Seyfarth Synopsis: Employers cannot ignore the recent amendments to state and local pay equity laws and increased attention on equal pay issues. Pay equity claims raise unique challenges, including the prevalence of statistical evidence and multi-jurisdictional compliance. This article addresses the advantages of conducting a pay audit and how the analysis, particularly a regression analysis, may be helpful to employers in litigation. It also discusses how an employer may use a plaintiff’s expert analysis to undermine the plaintiff’s own claim, as the Fourth Circuit addressed in a recent opinion.
Threshold Question: Should Employers Conduct A Pay Audit?
Conducting a proactive pay equity analysis is often the first and best step employers can take to ensure fair pay and diminish legal risk. Taking this step, however, should be approached with forethought and caution. Employers should make an informed decision about whether to conduct an audit.
A proactive pay equity audit is a valuable exercise when performed properly. It allows employers to identify and reduce risks, and can be used to substantiate an affirmative defense under some state-level pay equity laws. For example, the Massachusetts Equal Pay Act creates an affirmative defense to wage discrimination claims for an employer that has (1) completed a self-evaluation of its pay practices that is “reasonable in detail and scope in light of the size of the employer” within the three years prior to commencement of the action; and (2) made “reasonable progress” toward eliminating pay differentials uncovered by the evaluation. For federal contractors, evaluating pay practices on an annual basis is required, although the method for conducting the review is left up to the contractor. Moreover, conducting a pay analysis is aligned with organizational efforts to ensure equal pay in their workforces.
However, there are some key risks to be considered. If not adequately protected, an audit might be used against an employer in litigation under the federal Equal Pay Act or Title VII, which do not provide a similar affirmative defense. Thus, employers should work with counsel in order to protect the assessment process and results with the attorney-client privilege. Without these protections, a self-evaluation (and any wage differentials identified by it) may be discoverable in the event of a lawsuit. Employers should protect the audit at the outset and make an informed decision as to whether to waive the privilege in subsequent litigation. Counsel with experience and expertise in pay equity matters can also play a valuable role in shaping the scope and procedure for an audit to maximize its utility in identifying disparities that may become legal disputes and to ensure that the work product generated by the audit will make for effective evidence, if it is ever needed for use in court.
Do You Track the Data You Need For A Pay Audit?
Employers will, of course, need pay and demographic data to conduct an audit. This is typically readily available in HR information and payroll systems.
Other data points that could be used to explain differences in pay under the applicable federal and state equal pay laws are often not fully captured in employers’ information systems. This includes details about employees’ education, certifications and training, and prior relevant experience.
The federal Equal Pay Act – and many state equivalents – provide that employers may not pay unequal wages to employees in different protected classes who perform jobs that require equal (or, in the instance of some state laws, substantially similar or comparable) skill, effort and responsibility. Employers also sometimes lack the data needed to fully determine which jobs should be compared because of the “skill, effort and responsibility” involved. For example, “responsibility” may be measured by data not typically tracked in electronic information systems, such as amount of budget managed or the authority to execute legal documents.
While these are important limitations and employers would benefit from reviewing their data sources and discussing potential gaps in their data with employment counsel as part of a pay audit – and, indeed, we will delve more deeply into the issue of “data gaps” in future blog updates – do not let the perfect be the enemy of the good. There are often well-established proxies for some of the data points that could be missing. For example, for a proactive pay analysis, using age at date of hire as a rough “proxy” for prior experience is a common, and well-established practice.
While it is essential to consider these data gaps, a proactive pay equity analysis can still be extremely beneficial to identify employees whose pay can then be further evaluated. Even employers without perfect data – and our experience is this is almost all employers – can still benefit from a proactive pay assessment.
When Should Employers Use A Regression Analysis?
An employer’s selection of pay audit method depends on the scope and objectives of the review, including the number of positions and budget. It also depends on whether litigation is considered to be likely, and thus whether the method will be challenged in court.
Large employers that conduct a self-evaluation with the assistance of a professional labor economist typically perform a multivariate regression analysis. A regression analysis is a statistical technique used to model an organization’s compensation system based on data regarding factors expected to influence pay and determine to what extent gender or other protected characteristics may influence employees’ compensation. This is considered the “gold standard” in pay equity evaluations. If the pay difference between men and women measured for a group of employees has a high probability of occurring by chance alone, then the result is not considered “statistically significant.” However, when the size of the measured pay difference has a small probability to have occurred by chance, the result is considered “statistically significant.”
