On Tuesday, February 18, 2020 the National Industry Liaison Group (“NILG”) and the Office of Federal Contract Compliance Programs (“OFCCP”) held a Compensation Roundtable at the US. Department of Labor in Washington D.C. Director Leen introduced the program by emphasizing the need for advancement of females to higher level management positions, the right to parental leave for both men and women, and his continuing commitment to advancing his administration’s agenda. You may recall that on February 3, 2020 President Trump announced his intention to nominate Director Leen to be the next Inspector General at the Office of Personnel Management. Director Leen emphasized this will not impact his current agenda at OFCCP and that he expects OFCCP to issue many new directives or guidelines while he remains at the Agency, including:
- a Mediation Directive designed to assist the Agency and Contractors in resolving disputes before enforcement referral;
- a new Transparency Directive to provide for further disclosures of OFCCP findings prior to the issuance of Pre-Determination Notices;
- an Efficiency Directive supporting prompt and effective resolution of compliance reviews.
Director Leen also noted that the Agency remains “focused on focused reviews.” Contractors already have been selected for the next round of reviews, but the Agency still is working on preparation of the announcement letters.
Director Leen also shared that the Agency will be issuing new compensation guidance that he hopes will incorporate some of points to be made panel at the Compensation Roundtable.
The remainder of the event consisted of a panel discussion among a number of contractor representatives, including attorneys, labor economists and statisticians, and OFCCP representatives Robert LaJeunesse, the Branch Chief for Expert Services/Acting Director of Enforcement and Labor, and Economist David Garber.
All of contractor representatives acknowledged that much good has come from Directive 2018-05, OFCCP’s current Compensation Directive. For example, the panel fully supported the Directive’s:
- recognition of the importance of analytical differences between modeling for base v. other forms of compensation;
- focus on contractor’s own pay systems;
- emphasis on increased predictability and transparency;
- recognition of need for the use appropriate control variables,
- approval of outlier review; and
- use of age as a proxy for experience in appropriate circumstances.
The “Not So Good”
However, the panel also focused on the “not so good” side of Directive 2018-05 and its enforcement by OFCCP. The contractor representatives criticized OFCCP’s narrow focus on regression as the only tool for compensation analysis. This focus has caused OFCCP to force the aggregation of non-similarly situated employees into the same Pay Analysis Group (“PAG”) for the sole purpose of ensuring adequate pool size and statistical power. For example, OFCCP has formulated rules including that PAGs must have at least 30 employees, at least 10 employees for every variable included in the model, and at least 5 observations for every variable. Through strict adherence such rules OFCCP has inappropriately aggregated non-similarly situated employees, dropped variables that help explain pay variability, and combined dissimilar variables that are not appropriately interacted. The result is that the Agency: (1) creates PAGs that do not conform to Title VII of the Civil Rights Act of 1964’s mandate that comparisons be made only of similarly situated employees; (2) ignores important aspects of a contractor’s own compensation systems; and (3) combines dissimilar levels of categorical variables.
Other issues raised by the panel include OFCCP’s lack of a definitions related to indicators of systemic discrimination, practical significance and tainted variables. For example the panel posited that OFCCP is inappropriately assuming that mere “correlation” is enough to show a variable is tainted. In fact, OFCCP needs to show, among other things, how that a variable is related to a specific pay process that discriminates to establish that a variable is tainted.
To address these concerns, the contractor representatives recommended that OFCCP should:
- use regression analysis only when appropriate, and use other forms of analyses, such as rank sum, T-Tests adjusted for degrees of freedom, matched pair/cohort analyses, a combination of such tools, when not.
- carefully evaluate a contractor’s pay system and include what it learns in its models – i.e., how does the contractor pay; what factors does it consider, etc.;
- group similarly-situated employees carefully and in accordance with Title VII; OFCCP should not aggregate, add or drop groups or variables solely for the purpose of statistical power;
- accept the contractor’s units of analysis if at least 70% of the facility’s workforce is covered by some type of legitimate compensation analysis;
- provide a specific definitions related to systemic discrimination and tainted variables; and
- adopt the following rules for practical significance:
- if there is a less than a 2% pay disparity, regardless of whether statistical analysis results are significant, there should be no further review.
- if there is 2 to 5% pay disparity, OFCCP should review differences with understanding that result may not be practically significant.
Bob LaJeunesse and David Garber responded to many of the items noted above. Both emphasized that OFCCP is investigating for indicators of systemic discrimination, which are often not isolated to a single group of similarly-situated employees. According to OFCCP, its focus on systemic issues necessitates a broader view of PAGs, especially in cases where trend or other analyses show a pattern favoring one group over another. If there are a series of extremely small units of analysis, which show such a trend, then OFCCP cannot commit itself to using such small groups as PAGs because such tests do not allow one to control for legitimate differences in pay. In sum, OFCCP claims it is balancing the similarly situated standard with the “need” for statistical power.
In defending such practices, Mr. LaJeunesse and Mr. Garber emphasized that the Agency models its regression analyses in a very careful and deliberative way. For example, OFCCP uses a “nearest nesting” approach to avoid the aggregation of dissimilar variables. Mr. Garber stated that it is “common place in social science statistics to engage in variable fusion,” such as the combination of performance ratings into larger groupings. They both further emphasized that OFCCP looks at more than just differential distributions is determining a tainted variables and that its models “follow what we learn, not the other way around.”
The panel discussion succeeded in highlighting several points of disagreement between contractors and OFCCP regarding the best means to investigate potential compensation discrimination. At the same time, certain points of agreement were reached during the discussion. For example, both sides agreed that pay discrimination is an important societal issue that the Agency should be investigating and remedying. In addition, all agreed that contractors and OFCCP alike benefit from transparency and consistency when it comes to the process and the methods of compensation analysis.
At this point the Agency committed to taking the recommendations under review and continuing the dialog.