On August 24, 2018, OFCCP released its long-awaited Directive outlining the standard procedures OFCCP will use during the course of compensation audits of federal government contractors. Concerned that it failed to provide clear guidance to contractors, the Directive rescinds and updates Directive 307, OFCCP's most recent guidance concerning compensation audits. OFCCP explained that it "believes that fulsome guidance will further support contractors' ability to conduct meaningful self-audits so that they can proactively identify and address issues with their compensation practices." The Directive and its attachment and the FAQs are attached here.

OFCCP explained that the purpose of the new Directive was to (1) provide further transparency to contractors regarding the procedures used by OFCCP in conducting compensation audits; (2) support compliance and self-analysis by contractors; and (3) improve compensation analysis consistency and efficiency during compliance evaluations. The Directive was intended to provide more transparency to contractors regarding OFCCP's procedures and practices for establishing pay analysis groups (PAGs) and conducting statistical analyses and modeling.

OFCCP reiterated in the attachment to the Directive that it is focused on identifying and addressing compensation disparities that are the result of systemic discrimination, including pattern or practice discrimination and disparate impact discrimination. If OFCCP believes there are indicators of systemic discrimination, it will seek to understand the contractor's compensation system, policies and practices and undertake what it characterizes as a "holistic" review of the contractor's EEO, diversity and inclusion policies. The attachment explains that, in conducting such a holistic review, it will consider a variety of employment practices that can lead to compensation disparities between similarly-situated employees, including the promotion, training and advancement opportunities, and the assignments offered to employees.

OFCCP also described the categories of evidence it will consider during compliance reviews, including statistical and anecdotal evidence. Anecdotal evidence can be derived from the review of policies, compensation records and other documents and interviews with managers and workers. OFCCP explained that, in determining which cases to pursue, OFCCP will be less likely to pursue a matter if the statistical evidence is not corroborated by non-statistical evidence, unless the statistical evidence is especially strong or OFCCP has identified a pattern of compensation discrimination by the contractor in prior audits.

OFCCP will continue to use the OMB-approved scheduling letter to collect initial compensation data from contractors. Compliance officers will use the information provided by contractors to develop pay analysis groupings (PAGs) of comparable employees. According to the attachment to the Directive, it is OFCCP's objective to use or create PAGs that mirror the contractor's compensation system. It notes that, if a contractor provides information concerning its compensation hierarchy and job structure in the submission, OFCCP will attempt to design its analysis based on that structure. OFCCP expects the information provided by the contractor concerning its compensation structure to be reasonable and verifiable and will include groups of sufficient size to undertake a meaningful statistical analysis.

OFCCP declined to identify in the Directive the size of groupings it considers sufficient. However, in the FAQs issued with the Directive, OFCCP states that it will first review each of the PAGs to determine if they contain at least 30 employees under a similar pay system performing similar job functions. It will then try to determine if there are at least 10 employees per variable (e.g., gender, years in position). Thus, if a pay model had five control variables, the PAG would ideally have at least 50 employees. It also noted in the FAQs that, in rare circumstances, OFCCP may rely on non-statistical, or cohort, data to assess small job groups. If a contractor declines to provide proposed groupings or the information necessary for OFCCP to establish PAGs, OFCCP will conduct its preliminary desk audit based on either EEO-1 or AAP job groups, if they are reasonable and sufficiently large.

After it identifies appropriate PAGs, OFCCP will then control for factors that impact pay within the PAGs, such as sub-job groupings, functions, units and titles, as appropriate. During this preliminary analysis, OFCCP will also control for tenure, full-time status and other appropriate factors. If the results of the desk audit warrant additional review, OFCCP will seek additional information to understand the contractor's compensation systems, factors that drive compensation decisions, and job structure. Based on this additional information, OFCCP may broaden or narrow its preliminary PAGs.

OFCCP identified principles that it will use to conduct statistical analyses, both during and after the desk audit phase:

• It will use multiple linear regression analysis (MRA) to help reduce false negative and positive results;

• It will analyze (1) base pay, (2) total compensation and, if necessary (3) individual components of pay (e.g., bonuses or commissions);

• It will convert salary to a log of salary using a regression model; and

• It will analyze statistical outliers to test whether the PAGs have been composed properly.

In analyzing pay disparities using these models and principles, OFCCP will evaluate males and females in separate regression models. OFCCP noted, however, that it may evaluate the interaction of sex and race in future models. For race and ethnicity, OFCCP will create a variable for each race and ethnicity category with more than five employees. OFCCP provided the following guidance concerning the factors it will apply in undertaking a MRA:

• It will control for components of employee tenure (e.g., time in position, time with the company) separately;

• It will not consider squared terms (factors, such as time in position, that level out over time) until after the initial desk audit phase;

• It will control for education categories and performance ratings and rakings, combining them as necessary;

• It may use age as a proxy for prior experience only at the desk audit phase;

• Per the FAQs, it may control for job title if pay legitimately varies based on job title.

• It will control for job level or grade, combining them as necessary; and

• It will evaluate market salary surveys, if provided by the contractor, on a case-by-case basis if job and pay differentials are not sufficiently accounted for.

Before OFCCP will apply any of these (or other relevant) factors, it will test them for neutrality to make certain they do not have a disparate impact on any protected category. OFCCP also adopted the "rule of five" recently used by the National Office statisticians in compliance reviews and conciliation negations. It described this rule in the FAQs in discussing the circumstances when job title would be a relevant factor:

OFCCP will attempt to control for . . . job title if pay legitimately varies by job title. In many instances, controlling for factors like grade level, department, or business unit sufficiently distinguishes functional differences in job titles . . .. To capture meaningful pay differentials across the categories, OFCCP requires that each category contain at least five observations. If a category has fewer than five observations, OFCCP will join those observations with their ordinal counterpart (e.g., nearest grade or level) or to the category with the nearest average pay. With this approach, OFCCP meaningfully controls for pay differentials across job titles while minimizing the risk of suturing the model with low frequency employee controls.

At the end of the attachment, OFCCP identifies three practices it will use to facilitate transparency and consistency and the resolution of findings through conciliation:

1. At the conclusion of the desk audit, OFCCP will notify the contractor in writing of the general nature of any preliminary compensation disparities. For instance, a contractor may be informed that preliminary disparities have been found in "compensation policies and practices with respect to women in production, sales and management."

2. Consistent with Directive 2018-01, OFCCP will attach to any pre-determination notice (PDN) the individual-level data necessary for the contractor to replicate OFCCP's PAGs and regression results. The contractor will have the opportunity to offer a non-discriminatory explanation for the preliminary findings before a violation will be found. The PDN and the contractor's response will be reviewed by the National Office. The individual-level data will also be attached to any Notice of Violation (NOV) issued to the contractor after the National Office's review is completed.

3. To facilitate conciliation, OFCCP will include professional labor economists or statisticians from the Branch of Expert Services in the conciliation process to clarify OFCCP's variable coding, statistical methods and findings and to answer questions about the process and assumptions used in computing back pay.

With respect to point 2, OFCCP explained in the FAQs that, if a contractor provides factors that may explain the disparities in pay, OFCCP's statisticians and other field and national office staff will coordinate with representatives from the DOL's Regional Office of the Solicitor to decide on a preliminary analytical model. The contractor will be given an opportunity to provide additional information to OFCCP regarding its compensation systems and practices. Based on the additional information provided by the contractor, the preliminary model may be refined.

The Directive went into effect on August 24, 2018. It will apply to all compliance evaluations scheduled on or after that date. The Directive states that it will apply to "open reviews to the extent they do not conflict with OFCCP guidance or procedures existing prior to the effective date" of the Directive.