Statistical Issues in Employment Litigation By Steven W. Moore1 Statistics can play a key role in the adjudication of an employment litigation matter, particularly in the area of class and collective actions. In the area of discrimination law, the U.S. Supreme Court stated, “[o]ur cases make it unmistakably clear that ‘[s]tatistical analyses have served and will continue to serve an important role’ in cases in which the existence of discrimination is a disputed issue.”1 However, statistics can also be used by employers to prevent conditional certification or otherwise decertify a collective action under the Fair Labor Standard Act. This paper addresses some of the common themes and emerging developments regarding the use of statistics in employment class and collective actions. Discrimination Theories Although a number of theories have been used to litigate discrimination class claims, plaintiffs typically file these claims under two primary theories alleging violations of Title VII of the Civil Rights Act of 1964 and other related civil rights statutes. First, disparate treatment claims alleging intentional discrimination on a class-wide basis generally follow the pattern and practice theory first articulated by the Supreme Court in International Brotherhood of Teamsters v. United States, 431 U.S. 329 (1977), wherein the discrimination was alleged to be so pervasive that it was the standard operating procedure of the employer. If the class plaintiffs succeed in making a prima facie showing of a pattern and practice of discrimination, then the employer 1 Steven W. Moore (steven.moore@ogletreedeakins.com) is a shareholder in the Denver office of Ogletree, Deakins, Nash, Smoak & Stewart, P.C. must rebut that showing with evidence showing that the class plaintiffs’ statistics were either inaccurate or insignificant. Additionally, many discrimination class actions are filed under the disparate impact theory described by Griggs v. Duke Power Co., 401 U.S. 424 (1971), in which a neutral practice or policy was alleged to have adversely impacted a protected group. Under this theory, the class plaintiffs must make a prima facie showing that an employer’s neutral practice has a disparate impact on racial minorities, women, or other members of a protected group. Once the class plaintiffs make this statistical showing, the employer must show that the practice is job-related for the positions at issue and consistent with business necessity. The class plaintiffs would then have the opportunity to prove either (1) the practice is not job-related; or (2) the availability of another selection procedure that does not have the same adverse impact on the protected group. Class discrimination cases, whether filed under the disparate treatment or disparate impact theory, will devolve into situations with dueling experts debating whether there has been a statistically significant showing of some type of adverse treatment or impact in the hiring process. Some employers have successfully obtained court orders striking plaintiff’s statistical expert because that expert failed to control for relevant, non-discriminatory explanations for the apparent statistical disparities such as educational background, work history, licensure, certifications, and the proper labor market from which the employees at issue were drawn. While many class action claims relating to the hiring process are filed by private plaintiffs, employers may also face enforcement actions by the EEOC (or its state equivalents) alleging similar theories. Indeed, through its Eradicating Racism and Colorism From Employment (E-RACE) and Systemic Initiatives, the EEOC has announced ambitious plans to bring enforcement actions against employers using facially neutral criteria that have a disproportionate adverse impact on racial minorities. According to the EEOC, these criteria include arrest and conviction records, employment and personality tests, credit scores, and computer programs that flag residential addresses and zip codes. The EEOC has also stated that it will enforce gender-based equal pay laws through directed investigations and Commissioner Charges. Statistical Models In employment discrimination cases, experts often use statistical models focusing on “selection rate” comparisons, “population/workforce comparisons,” and multiple regression analyses.2 The selection rate methodology compares “the percentage of the complainant’s protected group that pass (or fail) a given test; met (or do not meet) a given employment requirement; or are (or are not) actually hired, promoted, or otherwise selected with the corresponding success (or failure) rate for the majority group.”3 Multiple regression is a methodology that estimates the effect of several independent variables (e.g., education, experience, performance) on a single dependent variable (e.g., salary). This methodology estimates the extent to which a particular independent variable (e.g., gender) influenced the independent variable (e.g., pay rate) – in other words, whether discrimination or other factors that caused the disparity.4 Some courts will also use what is known as the “four fifths rule,” described in the EEOC’s Uniform Guidelines on Employee Selection Procedures. Under the four fifths rule, “a selection practice is considered to have a disparate impact if it has a selection rate for any race, sex, or ethnic group which is less than four-fifths (or eighty percent) of the rate of the group with the highest rate.”5 However, this methodology, often described as a rule of thumb analysis, has been criticized because it can produce inherently