IntroductionUnder federal affirmative action regulations, federal contractors are required to conduct an annual self-evaluation of compensation with respect to gender, race, and ethnicity. Even though the self-evaluation is required, published affirmative action regulations have been virtually silent regarding the methodology federal contractors should use to perform this annual self-evaluation. This changed in June 2006, when the Office of Federal Contract Compliance Programs (OFCCP) released its final document regarding federal contractors’ examination of compensation practices with respect to gender, race, and ethnicity.
In these self-evaluation guidelines, the OFCCP outlines an analysis method by which contractors may comply that centers around five “standards”:
1. the self-evaluation must be based on “similarly situated employee groupings (SSEGs)
2. the employer must make a reasonable attempt to produce SSEGs that are large enough for meaningful statistical analysis
3. on an annual basis, the employer must perform some type of statistical analysis of the compensation system
4. the employer must investigate any statistically significant disparities in compensation (defined as two or more standard deviations) and provide appropriate remedies
5. the employer must contemporaneously create and retain the required data and make it available to the OFCCP during a compliance review
The analysis method outlined for contractors is essentially the same analysis method that the OFCCP itself will use to issue findings on allegations of compensation discrimination. The OFCCP will make a finding of compensation discrimination if multiple regression analysis reveals statistically significant disparities in the compensation of similarly situated employees, and that statistical evidence is supported by anecdotal evidence of compenstion discrimination.
Given that the OFCCP itself will follow this analysis method, it makes sense for federal contractors to build their compensation self-evaluation program around these same five standards. In order to design such a program, a deeper understanding of the standards is required.
It should be noted at the outset that a self-evaluation program should be designed and implemented in connection with corporate counsel and outside counsel. There are a variety of confidentiality issues, attorney-clint issues, and other privilege issues that must be conidered. Legal counsel should be involved in the self-evaluation process in its entirety to protect the interests of the employer and the interest of the employees.
Standard #1: Similarly Situated Employee GroupingsThere are no definitive rules for constructing similarly situated employee groupings, or SSEGs; the OFCCP proposed the following definition: “Groupings of employees who perform similar work, and occupy positions with similar responsibility levels and involving similar skills and qualifications.
The OFCCP notes that other “pertinent factors” should also be considered in the formation of SSEGs:
…otherwise similarly-situated employees may be paid differently for a variety of reasons: they work in different departments or other functional divisions of the organzation with different budgets or different levels of importance to the business; they fall under different pay plans, such as team-based pay plans or incentive-based pay plans; they are paid on a different basis, such as hourly, salary, or through sales commissions; some are civered by wage scales set through collective bargaining, while others are not; they have different employment statuses, such as full-time or part-time.
In addition to those mentioned above, other “pertinent factors” may include geography (or some other measure of location) and business unit or department. If locality adjustments or cost of living adjustments are given to employees working in certain geographic locations, this information should be incorporated into the SSEG construction. Similarly, payroll budgets may differ by business unit or department. This, in turn, may lead to different compensation between employees seemingly performing the same tasks with similar titles and functional responsibilities.
In order for the self-evaluation to generate meaningful results, it is important that each employee’s compensation is assessed against the appropriate peer group. It would be inappropriate, for example, to compare the compensation of the CEO of the organization and the compensation of the CEO’s administrative support staff. We would expect large differences in compensation between these two groups because they have different responsibility levels, serve different purposes within the organization, etc.
It should be noted that in the event of an investigation, the OFCCP will examine the manner in which the SSEGs have been constructed. Specifically, the OFCCP will review job descriptions, the actual work performed by employees, responsibility levels associated with each job, along with the skills, abilities, and qualification of the employees. Interviews with employees and managers may also occur. The OFCCP reserves the right to make the final determination as to whether employees in the same SSEG – as constructed by the employer – are in fact “similarly situated”.
Standard #2: The similarly-situated employee groups (SSEGs) should be large enough for meaningful statistical analysisIn the construction of SSEGs, employers should also keep in mind the number of employees being grouped together. While it would be inappropriate to place the CEO and the CEO’s administrative support staff in the same SSEG, it would be equally inappropriate to place each employee in the organization in his or her own SSEG. The SSEGs should be “large enough” that a meaningful statistical analysis can be performed.
The definition of “large enough” is somewhat subjective. At a minimum, there must be more individuals being studied than there are explanatory factors. If there are more explanatory factors than there are individuals being studied, the “effect” of each explanatory factor cannot be calculated using multiple regression analysis.
Assuming that the size of the SSEG meets this minimum threshold for regression analysis, determining whether the SSEG is “large enough” becomes a question of judgment. Although there are no definitive rules, the OFCCP offers the following guidance on this issue:
“… SSEGs must contain at least 30 employees and at least 5 employees for each comparison group (i.e., females/males, minorities/non-minorities…”
The OFCCP recognizes that not every employee can be appropriately placed in an SSEG. The guidelines indicate that these individuals should be excluded from the statistical analysis, and their compensation should be analyzed by non-statistical methods (such as mean or median analysis).
