'My image of being a lawyer was strictly the white hat'

By Roger Parloff, Fortune senior editor

"But my image of being a lawyer was strictly the white hat," he recalls. "I wanted it to have some meaning."

In 1980 he joined a private employment discrimination firm in San Francisco, excelled, became managing partner, and became rich. "Given my political background," he says, being wealthy "was fairly embarrassing at some level."

In addition, he was becoming frustrated by "the imperatives of being in private practice" - i.e., staying profitable. "The market dynamics of these cases [puts an] emphasis on producing a large damages pot as quickly as possible," he explains, "and not so much on what I was interested in, which was systemic change."

In 1991, at age 40, he resigned his partnership, and the next year he started the Impact Fund. By the late 1990s, Seligman was looking to use the fund to broker partnerships between nonprofits and private class-action firms to launch significant cases.

For-profit firms like 67-lawyer Cohen Milstein Hausfeld & Toll - which Seligman ultimately brought into the Dukes case as one of the Fund's five co-counsels - could provide the financial muscle and staff, while nonprofit partners lent the case credibility with the press and grass-roots groups, and made sure that the for-profits didn't grab a "cheap, early settlement," Seligman explains.

Upon getting the inquiry about Wal-Mart, Seligman called Marc Bendick, an economist who had then served as an expert in more than 75 employment cases. Bendick had a database of information compiled from forms, called EEO-1s, that large companies are required to file with the Equal Employment Opportunity Commission, breaking down their workforces by gender and race.

Bendick told Seligman that Wal-Mart's profile was skewed and that "compared to its major competitors, it stuck out like a sore thumb," as Seligman recalls.

Bendick eventually computed that women composed 63.4% of Wal-Mart's hourly (nonmanagerial) workers yet just 33.6% of the store's salaried managers. More damning, the company's numbers were far below those of its 20 top competitors, which averaged 56.5% women at the store manager level.

Though these numbers are powerful, some caution is in order. Such "benchmark" or "comparator" studies - in which a company's EEO-1 data is compared to that of a group of purportedly similar companies - are a staple of class-action employment litigation, and the outcomes depend on how one defines the comparison group.

Wal-Mart disputes Bendick's list of comparison companies and claims that when a broader group is used - reflecting Wal-Mart's wide geographic footprint and the variety of products and services it sells - it falls within the norm. It also claims that if Wal-Mart had counted its highest-level hourly-wage supervisors as "managers" on its EEO-1 forms, the way it believes several of Bendick's comparator firms do, the entire purported disparity vanishes.

Seligman filed the Dukes case in San Francisco federal court in June 2001. The mammoth discovery process then began, which, by mid-2003, had led to the taking of 200 depositions and the collection of millions of documents.

Through that process, Seligman maintains that he unearthed evidence that "biased attitudes were at the very top of the company." Such evidence would support a disparate treatment claim - i.e., the charge that Wal-Mart intentionally discriminated - as well as a disparate impact claim.

Seligman will produce proof, for instance, that at Monday morning meetings of high-level Sam's Club executives, female store employees were often referred to as "Janie Q's." A woman executive who found the term (of unknown origin) demeaning complained about it, but others continued to use it. While some in Bentonville, Ark., might have considered "Janie Q." endearing, it will likely sound condescending and offensive to a jury in San Francisco.

(Wal-Mart tried to move the case to Arkansas, but U.S. District Judge Martin Jenkins denied the motion. "This case represents Wal-Mart's worst nightmare," Seligman says. "Their policies are going to be judged by a San Francisco jury and a San Francisco judge." San Francisco is in the Ninth Circuit, whose federal appeals court is regarded as the nation's most liberal.)

Through the discovery process Seligman also harvested more statistical data, in which his experts detected more gender disparities. The most important related to pay.

According to computations performed by Richard Drogin, a statistician who by 2003 had served as a plaintiffs expert in about 30 cases, women store employees were paid less than men across all job positions (both hourly and managerial), even though women, on average, had more seniority and better performance evaluations.

He calculated that women hourly workers were paid, on average, $1,100 less per year than men, while women managerial workers received $14,500 less. The compensation disparity claims became the most potent in the case, because they could potentially require back-pay reimbursements to every class member, not just the fraction who might have become managers.

