
Are all the comparators and comparisons included in the OFCCP mathematical/computerized model appropriate for purpose of auditing adverse impact or any other type of discrimination? Not according to the administrative law judge (ALJ) that presided over OFCCP v. VF Jeanswear.
The OFCCP Computerized Approach of Manufacturing Comparators and Adverse Impact Gets Some Trimming in OFCCP v. VF Jeanswear
OFCCP filed an administrative complaint against VF Jeanswear Limited Partnership (“VF” hereinafter). The complaint alleged that VF violated Executive Order 11246 by discriminating in favor of “Asian” applicants and against “non-Asian” applicants (all other racial and ethnic groups of comparators combined) when VF was hiring into its Operative Job Group. According to OFCCP pleadings, the selection rate of non-Asian applicants was 15.8% and the selection rate of Asian applicants was 43.5%. This disproportionate selection rate was statistically significant at the level of 6.79 standard deviations and produced a shortfall of 31.
The ALJ took issue with the assumptions and premises of the OFCCP comparator and comparison on which its case rested. Notably, the assertion that “non-Asian” is a “race” was found by the ALJ to be baseless, because it is not a category recognized in the regulations implementing the Executive Order. The ALJ found that OFCCP had made up, out of thin air, a group of comparators to the sub-category of Asian and applied statistical analysis to this made-up aggregate group (Non-Asians). The ALJ could not identify any rationale that could justify OFCCP deviation from its own regulations.
There is no doubt that the ALJ decision in VF trimmed some of the OFCCP’s abuses and excesses in manufacturing comparators, comparisons and adverse impact during audits. However, there is still some uncertainty on whether VF should be interpreted narrowly or broadly. The minimalist interpretation reads the ALJ decision as a rejection of the “non-Asians” group of comparators as well as “non-Hispanics,” “non-Blacks,” “non-Native American,” and “non-Pacific Islanders.” The maximalist interpretation, on the other hand, reads the ALJ decisions to also include “non-Whites.”
So what is the conclusion from the VF case? The OFCCP can’t create groups of comparators that have no support in its own regulations without probative and valuable contextual evidence. The OFCCP made a strategic decision not to appeal the ALJ ruling and left it standing. This was certainly a setback that trimmed the OFCCP computerized model of manufacturing comparators, comparisons, and adverse impact. It exposed its excesses and abuses, disallowing many comparisons of “sub-group v. all other” from the OFCCP arsenal. However, VF did not reign in, as the issue was not before the court, the OFCCP’s most lethal weapon: Any sub-group compared to the sub-group with the highest selection rate.
Regulatory Support for the Use of “Most-Favored-Nation” or the Comparator with the Highest Selection Rate
I described above the approach that compares the selection rate of the sub-group with the highest selection rate to the selection rate of any other sub-group as the most lethal weapon in the OFCCP arsenal. Statistically speaking, it is the approach that produces the maximum disparities in selection rates, yielding the highest measure of statistical significance, and the highest shortfall. As many federal contractors painfully learned from their interactions with the OFCCP, the higher the shortfall, the higher will be the damages assessed against contractors.
Naturally, any argument that will try to justify the “most-favored-nation” approach – because it is the approach that maximizes the size of the damages – will likely be a losing argument without some other rationale that the court can accept. What is the OFCCP’s rationale for the “most-favored-nation” approach to comparators and comparisons?
There is certainly a regulatory basis for using the approach that compares the group with the highest selection rate with any individual sub-group. The Guidelines on Employee Selection Procedures actually define adverse impact as “a selection rate for any race, sex, or ethnic group which is less than four-fifths (4/5) (or eighty percent) of the rate for the group with the highest rate…”. 41 C.F.R. § 60-3.4(D). While not binding on contractors, the Federal Contractor Compliance Manual suggests to Compliance Officers that comparators and comparisons should be framed using the sub-group with the highest selection rate compared to each or any individual sub-group.
