THE INCREASED LIKELIHOOD OF FALSE POSITIVES UNDER DIRECTIVE 307

In an ongoing series of blog posts, we will be dissecting portions of Directive 307 to highlight the potential pitfalls with the Directive’s prescriptions for conducting pay equity analyses. At issue this week is OFCCP’s oversight regarding the increased likelihood of false positives resulting from “pay analysis groups”, the new unit of pay equity analysis, as opposed to the former similarly situated employee groupings (“SSEGs”) required by the 2006 Compensation Standards and Voluntary Guidelines.

In the Notice of Final Rescission, OFCCP indicates that “the Standards require a specific technical approach that substantially increases the risk that OFCCP would fail to detect improper pay disparities.” Presumably the OFCCP is referring to the requirement that regression analyses be conducted using similarly-situated employee groupings (SSEGs). More insight is offered in the Questions and Answers to Directive 307 in which OFCCP argues that “…both the Compensation Standards and Voluntary Guidelines…fragment the analysis as much as possible, making it harder for either the agency or contractors to identify broad patterns of discrimination that cut across individual jobs.” Essentially, OFCCP is arguing that analyses conducted at the SSEG level are problematic because the SSEGs do not have large numbers of employees and the regression analysis is not able to effectively detect compensation discrimination when it actually exists. This is referred to as a false negative in the realm of statistics.

Although OFCCP’s claims regarding the limitations of SSEG-based regression analyses are debatable, perhaps the larger issue is that the alternative grouping strategy that they offer may substantially increase the potential for the opposite type of error: false positives. OFCCP notes that “a pay analysis group may combine employees from multiple job titles, units, categories and/or job groups in order to perform a pooled regression analysis, with statistical controls added as necessary to ensure workers are similarly situated.” A major problem with conducting regression analyses using groups composed of employees with very different job characteristics is that effectively controlling for legitimate, non-discriminatory explanations of pay disparities becomes increasingly difficult as the groups become more diverse. That is, the less the groups contain positions that are similar in type of work performed, skills/qualifications required, and levels of responsibility, the less likely that the legitimate, non-discriminatory variables can be effectively controlled for via statistical means. In some cases it may be impossible to control for all critical factors across a broad group. As a result, disparities due to legitimate, non-discriminatory factors will be attributed to illegal factors such as gender or race discrimination, and OFCCP will infer discrimination when it actually DOES NOT exist. Such an error is a false positive in the realm of statistics.

In its haste to reject the Bush-era Standards and Guidelines, OFCCP has introduced a grouping and analytic method that will likely result in many more incorrect conclusions of compensation discrimination than would have arisen via methods consistent with the Standards and Guidelines. In theory, fewer incorrect conclusions of no discrimination will likely be made. However, it is critical to balance the likelihood of either type of error with the potential consequences, and as such, it is reasonable for federal contractors to be concerned about OFCCP’s new approach where the agency accepts more incorrect conclusions of compensation discrimination.

We will be presenting a more detailed analysis of the false positive issue in a short webinar to be posted on the DCI website. Stay tuned!

by Kayo Sady, Ph.D., Consultant and Amanda Shapiro, M.S., Consultant, DCI Consulting Group

    Stay up-to-date with DCI Alerts, sign up here:

    Advice, articles, and the news you need, delivered right to your inbox.

    Expert_Witness_1st_Place_badge

    Stay in the Know!