Duking It Out With Wal-Mart

Thanks for inviting me back to Concurring Opinions. Last visit, I spent much of my month ranting about Ricci v. DeStefano – the New Haven firefighters case. This year I will try to avoid doing the same about Wal-Mart v. Dukes but I do want to say something. The decision is a major procedural decision limiting the availability of class actions for employment discrimination claims but also for class actions generally. To reach its decision, the Court indicated that it had to address the underlying substantive law which involved claims of systemic disparate treatment and disparate impact. The discussion of substance was in the context of deciding whether there were common questions of law and fact needed to satisfy Rule 23(a) and that is certainly not the same as discussing the substantive law directly. Nevertheless, I think the way the substantive law was discussed may well be a foreshadowing, a grim foreshadowing, of how the lower courts and the Supreme Court will treat the two systemic theories of discrimination in the future. The hope is that Wal-Mart will be treated “only” as a procedural class action case.

The plaintiffs’ claim was that Wal-Mart had a policy of granting unfettered discretion to its store managers to make pay and promotion decisions and it operated as a pattern of pay and promotion discrimination . The discretion policy is in sharp contrast to the general way in which Wal-Mart operates. Wal- art has been heralded as having developed the most sophisticated systems yet for collecting, analyzing and acting upon data flowing to its Bentonville headquarters in real time in all aspects and all locations of its business. If, for example, a freezer unit in a Wal-Mart location in Shanghai starts drawing electricity beyond established parameters, that information is transmitted to Bentonville, analyzed and the local facility is notified and ordered to deal with whatever problems that heightened power usage reveals. Like the rest of the data generated in the operation of the business generally, the pay and promotion data is collected in the Bentonville. The difference is that nothing is done about what were concededly dramaticshortfalls in pay and promotion of women working at Wal-Mart stores.

The evidence, which was unchallenged, showed that women filled 70% of the hourly jobs but only 33% of management jobs, with most promotions coming from the pool of hourly workers. Further, it took women longer than men to rise into the management ranks and the higher in the management hierarchy the fewer the women. Finally, women were paid less than men in every region and that salary gap widened over time, even for men and women hired into the same jobs at thesame time. Based on that basic statistical evidence, plaintiffs claimed that this system of making pay and promotions was a pattern of systemic disparate treatment discrimination and the discretion policy operated as an employment practice that resulted in disparate impact to women. This post will deal with one aspect of the systemic disparate treatment claim – the failure of the Court to confront the statistical evidence of discrimination that plaintiffs presented. 

Reading the transcript of the oral argument as well as the resulting opinion for the Court makes it clear that the majority appears not to be able to get their minds around the concept that the thrust of the case was aimed at the way Wal-Mart operated as to pay and promotion and was not looking only at a series of individual pay and promotion decisions made by individual store managers. By looking only at these individual decisions, and not ever looking at the aggregation of all the decisions, the Court found itself able to conclude that there were no common question of law or fact since not each and every pay and promotion decision was discriminatory. The majority just could not see plaintiffs’ claim: Because all the store managers each had this unfettered discretion all the women workers faced the risk that the discretion would be exercised in a discriminatory way and the existence of that risk was the common question of  act underlying plaintiffs class action claim. The aggregated data that Wal-Mart collected reveals that there is an extreme shortfall for women in pay and promotions.  Plaintiffs do not claim that all the individual managers discriminated all the time. But plaintiffs do claim that this statistical evidence at least raises a strong suspicion that Wal-Mart’s policy of store manager discretion allows a considerable amount of discrimination to occur. Plaintiffs pointed to significant additional evidence that supports drawing the inference that the operation of the policy amounts to systemic disparate treatment.

Writing for the Court, Justice Scalia describes how he thinks the discretion policy operates:  He intuits that most managers, knowing that Wal-Mart has promulgated a formal policy prohibiting discrimination, do not discriminate, others discriminate intentionally and others discriminate by relying on stereotypes or unconscious bias. Accepting his unsubstantiated hunch about how this works, supports, rather than undermines, the inference that all the women working in the Wal-Mart stores faced the real risk of pay and promotion discrimination.

The Court did not put the policy together with the results of its operation because it does not even mention the statistical evidence that makes the connection. After an unexplained detour from the structure of a previously accepted systemic disparate pattern or practice case to essentially trash the use of “social framework” expert testimony, Justice Scalia then does discuss the sophisticated statistical evidence plaintiffs relied on. The plaintiffs’ experts concluded, at a statistically significant level, that pay and promotions were related to the sex of employees in every region and across all regions of the company. Justice Scalia rejected the probative value of that evidence because it was “insufficient to establish that [plaintiffs’] theory can be proved on a classwide basis. . . A regional pay disparity, for example, may be attributable to only a small set of Wal-Mart stores, and cannot by itself establish the uniform, store-by-store disparity upon with the plaintiffs’ theory of commonality depends.” Again plaintiffs’ claim is not that there was uniform store-by-store disparity. What was uniform across all the stores that all the women faced was the risk of pay and promotion discrimination when their managers make pay and promotion decisions.

