Sex Discrimination Class Actions and Merit-Based Compensation: Is Your System at Risk?

Over the past few years, the plaintiffs' bar has developed a potentially powerful theory of sex discrimination which has particular significance for merit-based compensation systems. In numerous class actions predicated on this theory, employers across diverse industries -- retail, financial services, higher education -- have paid significant sums for apparent gender disparities in pay and promotions within their ranks. For example, Merrill Lynch recently settled a sex discrimination class action for an amount reportedly in excess of $100 million.1 In 1997, Home Depot paid out $87.5 million,2 and in 2002, American Express Financial Advisers, Inc., agreed to pay $31 million3 Morgan Stanley4 and Wal-Mart5 are presently fending off similar claims. The Wal-Mart suit, if certified, will constitute the largest sex discrimination class action in history, with hundreds of millions of dollars at stake.

The Gender Gap

The new theory (which we call the "unconscious discrimination" theory) constitutes a formidable plaintiffs' tool, because it permits discrimination cases to succeed largely based on numbers. Numbers-based approaches tend to favor plaintiffs, because they obviate the need for evidence of discriminatory intent, which once was the hallmark of such cases. Further, most statistical analyses comparing men and women reveal that men still possess significant advantages in the workplace. It is well known, for example, that many occupations tend to be segregated by gender (e.g., nurses, welders),6 that occupations in which women predominate tend to be less well paid,7 that within occupations women are clustered at the lower levels of the hierarchy,8 and that women overall earn 72 percent of what men earn (the so-called "gender wage gap").9 Less clear are the reasons for such discrepancies, and who -- if anyone -- may be liable for them.

Economists have tended to analyze such gender disparities in terms of two broad categories: "supply-side" explanations, which focus on the preferences and qualifications that men and women bring to the labor market, and "demand-side" explanations, specifically, labor market (i.e., employer) discrimination. "Supply-side" explanations include so-called "pre-labor market" or societal discrimination; for example, the social influences that might cause a woman to become a secretary rather than an attorney, thus adversely affecting her earning power. In contrast, "demand-side" explanations focus on employer discrimination, where similarly situated men and women are treated differently because of gender.10

A class action lawsuit generally involves a complicated quest to untangle the respective contributions of these various influences. Through expert analyses and testimony, the parties attempt to persuade a judge or jury as to whether gender differences exist, and if they do, whether they are attributable to the employer. The analytical tool of choice is a form of statistical inquiry called a multiple regression analysis. It attempts to estimate how much of the variation in salary among employees is attributable to "legitimate" factors such as seniority, job type, degree level, degree field, years since graduation, and so forth, and how much is attributable to gender. Multiple regression analysis does not measure discrimination, however. It measures instead how much of the variation in men's and women's pay is attributable to the non-gender factors that are available in the data. What isn't attributable to the available "legitimate" factors gets assigned to gender by default. Plaintiffs then argue that all of the differences assigned to gender are the result of discrimination by the employer.

In litigating class claims, plaintiffs generally minimize the number of factors used in the regression. For example, they may ignore factors outside the workplace (e.g., total work-related experience) and attempt to discredit certain factors within the workplace, claiming that anything under the control of the employer, such as job assignment, level within the organization, or performance rating, is potentially tainted by discrimination and therefore should not be included. By reducing the number of factors used in the analysis, plaintiffs increase the differences attributable to gender. Employers, in contrast, try to include as many salary-related variables as possible, which has the effect of reducing, if not eliminating, disparities.11 From the employer's perspective, however, the analysis is complicated by the fact that its data rarely capture all relevant factors (the "missing variable" problem). For...