unreliable results in some situations particularly with small sample sizes. In promotion and unequal pay cases, some experts may also utilize a “cohort” analysis, which analyzes the progression of a group of employees who started together at some specific level or point. The advancements in compensation and promotions are evaluated over a period of time between the cohorts.6 The Expert’s Opinion Must Be of Legal Significance In order for a statistics-related opinion to be admissible in court as probative evidence, the opinion need not reach the level of scientific certainty, but rather, it must be of legal significance.7 In order to meet this standard, the opinion “must take into account the appropriate labor pool and account for significant nondiscriminatory factors, typically by using multiple regression analysis.”8 For example, in a hiring case alleging racial discrimination, the expert must first determine the appropriate labor pool. In a situation where racial minorities have not been discouraged from applying for employment, the appropriate labor pool may be drawn from the actual applications for employment submitted in response to a job posting. This would enable the expert to analyze hard data to see if there is any disparity in hiring patterns between whites and racial minorities. The expert would also need to control for other significant, nondiscriminatory factors such as education and work experience, depending on the occupation at issue. Indeed, the statistical analysis would be skewed if it factored in applicants who did not have the requisite experience for the available positions. In situations where applications for employment cannot be used for determining the appropriate labor pool (such as where racial minorities were discouraged from applying for employment9 or where all applicant information was not retained), statistical experts employ other techniques for determining the appropriate labor pool and analyzing comparable or similarly-situated employees from different protected groups. For example, in Teamsters v. United States, a statistical model was used to show a gross disparity in the percentage of African Americans in the population with the percentage of African American drivers employed by the defendant employer.10 This type of comparison is best used where unskilled, entry-level jobs are at issue. Indeed, in Hazelwood School District v. United States, the Supreme Court cautioned that, “When special qualifications are required to fill particular jobs, comparisons to the general populations (rather than to the smaller group of individuals who possess the necessary qualifications) may have little probative value.”11 Instead, where the employment position at issue requires certain skills or qualifications, a statistical model should compare, for example, the racial composition of those who are qualified in the labor market (rather than the general population) to those holding the jobs at issue.12 In other words, the proxy pool must be that of the local labor force possessing the requisite skills for the position at issue.13 This information may be derived from census data and unemployment data.14 Note, however, that applicant flow data, if it is available, is the preferred source for a statistical study.15 Mirroring the Business Model – Decentralized or Centralized In constructing a statistical model, the expert should fully understand the business model that an employer utilizes in undertaking the employment practice at issue. For example, if there is a class action challenging a company’s promotion practices on a nationwide basis, the expert must first determine where and by whom the promotion decisions are being made. If the decisions are being made out in the field on an office-by-office basis by local management, the statistical model must analyze the statistics on a disaggregated basis, controlling for each office. On the other hand, if the business model is such that the actual decision making occurs from a centralized location, such as the company’s headquarters, then an aggregated statistical analysis may be appropriate. The leading case on this statistical concept is Wal-Mart v. Dukes16 where the Supreme Court rejected the statistical opinion proffered by the plaintiffs. There, the plaintiffs launched a nationwide gender discrimination class action, challenging Wal-Mart’s hiring and promotion practices which are carried out on localized and discretionary basis. In rejecting the opinion, the Supreme Court explained, “[i]nformation about disparities at the regional and national level does not establish the existence of disparities at individual stores, let alone raise the inference that a company-wide policy of discrimination is implemented by discretionary decisions at the store and district level.” Controlling for Nondiscriminatory Explanatory Factors As mentioned above, a proper statistical model takes into account nondiscriminatory factors that are used in the decision making of the personnel action at issue. There are, of course, situations where plaintiffs’ experts fail to control for certain nondiscriminatory factors to achieve desired results. As one labor economist previously noted, “An analysis that includes a very large number of decisions without considering the job-related factors that affected the decision is very likely to result in an adverse difference that is not consistent with policies that produce parity.” 