Recent headlines like "Merrill Lynch $160M Deal to End Merrill Race Bias Suit Gets Final OK" should not just catch our attention, but hold it. How, after all, in a post-Dukes world could allegations regarding decentralized decision-making support a class case? And how, in a post-Comcast world, could class claims requiring individualized determinations of lost compensation lead to the third biggest settlement of race-bias litigation in the United States? In his class certification opinion, Judge Posner of the Seventh Circuit tells us how, but, his explanation leaves more questions than answers.

First, consider the plaintiffs' case. More than eight years ago, the McReynolds plaintiffs sued on behalf of hundreds of African American financial brokers, alleging that Merrill Lynch systematically channeled business to white employees. The district judge denied class certification three times. Then, offering Wal–Mart Stores, Inc. v. Dukes, 131 S.Ct. 2541 (2011) as a reason for class certification, the plaintiffs persuaded the Seventh Circuit to take another look. On appeal, they argued that two practices disparately impacted Merrill Lynch's African American brokers: 1) a "teaming policy" that allowed brokers to work on teams with other brokers they chose; and 2) an "account distribution" policy that considered past success in reassigning the accounts of departing brokers. The plaintiffs alleged that the voluntary teams tended to exclude African American brokers, and so lowered their compensation and access to redistributed accounts.

Next consider Dukes, the Supreme Court's decision overturning certification of a company-wide class of women employees who complained about decentralized decision-making at Wal-Mart. The primary theory in Dukes was that Wal-Mart's practice of giving local managers discretion to make pay and promotions decisions "disparately impacted" women. The Supreme Court rejected that theory. Justice Scalia reasoned that "allowing discretion by local supervisors"—a practice "against having uniform employment practices"—would inflict the same injury on class members only if all of the company's far-flung managers exercised that discretion in the same way—i.e.with the same bias. But universal bias was not something the Court would simply assume, and it found the plaintiffs' statistical and anecdotal evidence too weak to suggest that all managers made biased decisions. The upshot was no "commonality." Manager discretion alone, said the Court, could not bind the women of Wal-Mart together into one class.

On the heels of DukesMcReynolds complained that a company practice of allowing discretion led to class-wide discrimination. The plaintiffs nevertheless cited Dukes, saying it favors class certification. And the Seventh Circuit agreed. Judge Posner wrote:

[T]o the extent that these regional and local managers exercise discretion regarding the compensation of the brokers whom they supervise, the case is indeed like Wal–Mart. . . . [But] permitting brokers to form their own teams and prescribing criteria for account distributions that favor the already successful . . . are practices of Merrill Lynch, rather than practices that local managers can choose or not at their whim. Therefore challenging those policies in a class action is not forbidden by the Wal–Mart decision; rather that decision helps (as the district judge sensed) to show on which side of the line that separates a company-wide practice from an exercise of discretion by local managers this case falls.

In short, the Seventh Circuit found McReynolds different from Dukes because plaintiffs' co-workers rather than managers exercised discretion in an allegedly biased way.

The $160 million question is, why? On its face, Wal-Mart's practice of delegating pay and compensation decisions to its managers looks like Merrill Lynch's practice of delegating team-building decisions to brokers. And the alleged bias of Wal-Mart managers seems to function like the alleged bias of brokers forming teams. So unless brokers exercise discretion in the same biased way, how does giving them team-building discretion inflict the same injury across the entire class? The Seventh Circuit does not quite say, but finds:

[T]he brokers choose as team members people who are like themselves. If they are white, they, or some of them anyway, are more comfortable teaming with other white brokers.. . . [T]here is bound to be uncertainty about who will be effective in bringing and keeping shared clients; and when there is uncertainty people tend to base decisions on emotions and preconceptions, for want of objective criteria…

So whereas Dukes found the plaintiffs' evidence of bias insufficient, McReynolds seems to assume a bias among brokers, citing no evidence at all.

The apparent gap between McReynolds and Dukes only grows when we move from the why of class certification to the how. In Dukes, the Court rejected the use of Rule 23(b)(2) as a basis for certification because it only allows for certification to award uniform injunctive or declaratory relief, not individualized money damages. Individual claims for lost pay, the Court explained, must meet the stricter requirements of Rule 23(b)(3). But facing the same problem, the Seventh Circuit stepped lightly around it. It first certified a class under Rule 23(b)(2) for the limited purpose of deciding whether Merrill Lynch's practices violated federal law. It then explained that, if the practices were found to have violated federal law, the court could hold mini-trials to assess individual damages.

If this looks like an exception that could swallow Rule 23(b)(3), it may be. The Fifth Circuit has said as much, and has rejected "issues classes" for this reason. Comcast Corp. v. Behrend, 133 S. Ct. 1426 (2013), the Supreme Court decision bringing damages fully within the Rule 23(b)(3) analysis of common versus individual issues, may signal a similar view. Comcast that class representatives must meet the requirements of Rule 23(a) and at least one prong of Rule 23(b). Period. Still, the Seventh Circuit is not alone in allowing "issues classes" when damages are individualized. And even after Comcast, these partial class actions live on in the Sixth, Second, and Ninth circuits and have already gained some acceptance in others.

So what, if anything, can we make of this? What of a $160 million pay-out in McReynolds v. Merrill Lynch? Maybe just this: the borders of Rule 23 remain surprisingly porous. In the long run, Dukes, Comcast, and other Supreme Court decisions may sharply limit the kinds of employment cases that can be brought on behalf of a class. But in the long run, as Keynes famously said,[1] hundreds of millions of dollars may have been spent on the ones than can't.