Regression analysis is the statistical method of analysis most often used by OFCCP to analyze systemic patterns in compensation. Regression analysis builds upon descriptive statistics and anecdotal evidence by further isolating the comparability of the workers at issue. OFCCP compares similarly situated employees by forming pay analysis groups and controlling for relevant factors influencing their pay.
a. Similarly Situated Employee Group (SSEG). A similarly situated employee group (SSEG), is a group of employees who are comparable for purposes of analyzing a contractor’s pay practices. Note that OFCCP also uses the term “pay analysis group” and views the SSEG and pay analysis group, or PAG, as equivalent terms. An SSEG may be limited to a single job or title, and regression analysis may be performed separately on distinct units or categories of workers. Alternatively, an SSEG may aggregate employees from multiple job titles, units, categories and/or job groups to perform a regression analysis, with statistical controls added as necessary to ensure workers are similarly situated. For example, an SSEG could cover an entire unit or division of the organization but control for job title or pay grade to ensure that the analysis compares similarly situated workers. Statistical testing for practices that impact pay such as job assignment may require a different analytic grouping than testing for pay differences within a single job.
At the desk audit stage, the CO forms preliminary SSEGs by considering the industry, the types of jobs and compensation at issue, the contractor’s actual compensation practices and the available data. If a contractor provides its compensation hierarchy and job structure with its Itemized Listing submission, the CO must attempt to design its pay analysis based on that structure. In the absence of information about a contractor’s compensation system, the CO should conduct its desk audit analysis using AAP job groups as SSEGs, provided they are reasonable,73 meet the requirements of 41 CFR 60-2.12, and are of a sufficient size to conduct a meaningful systemic statistical analysis. If those conditions are not met, the CO should use EEO-1 job categories as SSEGs. Salary structures and levels or stacking lines of progression (e.g., Accountant 1, Accountant 2, etc.) may also be viable options for SSEGs. Working with BES, the CO must control further for sub-job groupings, functions, units, or titles, as appropriate. During the desk audit analysis, OFCCP will also control for tenure and full-time status as well as other factors, as appropriate.
b. Statistically Controlling for Factors. In a statistical analysis of compensation, using controls is a way of ensuring comparison of similarly situated workers and accounting for potential explanations of pay differences. Controlling for education, for instance, accounts for the effect of education on any differences in pay. OFCCP can control for observable employee factors such as education, tenure or performance, as well as institutional factors such as department or job title. It is only appropriate to control for factors in a statistical model if they have been implemented fairly and consistently applied and are relevant to the contractor’s compensation practices. At the desk audit stage, the CO should utilize a variety of desk audit tools and consult with the regional BES expert to decide on which factors to include in the preliminary analytical model.
For example, if a contractor awards bonuses based on performance, employees may be similarly situated if their performances are assessed according to the same process and criteria, and if they have similar performance ratings. Considering these facts, a statistical analysis of bonuses could compare similarly situated employees by grouping together all employees who are subject to the same performance evaluation process into one pay analysis group, and then statistically controlling for individual differences in performance ratings, as well as any other relevant factors. In some instances, the pay analysis group appropriate for an analysis of bonus pay may differ from the pay analysis group used to assess base pay.
73. FCCM 1F01.