Seyfarth Synopsis: In a decision with far–reaching implications for workplace class actions, the D.C. Circuit recently affirmed the denial of class certification of a Rule 23(b)(3) class on the grounds that the proposed class contained uninjured class members in the case of In Re Rail Freight Fuel Surcharge Antitrust Litigation, Dakota Granite Co, et. al. v. BNSF Railway Co., et. al. (decided August 16, 2019) (“Rail Freight”). In so doing, the D.C. Circuit joined the First Circuit in a decision issued last year, which denied certification on the same grounds. See In Re Asacol Antitrust Litig., 907 F. 3d 42, 51-58 (2018). The issue of whether a certified class may contain uninjured class members was left open in the Supreme Court’s decision in Tyson Foods Inc. v. Bouaphakeo, 136 S. Ct. 1036 (2016), and now has been answered in the negative by two federal Courts of Appeal. It should be required reading for any corporate counsel involved in workplace class action litigation.
Background Of The Case
The Rail Freight case was brought before an MDL panel on behalf of a class of over 16,000 shippers, claiming price fixing by the nation’s largest freight railways in violation of the Sherman and Clayton Acts. Plaintiffs’ key evidence to establish common causation, injury, and damages consisted of a regression analysis from their economist, Dr. Rausser, which estimated “negative damages” for 12.7% of the putative class members or more than 2,000 shippers – in other words, the model estimated that a significant portion of the putative class suffered no damages whatsoever. Id. at 8. Lacking any other proof that they were injured by the alleged price-fixing violations – and lacking any “winnowing mechanism” to segregate these uninjured class members – the D.C. Circuit ruled that plaintiffs’ statistical evidence, though admissible under Daubert, failed to show class-wide injury and therefore did not make the necessary showing of commonality and predominance under Rule 23. Id. at 11.
The D.C. Circuit’s Reasoning
The D.C. Circuits also reiterated a prior holding in the case that common evidence must “show all class members suffered some injury.” Id. at 9. Even accepting for the sake of argument that predominance might exist despite a de minimis number of uninjured class members, the D.C. Circuit suggested that it was the “raw number” here and not the percentage that was troubling because presenting individualized evidence for 2,037 class members was incompatible with the requirements of Rule 23. Id. at 10-12. Finally, the D.C. Circuit noted that questions of overly broad classes cannot be deferred to a post-certification stage, but must be confronted up front as “part-and-parcel of the ‘hard look’ required” by the Supreme Court for statistical models that purport to show predominance. Id. at 4, citing Comcast Corp. v. Behrend, 569 U.S. 27 (2013).
Implications For Employers
The type of statistical evidence which was critical to plaintiffs’ case in Rail Freight is also essential in workplace class actions asserting claims of adverse impact, and raises the same problem of uninjured class members. The standard regression models supporting class certification in adverse impact cases address the question of whether an alleged discriminatory policy or practice adversely affects a protected class, on average, at a statistically significant level, after accounting for all major non-discriminatory variables. By way of example, plaintiffs might challenge an employer’s crediting of certain levels of education in deciding whether to promote candidates to a supervisory position, arguing that this factor is not sufficiently job-related and favors men. Plaintiffs might then offer a statistical model that purports to isolate the effect of education on promotions of men and women (controlling for other major factors), with the goal of establishing that this factor has a statistically significant adverse impact on women candidates.
If plaintiffs are able to produce such a model, is that sufficient for class certification? The Rail Freight decision suggests that the answer may be no.
By its nature, a standard regression model only reflects average disparities (that is, the mean value of a normal distribution of observations). Class members may be more or less affected, and it is likely that some were not affected at all. To use our example, there will be some women who had credited levels of education and therefore received the benefit of its consideration (even if they were not ultimately promoted) and others who even with the credential would never have been promoted because their other qualifications were lacking in some other way. Neither of these categories of women have suffered any damages under plaintiffs’ case theory.
In adverse impact class-based litigation, courts have nevertheless allowed cases like this to proceed as class actions. Courts have reasoned that even if individualized damages are typically not amenable to class treatment, damages can be determined at a later phase where individuals pursue their own claims with a favorable presumption based on class liability, subject to the employer’s right to produce exonerating evidence. Rail Freight, however, instructs that the overinclusion of uninjured persons must be confronted at class certification and not deferred to a later stage. This standard applied to adverse impact litigation would make class certification more difficult, because the very statistical evidence used to support class certification, when given a “hard look,” may also defeat the showing of commonality or predominance in the first instance.