Why the Courtfails to be able to understand a fairly straightforward systemic disparate treatment claim is hard to understand.  The precedent established back in United States v. Teamsters, Hazelwood School District and Bazemore v. Friday has not been formally overruled. But the question for the future is whether or not those cases will not be viewed as undermined by the decision in Wal-Mart. Consistent with its failure to look at how the statistical evidence was connected to the discretion policy, the majority appears to understand discrimination as only occurring at the individual decision making level — the “few bad apples” perspective on the perpetrator that disregards the victims rather completely.

Further support for the Court’s narrow vision of how discrimination can occur can be seen by juxtaposing Wal-Mart’s failure to understand a straightforward systemic claim with the Court’s individual disparate treatment and retaliation decisions that are more favorable to plaintiffs than the lower courts’ approaches suggests that the Court really thinks discrimination only occurs at the individual decisionmaking level. While the antidiscrimination statutes are generally tort like in their structure, even tort law has escaped the straightjacket limiting tort claims to individual occurrences. Would that the Supreme Court majority raise its eyes a bit higher.

Having refused to certify plaintiffs’ class claim, the Court remanded to determine plaintiffs’ individual claims. It would be interesting to see if the Wal-Mart plaintiffs will continue to assert their systemic claims, even in absence of a class action. If they would, it would be interesting if  Wal-Mart had to explain its general obsession with data collection, analysis and centralized action as to all of its operations but its failure to do anything once the collected and analyzed data revealed such a dramatic impact on the pay and promotion of women workers at its stores.

You may also like...

4 Responses

  1. Joey Fishkin says:


    Thanks for this post. There was actually a bit of recent news from the plaintiffs’ attorneys regarding how they intend to proceed: state-level lawsuits. See here:


  2. Brett Bellmore says:

    “Why the Court fails to be able to understand a fairly straightforward systemic disparate treatment claim is hard to understand.”

    Perhaps the Court is starting to get tired of statistical cases for discrimination which don’t bother to prove any particular individual instance of it? Maybe they even think that taking statistical disparities as proof of discrimination is uncomfortably close to instituting quotas.

    In any event, “fails to understand” and “rejects” are different things. People, and courts, are quite capable of rejecting things they do, in fact, understand.

  3. Brett Bellmore says:

    I suppose I ought to be more specific about why one might reject, (Rather than fail to understand) a statistical case for discrimination.

    Whenever you make a statistical case for “X” causing “Y”, (Discrimination causing discrepancy, in this instance.) you have to exclude the possibility of a spurious correlation being produced by both “X” and “Y” being correlated with some third factor.

    Now, they went to a little effort to preclude such obvious confounding variables such as hours worked. But, as is typically the case, they more or less just rejected out of hand the possibility that, on average, there were a larger percentage of highly meritorious employees among the men, than among the women.

    This is significant because a good deal of research has demonstrated that, while the average performance of men and women on intellectual tasks is close to the same, men have considerably more dispersion in their performance. More unusually good performers, AND more unusually bad.

    Since the people who get promoted are generally supposed to be the unusually good performers, this would imply more promotions among men than women, even if the two groups averaged the same on a number of metrics.

    How would you avoid this sort of thing producing a spurious conclusion that discrimination was taking place? You have to examine individual cases. Exactly what the advocate of statistical cases for discrimination doesn’t want to do.

    But what the Court does want done.

  4. Mike Zimmer says:

    Thanks to Joey for the update. The new amended complaint refocuses the original claim to the Wal-Mart stores in its California Region rather than nationwide. Here is the link: http://www.walmartclass.com/staticdata/Fourth%20Amended%20complaint%20final%20with%20exhibits.pdf.

    As for Brett’s point, the Court in Wal-Mart was unanimous in holding that a formula-based remedy would violate Wal-Mart’s right, and presumably the class members’ right to have their individual cases decided individually.

    Those procedures would only come into play once Wal-Mart’s liability was established and the burden of persuasion, pursuant to the Teamsters case, would shift to Wal-Mart to prove that, as to each individual class member, it did not discriminate in pay and promotion decisions.

    The Teamsters, Hazelwood School Disctrict and Bazemore v. Friday cases are all precedent for when liability is established so that the burden shifts.