Social scientists, labor economists – and the Supreme Court – generally deem results as statistically significant at approximately two standard deviations (i.e., 1.96) or higher. A finding of 1.96 standard deviations (assuming a “normal distribution” manifested by the familiar bell curve graphic) indicates that a given pay difference would be expected to occur by chance 5% of the time if pay was set in a gender (or race)-neutral environment and if the grouping is appropriate and the regression model correctly incorporates all of the legitimate, business-related determinants of pay. Courts have approved this standard in employment discrimination cases. See e.g., Adams v. Ameritech Servs., Inc., 231 F.3d 414,424 (7th Cir. 2000) (noting that in employment discrimination cases, “[t]wo standard deviations is normally enough to show that it is extremely unlikely … that [a] disparity is due to chance.”); Cullen v. Indiana Univ. Bd. of Trustees, 338 F.3d 693, 702 (7th Cir. 2003) (explaining in Equal Pay case that “generally accepted principles of statistical modeling suggest that a figure less than two standard deviations is considered an acceptable deviation”).
A regression analysis is widely accepted by courts as reliable, is easily customized, and is an effective way to isolate the association of gender (or race) and compensation. However, it cannot be used to analyze job groups with few employees (typically fewer than 20-30) or heterogeneous groups that do not include at least a critical mass of employees of each gender (or race).
Other common methods are an average pay ratio (“APR”) (sometimes referred to as the “adjusted pay gap” or “adjusted pay difference”) and a cohort study. APR is a calculation of the average pay of women, compared to the average pay of men, conducted in groupings that may range from certain selected business units to an entire organization, after controlling for factors that are relevant to employee compensation.
Finally, a cohort study is a comparison of employees within a narrow group. It does not require statistical analysis and thus is less costly, but it typically includes some inherently subjective assessments and thus may be more difficult to defend in litigation. Also, it typically takes significantly more person-hours to evaluate pay using the cohort method.
Thus, employers often use a regression analysis for larger job groups, supplemented by a cohort analysis for smaller groups.
Regression Analysis As Evidence In Pay Equity Cases
Regression results can be helpful in defeating equal pay cases. Courts have dismissed claims under the Equal Pay Act when the evidence shows no systemic discrimination, i.e., no statistically significant differences in pay based on gender. See e.g., Spencer v. Virginia State Univ., No. 3:16CV989-HEH, 2018 WL627558, at *10 (E.D. Va. Jan. 30, 2018), aff’d, 919 F.3d 199 (4th Cir. 2019). In Spencer, a sociology professor claimed that she was paid less than male colleagues in other departments. The court entered summary judgment for the University, noting that “the regression analysis performed by … Plaintiff’s own expert, makes clear that VSU did not suffer from systemic, gender-related wage disparity,” and noting that the plaintiff had failed to point to any male comparator who earned more. The court explained that “[w]hile the lack of systemic discrimination, standing alone, may not be sufficient to disprove an EPA violation, … the absence of systemic discrimination … combined with … improper identification of a male comparator suggests a failure to establish a prima facie case.” Affirming, the Fourth Circuit explained that the plaintiff’s expert’s failure to uncover any statistically significant disparity within each school of the university undermined Plaintiff’s claim. 919 F. 3d at 206.
The Spencer case notes one limitation of a statistical model in defending individual pay discrimination claims: the absence of a statistically significant group-level disparity does not preclude the possibility of individual employees claiming that their compensation was lower than that of a particular comparator of the opposite gender. However, a regression analysis that also includes an individual-level assessment by providing lists of employees who are “outliers” as to pay, allows employers to review and address the compensation of individual employees who may raise pay equity issues, even if they are in groups that show no disparity.
Finally, as to the law in Massachusetts and other laws in places like Oregon that provide an affirmative defense or a partial affirmative defense for employers who conduct reasonable audits, there is little guidance as to what is “reasonable.” Employers conducting audits should ensure the audits are as comprehensive in scope as the data allows, based on a methodology vetted by appropriate legal and economic experts. Employers should take special care at the outset of the audit in determining appropriate groups of employees for comparison purposes. And in light of the limitations of regression analyses, employers should also consider including an individual-level assessment of employee pay.
Conclusion
A regression analysis that finds no statistically significant difference in pay on a systemic basis and also includes an individual-level assessment is helpful for a defense to a pay equity claim. Employers considering whether to conduct an audit should do so only under the protection of the attorney-client privilege, so they can examine whether to waive the privilege and rely on the results in litigation.
For 20 years, Seyfarth’s Pay Equity Group has led the legal industry in fair pay analysis, thought leadership, and client advocacy.
As we reflect on the developments in equal pay laws and litigation in the past year, we continue to see a legal landscape that is rapidly evolving. To keep you up-to-date, we have created an Equal Pay-focused blog series to disseminate this information.