But the regulations cited above are specifically limited to adverse impact. Is there a basis in OFCCP regulations that supports “the most-favored-nation” approach in systemic and individual cases of disparate treatment when multiple comparators with different selection rates are present? I found none.
As an illustration, let’s consider the example of an applicant flow in which Asians were selected at the rate of 30%, Whites were selected at the rate of 15%, Blacks were selected at the rate of 12%, and Hispanics were selected at the rate of 10%. After the OFCCP compared selection rates, it found that the difference in selection rates between Asian v. White, Asian v. Black, Asian v. Hispanic, were statistically significant. With these numbers, OFCCP will attempt to pressure the contractors to enter into a conciliation agreement under which the victims are the single sub-group of White, Black, and Hispanic. Were the comparators and the comparisons conducted by OFCCP appropriate, suitable, or valid? Could the OFCCP have compared the selection rate of each group to the average selection rate of all the groups? In this example the average selection rate is 16.75%. So this alternative approach could have consisted of comparing the selection rate of each sub-group to the average selection rate of 16.75%. What if the comparison of each sub-group selection rate to the average selection rate does not reveal any significant difference in selection rate? Can the OFCCP’s approach be justified only because it yields the maximum return (getting the contractor in the deepest trouble possible)?
These are all good questions. However, without contextual evidence, it is not possible to determine whether the comparators and the comparisons selected by OFCCP are appropriate, valid, or suitable.
The Case of Intersectional, Trait-Plus, and Identity Discrimination – AKA “Complex Bias Claims”
Title VII prohibition against discrimination has been interpreted by some courts not to be limited to only one of the immutable and distinct traits listed in the statute.
There are intersectional claims (the intersection of two or more protected traits), arising when an individual seeks to show that the employer discriminated because of the individual’s particular combination of protected traits, rather than simply trying to show that the employer discriminated on the basis of one or two distinct traits. The EEOC has brought intersectional cases, in which discrimination in the workplace was alleged on the basis of Race & Age, Race & Disability, Race & Gender, Race & National Origin, Race & Pregnancy, and Race & Religion. For example, a plaintiff who files an intersectional claim under Title VII will try to prove that an employer discriminates against Black women, even if the employer does not discriminate against White women or Black men. Likewise, an intersectional claim filed by an Asian American woman will attempt to prove that the employer discriminates based on stereotypes and assumptions about Asian American women, even in the absence of discrimination against Asian American men or White women.
There are also various cases – known as “trait-plus” – that relate to situations in which an employer discriminates against (e.g. imposes a rule on members of one group) a group in a workplace based on a combination of one or more of their protected traits and some other unprotected attribute. Examples of trait-plus groups include but are not limited to men/women with young children, men/women married to a fellow employee, men/women with family responsibilities, married men/women, unmarried men/women, unmarried black women with children, men/women living in common-law relationships, black men with criminal background.
There are also the identity discrimination cases when an employer discriminates against an employee because of his/her gender identity or sexual orientation, that are brought within Title VII through the vehicle of gender stereotypes.
Courts have been struggling with the application of comparators and comparisons as a methodological device for evaluating complex bias discrimination. Complex bias cases either present the court with too many comparators, or an absence of comparators. For example, courts had to determine whether the proper comparator may only include a person outside of the protected class who has the same “trait-plus characteristic” as the plaintiff (in this example, a male with young children) or whether the comparator may include any person (male or female) who has the reverse trait but lacks the “plus” characteristic (in this case, a female without young children). In some complex bias cases, it is almost impossible, as a practical matter, for an individual to find his or her negative mirror image to show that discrimination has occurred.
The multiple comparators present special problems in religious accommodation cases. Large employers typically deal with multiple comparators and comparisons when they try to manage a large number of religious accommodation requests from their employees. Different religious practices present different levels of conflict with workplace rules. This reality can create unsurmountable challenges to employers who try to consistently and evenhandedly manage the religious accommodation requests they receive from their employees, or the court, when trying to assess a religious discrimination claim through multiple comparators methodology.