The primary mechanism that was allegedly producing all these gaps was, once again, the granting of too much discretion to store managers, which predictably gave their unconscious bias free rein. To provide testimony on the literature relating to unconscious stereotyping and an analysis of why Wal-Mart's procedures failed to channel managerial discretion adequately, Seligman retained Bielby.

Bielby would provide testimony on one additional critical issue. Seligman needed to finesse what defendants claim is an internal contradiction in all class actions that challenge a company's granting of "unfettered discretion" to managers. If each store's local manager is being granted unbridled discretion, one might expect to see significant differences in how each store treats men and women.

Stores run by female store managers, for instance, might show less gender discrimination than those run by men. Some male managers might also be better able to resist subconscious stereotyping than others. How can a court treat 4,000 store managers as acting identically for purposes of a class action when the plaintiffs' whole theory of the case is that those store managers are being granted too much autonomy?

In a rebuttal to that sort of argument, Bielby opined in his report that "a strong and widely shared organizational culture promotes uniformity of practices throughout an organization," and that Wal-Mart had such a culture.

That could be inferred from such factors as Wal-Mart's "emphasis on the company's founder and its history, a mission statement defined by core values, frequent communication about the culture to employees," and so on. (None of those traits is unusual among big corporations, and Bielby has provided similar testimony in cases against, for instance, Home Depot, Brookshire Grocery Co., US Bancorp Piper Jaffray, and FedEx Express.)

To mount its statistical defense, Wal-Mart retained Joan Haworth, an economist who had, by then, served as an expert in more than 65 employment cases. As for the purported promotional disparities the plaintiffs experts had found - that women's presence in salaried managerial positions was far lower than their presence in the hourly workforce from which they were drawn - she found that the plaintiffs were simply failing to take into account employees' interest in such promotions.

Her data showed that women were actually being offered store manager jobs in numbers that exceeded their percentage in the applicant pool. During the years 1999 to 2002, for instance, 12% of the applicants were women, while 17% of the offers went to women.

As for the plaintiffs' claims of pay disparity, defense expert Haworth likewise reached very different conclusions from plaintiffs' expert Drogin. She claimed that Drogin's analyses did not adequately take into account crucial factors, like the number of hours worked and whether they included night-shift work, which pays more.

But her overarching criticism was that his approach amounted to pretending that a single person was making all promotion and pay decisions throughout Wal-Mart nationwide, when, according to depositions, most pay determinations were made at the store manager level or, in the case of certain specialty department employees, at the district manager level.

When she performed statistical analyses at the store and district levels, reflecting Wal-Mart's claims about the way that decisions are made, she found that 92.8% of all the standard Wal-Mart and Supercenter stores showed no statistically significant pay disparities. Of the remainder, 5.2% showed disparities favoring men, while 2.0% showed disparities favoring women.

In other words, more than 90% of class members worked at stores where women were statistically no worse off than men. Wal-Mart's argument, then, was that if a class action must be filed, it should be brought against the specific stores with disparities favoring men.

By September 2003, the case was ready for the all-important arguments on class certification, at which a judge considers whether to let the case proceed as a class action. This is often the decisive event in such a case, since disapproval of the class renders litigation impractical for most class members, while certification creates such large exposure for the defendant as to force settlement.

In the Wal-Mart case an important part of the class certification proceeding was a challenge by Wal-Mart to the admissibility of psychological and sociological testimony on unconscious bias, which it challenged as junk science.

Such challenges have been routine features of class action employment cases since at least 1997, when one was brought by lawyers for Home Depot. (Home Depot (Charts, Fortune 500) lost its bid to exclude the testimony and eventually settled for $87.5 million.)

The crux of the defense argument is always that the effects of unconscious stereotyping in the workplace are still too unproven to be the subject of helpful courtroom testimony. Defense experts argue that the plaintiffs' experts ignore or gloss over studies that cut against the plaintiffs' thesis, and that they overgeneralize from lab experiments that may have little bearing on real-world work settings.

A typical stereotyping study, for instance, might involve college students who are asked to choose the more qualified applicant from among two fictional candidates - one male and one female, say - on the basis of résumés alone. Such experiments might not predict how real managers behave in work settings when choosing among employees that they've worked with closely for months.