17 Some plaintiffs’ experts fail to account for various explanatory variables that include, for example, the following: (1) differences in qualifications; (2) performance; (3) education; (4) work experience; (5) job title or rank; (6) licensure; and (7) skill or specialty. If a plaintiff’s expert fails to take into account important variables considered in the decision making process, the opinion may be found to be insufficient to establish the plaintiff’s prima facie burden in a class case. Note, however, that some plaintiff-side experts will argue that certain variables are tainted and should not be considered if they are the product of direct or indirect discrimination. Fair Labor Standards Act Theories In recent times, collective actions under the Fair Labor Standards Act of 1938 have been one of the fastest growing areas of representational litigation, far outpacing discrimination class actions in terms of case filings across the nation. In a typical Rule 23 class action, the battle is often won or lost at the class certification stage because defendants cannot fathom the risk of losing a class case at trial or because class counsel has lost the incentive to continue in a case if the theory relies heavily on the use of costly experts. However, because of the lenient standard in conditionally certifying FLSA collective actions, the real battle in those cases is at the end of the discovery period when the employer files a motion to decertify the collective action. In FLSA collective action litigation, employee challenges have focused on two primary theories. First, employees often allege that they are misclassified as exempt when in fact they are really non-exempt employees who are owed overtime compensation for hours worked in excess of 40 in a given workweek. These claims are generally premised upon the theory that the employees at issue do not fall within one of the white-collar exemptions for executive, administrative, or professional employees, or the theory that they were not paid on a true salary basis as the FLSA requires. Second, non-exempt employees often allege that they have not been paid overtime for off-the-clock compensable work, ordinarily in the context of pre-shift and postshift work or work performed during meal and rest periods. In recent times, there has been a steady increase of “donning and doffing” cases where employees claim that the time spent putting on and taking off their work-related clothing and/or other gear is compensable time under the FLSA. While donning and doffing theories first became popular in the poultry industry, the theory has been expanded to other industries where workers customarily wear specialized protective gear. Statistical Issues in FLSA Litigation Employers are beginning to turn to economists and statisticians to present expert testimony in support of motions to decertify FLSA collective actions. These experts, in an off-the-clock case, can analyze certain data that may exist in the work place that measures time – such as the use of computers, phone systems, security systems, door entry systems, and other data that records a time stamp of entry, use, or access. In off-the-clock litigation, economists and statisticians also often analyze payroll data to explain “how specific wage and hour practices are applied to different group of workers,”18 which, in turn, can be used to decrease the size of the class or decertify it altogether. For example, payroll data can explain variations in pay practices from various regions or districts of a company. Such an expert opinion may be used to disabuse the notion or allegation that practices occurring in a small number of districts or regions are also applying companywide. Additionally, in exemption misclassification cases, some experts will analyze data that has been gathered through workplace surveys, job analyses, or work process studies. This data may reflect the proportion of time spent on certain job duties (exempt vs. non-exempt), which, in turn, can be used to prove that the subject employees are properly classified as exempt executive or administrative employees, for example.19 1 Teamsters v. United States, 431 U.S. 324 (1977). Employment Discrimination Law (5th Ed.) 35-17. 3 Employment Discrimination Law (5th Ed) 35-18. 4 Employment Discrimination Law (5th Ed) 35-28. 5 Stout v. Potter, 276 F.3d 1118, 1124 (9th Cir. 2002). 6 Employment Discrimination Law (5th Ed) 35-34. 7 Bazemore v. Friday, 478 U.S. 385, 400 (1986); Employment Discrimination Law (5 th Ed.), 35-12. 8 Employment Discrimination Law (5th Ed.), 35-12. 9 Dothard v. Rawlison, 433 U.S. 321 (1977). 10 431 U.S. 324 (1977). 11 433 U.S. 299, 309 (1977); see also, Wards Cove Packing Co. v. Atonio, 490 U.S. 642, 650-51 (1989). 12 See e.g., Scales v. Slater, 181 F.3d 703, 709 n.5 (5th Cir. 1999); Carney v. City and County of Denver, 534 F.3d 1269, 1275-76 (10th Cir. 2008). 13 Moore v. Hughes Helicopters, Inc., 708 F.2d 475, 482-83 (9th Cir. 1983). 14 Long v. City of Saginaw, 911 F.2d 1192 (6th Cir. 1990); Phillips v. Joint Legislative Comm., 637 F.2d 1014 (5 th Cir. 1981). 15 Malave v. Potter, 320 F.3d 321, 326 (2d Cir. 2003) (“in the context of promotions, we have held that the appropriate comparison is customarily between the composition of candidates seeking to be promoted and the composition of those actually promoted.”) 16 131 S. Ct. 2541 (2011). 17 Joan G. Haworth, Ph.D. “Making Decisions in the Current Employment Environment: The Role of Economic and Statistical Analysis.” ERS Group. 18 Charles J. Mullin and Bo S. Shippen, “Leveraging Employment Data To Fight Class Certification,” Law360 (June 19, 2014). 19 Joan G. Haworth, Ph.D. “Making Decisions in the Current Employment Environment: The Role of Economic and Statistical Analysis.” ERS Group. 2