The increased diversity in the workplace and in today’s society suggests to us that the number of complex bias claims will increase. With the effective date of the LGBT Executive Order behind us, the OFCCP is now enforcing non-discrimination requirements against the categories of: Individuals with Disabilities, Veteran, Black, White, Hispanic, Asian, Pacific Islander, American Indian, Two-or-More Races, Gender Identity, Sexual Orientation, Sex, Religion, National Origin, and Color. The OFCCP just announced that it will be investigating LGBT complaints from employees of federal contractors. The OFCCP will be operating like the EEOC in handling LGBT complaints. Whether ready for it or not, the OFCCP has entered the complex claim arena. Sooner or later OFCCP will have to deal with intersectionality, trait-plus, and identity claims. These claims will expose the limits of the OFCCP computerized approach of manufacturing comparators, comparisons, and discrimination.
Contextual Considerations are Critical in Identifying Appropriate Comparators
There are almost infinite possibilities in constructing comparators and comparisons, as I discussed above, when we account for the different possibilities opened by complex bias claims. The explosion of diversity and identities are putting intense pressure on the limited numbers of traits listed in current anti-discrimination laws (we will leave this discussion for another day). The courts are struggling with the crafting of standards that will introduce some quality control and quality assurance in the process of constructing comparators and comparisons to avoid – among other things, sliding into empty formalism. What should be the standard to assess and judge whether a comparator and a comparison are suitable, valid, or appropriate?
An answer that has been proposed by many is taking into consideration the context. Context is the sociology of the workplace; it includes understanding social relations in the workplace, the demographics of the workforce, the diversity of the workforce, the demographics of the areas from which the organization recruits and selects, the patterns of exclusion and inclusion, the networks of power, the dynamics of the relationship between the employees, and the policies and workplace governance etc. All could be deemed worthy of consideration in deciding the suitability, the appropriateness, or the validity of the comparator and the comparison, and ultimately whether to allow a discrimination claim to proceed under certain types of comparators and comparisons.
Context is hardly a novel concept. The court has considered contextual evidence and given weight to the context in harassment cases, adverse impact cases, disparate treatment cases, and stereotyping cases. The context helped the court decide on relevance, suitability and appropriateness to the comparator and the comparison.
Griggs v. Duke Power had its contextual background. The Griggs court recognized that Black petitioners have long received inferior education in segregated schools. The Griggs’ court took judicial notice of its own previous case (Gaston County v. United States), in which the court barred the institution of a literacy test for voter registration, on the grounds that the test would indirectly abridge the right to vote on account of race, because of the inferior education received by Blacks in North Carolina. This was the obvious context for comparing Whites and Blacks.
More recently, criminal background cases are anchored in context. Racial and ethnic disparities are reflected in incarceration rates. One in 106 white men, one in 36 Hispanic men, and one in 15 African American men are incarcerated. This context should justify comparing White males to Black males or White males to Hispanic males.
When context is taken into consideration, comparators and comparisons offer a seemingly bright-line framework for identifying elusive facts and resolving complex social judgments. Justice Brennan, in his dissent in Gilbert, rejected the formalism of the comparative exercise of the majority. He reminded us that discrimination is “a social phenomenon encased in a social context,” rather than simply a matter of formal line-drawing.
Contractors should first ensure that the comparators and comparisons used by the OFCCP during audits are supported by its own regulations and second, and require that the agency produces contextual evidence that supports its choice of comparators and comparisons. So far, the OFCCP has had a free pass to run its computerized approach to manufacturing, comparators, comparisons, and discrimination to yield the maximum damages without any serious challenge from contractors.
This article was prepared by the author in his personal capacity. The views and opinions expressed in this article are the author’s own and do not reflect the views of his current or former employers.