There's no question that a large body of scientific literature on unconscious stereotyping exists; the dispute is over what, if any, inferences can safely be drawn from it to particular workplace situations.

Wal-Mart lost its attempt to exclude the stereotyping testimony. Judge Jenkins ruled - as have most judges - that the defense experts' critiques went to the weight to be given the evidence, not its admissibility, and weight could be assessed by the jury.

Similarly, Judge Jenkins declined to resolve the numerous statistical disputes that had been presented to him. He ruled that at a class certification hearing he was not supposed to resolve such issues, which were to be deferred to a later stage of the litigation. The plaintiffs only had to present a theory that was plausible, which they had done. That's the ruling that was upheld by an appeals panel this past February, and which Wal-Mart is still appealing.

Meanwhile, the suits keep coming. In the 15 years since the Lucky Stores case, sociologist Bielby estimates that he has served as an expert witness for plaintiffs in "50 or 60" employment discrimination cases. (One of the cases is against a group of Hollywood film and TV studios, including entities owned by Time Warner, which also owns Fortune's publisher.)

During that period he has obviously encountered many companies that were vastly more sophisticated than the family-run grocery chain he first critiqued. These have included Fortune 500 companies, many of which have written antidiscrimination policies, substantial human resources operations, and in-house legal departments counseling them on how to behave.

Nevertheless, Bielby keeps finding failings in the companies' procedures. "The issues in the more sophisticated companies," he explains, "often have to do with, Are there specific vulnerabilities you can point to?"

In a race discrimination class action against FedEx (Charts, Fortune 500), Bielby critiqued a performance review system that was based in part on objective factors spelled out in employee manuals.

Bielby nevertheless found the process to be inadequate for a litany of reasons: "Judgments about the weights to be assigned to individual components of a performance dimension in order to obtain an overall rating ... are left to the discretion of individual managers"; FedEx did no auditing to ensure that managers were adhering to the written guidelines; it didn't monitor whether there were racial disparities in the outcomes of these reviews; its training of the people performing the reviews lacked "hands-on rating exercises with instructor feedback" and was not "repeated at regular intervals"; and FedEx was not holding the raters accountable for how they performed the rating task, "with real consequences for inadequate performance."

The class was certified in September 2005, and the parties announced a $53.5 million tentative settlement in April.

Just what sort of procedures would win Bielby's blessing? He concedes that no neutral procedural mechanism in itself will ever be sufficient; it must also be accompanied by auditing of the decision-making process, monitoring of the racial and gender outcomes, and holding decision-makers "accountable" for their "contributions" to diversity.

And therein lies the rub. To a pragmatic businessman, it may sound like Bielby is saying that companies should dock the pay of managers if they don't meet numerical diversity goals - creating a powerful incentive for managers to adopt surreptitious quotas and preferences. Bielby maintains that he's not recommending any such thing. The monitoring of outcomes that he's advocating is merely diagnostic testing, he insists.

As for the "accountability" piece of it, he says, "make it part of the manager's job to contribute to the company's EEO goals. And by that I don't mean quotas or the number of men vs. women promoted. [Make clear] what the company's vision is of EEO and how that translates into specific responsibilities and duties for someone that has authority to make decisions about hiring, promotion, pay, and so on. And do that in a meaningful way - that's part of their performance evaluation - so they know that it's not just feel-good talk."

SInce Wal-Mart filed its reconsideration motion in late February, both the appellate courts and the Supreme Court have issued a series of rulings that make Judge Jenkins's certification of the Wal-Mart class action appear ever more precarious. In its past term, for instance, the Court rendered two highly controversial, quite conservative 5--4 rulings in the area of antidiscrimination law.

"Neither case directly applies to our case," says Seligman, "but both cases say something deeply troubling about the direction of the U.S. Supreme Court." In the second of them, the Court struck down voluntary, race-conscious school-assignment programs adopted to achieve racial balance. "Racial balancing is not permitted," Chief Justice John Roberts wrote.

If Dukes v. Wal-Mart does reach the Supreme Court, the suit will present a blockbuster test case, with potential to change the face of contemporary employment discrimination class actions. Given their near-universal exposure to such suits, Fortune 500 companies should be praying that it does.

REPORTER ASSOCIATE Susan M. Kaufman contributed to this article. Top of page