Nathan D. Woods, Jora B. Stixrud and Doug H. Owens, authored “Narrowed Classes: Statistical Differences and Damages” which focuses on statistical implications of more narrowly defined employment class actions in the wake of the Supreme Court’s decision in Wal-Mart v. Dukes. (download)

[Reprinted with permission from Law360.com]

Narrowed Classes: Statistical Differences And Damages

Law360, New York (August 8, 2011) -- In the weeks since the U.S. Supreme Court handed down its decision in Wal-Mart Stores Inc. v. Dukes et al., observers have offered a number of legal insights regarding the verdict’s potential impact on employment class actions in the future.

As analysts of data in employment class actions, we see an additional implication arising from one outcome of the court’s decision on which all observers seem to agree. That is, future class actions will need to be more tightly focused around particular policies or practices, organizational units, geographies, positions or supervisors associated with the alleged illegal activity.

Assuming this is correct, it stands to reason that potential class members will be more homogeneous on the dimensions affected by the alleged discriminatory actions. In contrast to a widely defined class of employees, we might expect a more narrowly tailored class to include those parts of a company which exhibit results apparently adverse to a protected group, and leave aside those parts of the company where that same group appears to be treated the same, or is advantaged, relative to a nonprotected group.

If this occurs, assuming this portion of the protected group really is harmed, an empirical analysis may show a more compelling finding of substantively and statistically important aggregate differences and a more persuasive statistical pattern across a narrowly proposed class.

Statistical Differences and Damages

In almost any potential class of protected group members the extent to which each employee appears to be harmed by alleged discriminatory actions will not be the same. However, the common practice of demonstrating whether harm has occurred through a comparison of overall average outcomes between protected and nonprotected groups ignores this underlying variation.

The estimated aggregate difference is based on the entire population studied and includes subgroups with differences apparently to the disadvantage of a protected group, but also subgroups with differences that are neutral or even advantage a protected group.

Using pay as an example, suppose we estimate a difference of $1,000 a year, on average, across a broadly defined class. This does not mean that all protected group members show the same $1,000 difference. Rather, looking at the individual differences estimated from the same model, some protected group members might appear to be underpaid by $4,000 or $5,000, while others appear to be overpaid by $2,000 or $3,000, and others exhibit small differences centering on zero.

Summing all of these differences together for purposes of calculating an overall average difference in pay between protected and nonprotected employees has the effect of diluting relatively extreme pay differences to the disadvantage of some protected group members with the neutral or even advantageous differences elsewhere among the broadly proposed class.

Going forward, the general consensus is that class actions will be smaller with regard to the number of employees included, and specifically tailored to employees subject to a particular policy or practice, or working in a particular geography, organizational unit or job. This would seem to increase the burden on plaintiffs’ counsel to identify the particular practice or segment of a company associated with the alleged harm.

However, assuming the narrowed class is carefully defined by plaintiffs’ counsel, the data analyzed for this subset of protected group members should exhibit more of the alleged harm on average, since many of the protected group members who would show little or no difference, or are advantaged, will be excluded.

Therefore, the estimated average difference within this more narrowly defined class will almost certainly be higher, on average, than it would have been under a broader class definition. Again taking pay as an example, rather than a $1,000 difference across a wider class we might see a $4,000 difference estimated across a narrow one.

Furthermore, assuming the narrowed class is carefully defined, we should expect the percentage of the class members who appear to have suffered some harm will be higher than in the larger group. This is because more of those who exhibit neutral results or who appear to be advantaged would now be excluded, leaving a larger share of apparently disadvantaged employees. The data will therefore be more likely to show both that a difference exists on average, and that underneath the average difference a larger share of the individual class members are apparently disadvantaged.

A final point to consider in this regard is that, depending on the case and the methods of damage calculations, it is possible that estimated aggregate damages for a smaller, focused class could be as high as or higher than they would have been for a larger, more broadly defined class. Consider the example of a broadly defined class of 10,000 protected group members who, based on an analysis of pay, appear to be underpaid by $1,000 per year on average (recall that this average includes some positive and some negative differences, but the average across everyone is $1,000).

Under one very simple approach to damages where the average difference applies to all protected group members, the damages would be $10 million (10,000 employees multiplied by $1,000). Now consider restricting this broad class to a more narrowly defined class of 2,500 protected group members who are subject to the same allegedly harmful policy, or work under the same decision maker, for example.

If this subset of 2,500 were selected strategically we would expect to see that the estimated average difference in pay would be higher for this group than it was for the larger group of 10,000. If it were an average difference of $4,000 per year (recall now that this average still includes some positive and some negative differences, but substantially more negatives than in the broadly defined example), then following the same simple approach the damages for this group would also be $10 million (2,500 employees multiplied by $4,000).

Simply put, if the measured average difference is substantially larger among the remaining protected group members, a smaller class does not necessarily imply a smaller potential for damages.

Outlook for Plaintiffs and Employers

The class action landscape will change as a result of the Supreme Court’s decision in Wal- Mart Stores Inc. v. Dukes et al. From our vantage point as analysts of data in these matters, more narrowly defined classes offer the possibility of a more compelling demonstration of statistical evidence to the disadvantage of protected group members.

Assuming plaintiffs’ counsel carefully investigates and tailors a class, establishing a common pattern in the data and demonstrating the merits of the claim through analysis should be more straightforward. And if a class is certified and the merits of a claim proven, the potential downside for plaintiffs of a smaller class will likely be mitigated somewhat or entirely by the higher per person damages.

Employers need continued vigilance to ensure company policies, and the decisions being made on the ground are not exposing them to the risk of discrimination litigation. It is even more important now to perform regular audits of employment practices and to ensure those audits delve below aggregate-level analyses to address whether any particular policy, practice, organizational units, geographies, positions or supervisors are associated with possibly discriminatory activity.

These more narrowly defined entities may well be the specific focus of future litigation, and not simply one part among many in a larger class action where their negative contribution to an aggregate disparity is counterbalanced by advantageous differentials elsewhere in a company.

--By Nathan D. Woods, Jora B. Stixrud and Doug H. Owens, Welch Consulting

Nathan Woods, Ph.D., is an economist and director of the Washington, D.C., office of Welch Consulting. Jora Stixrud, Ph.D., and Doug Owens, Ph.D., are economists in the firm's Washington office.

The opinions expressed are those of the authors and do not necessarily reflect the views of the firm, its clients, or Portfolio Media, publisher of Law360. This article is for general information purposes and is not intended to be and should not be taken as legal advice.