However, the OFCCP cautions that statistical analyses encompassing less than 70% of the organization’s workforce would be subject to “careful scrutiny”.
The construction of SSEGs is one of the most important components of the compensation self-evaluation. Errors in groupings can render the results generated from the self-evaluation meaningless. Additionally, the SSEGs constructed serve as a memorialization of the organization’s view of its employees and the functions they serve. Legal counsel should be involved in the self-evaluation process from the onset to advise on issues of privilege, work product, discoverability, etc., to ensure that the interests of both the employer and the employee are protected.
Standard #3: An annual statistical analysis of the organization’s compensation systemThe official guidelines indicate that the statistical model used should incorporate legitimate factors that explain compensation differences within SSEGs. These explanatory factors typically include measures such as length of service, time in job, relevant experience in previous employment, education and certifications, and location. Collectively these factors are referred to by labor economists and statisticians as “edge factors”.
Some edge factors will be readily measureable – for example, lenghth of service with the employer – while others are more difficult to quantify. If an factor believed to affect compensation is difficult to quantify a proxy variable can be used. A good proxy variable is one that is easily measurable and is highly correlated with the edge factor for which it is being substituted. Caution should be exercised in the selection and use of proxy variables, as they may not truly reflect what one is intending to measure.
For example, age at hire is sometimes used as a proxy for relevant prior experience if previous employment information is not available. Age at hire is easily measurable, since hire date and date of bith date are typically maintained in human resources databases. We would expect that age at hire would be somewhat correlated with prior experience; “older” workers typically have more prior experience than “younger” workers. However, age at hire may not reflect relevant prior work experience. Further, age at hire does not consider periods of absence from the labor market for such reasons as illness, education, personal reasons, etc., The use of age at hire may introduce a gender bias into the model, as women typically experience greater absense from the labor market than men due to childbearing and child rearing. Thus, using age at hire may overstate the true revelant prior experience for some individuals.
The statistical tool commonly selected for analysis is multiple regression analysis. Multiple regression analysis is one of the preferred statistical techniques because the calculations involved are relatively simple, the interpretation of estimated gender or race “effects” is straightforward, and the entire compensation structure can be expressed with one equation.
The beauty of multiple regression analysis is that this technique estimates the effects of each factor net of all the other factors in the model. In other words, it allows one to estimate how many more dollars of compensation an individual would be expected to receive if (s)he had one additional year of length of service, holding all other factors (such as time in job, education, etc.) constant. This allows the effects to be separated out and examined individually.
When reviewing and evaluating the results of multiple regression analysis, is it important top keep two issues in mind: (1) practical significance and (2) statistical significance. Practical significance refers to the size of the estimated effect to (in this case) compensation. An estimated effect is said to have practical significance if the effect is “big enough to matter”. Statistical significance refers to whether the observed effect is the likely outcome of a gender- or race-neutral process. Unlike practical significance, there is a generally accepted “rule” for determining whether an effect is statistically significant. An observed outcome is said to be statstically significant if the probability (or likelihood) of that outcome is “sufficiently small” such that it is unlikely to occur under a gender- or race-neutral process. The commonly accepted definition of “sufficiently small” is 5%, which is equivalent to approximately 2 units of standard deviation.
The guidelines indicate that contractors with 500 or more employees must use multiple regression analysis. However, given the advantages of multiple regression analysis, employers with fewer than 500 employees should consider the use of this technique in their compensation self-evaluation.
Standard #4: the investigation of statistically significant disparities and the remedies of such disparitiesAccording to the guidelines, any statistically significant disparity (i.e., disparities of two or more standard deviations) must be investigated and appropriate remedies must be provided. This guideline has both a concrete component (two or more units of standard deviation) and components that are more subjective (investigation and remediation).
An observed compensation disparity is said to be “statistically significant” if it is unlikely that this difference occurred under a gender- or race-neutral process. The general rule of thumb used to define “unlikely” is two or more units of standard deviation. A disparity of two standard deviations means that over repeated sampling, a disparity of the size observed will occur approximately 5% of the time.
It should be noted that from a statistical perspective, any disparity that is not statistically significant is not different from a disparity of zero. That is, we cannot say that the disparity is not occurring by chance, and no inference of discrimination can be drawn.
Assuming that statistically significant disparities are observed, the guidelines provide the employer with flexibility in structuring their investigation methods and appropriate remedies. Regardless of how the employer chooses to investigate and remediate the statistically significant disparities, it is imperative that investigation precedes remediation. It may be the case that the investigation reveals an “edge factor” or other legitimate factor that was not considered in the original statistical model but nonetheless explains variation in compensation. These factors – if they exist – should be identified and considered before any changes in compensation are instituted.
The manner of investigation differs among employers, and depends to some extent upon the types of employee data available to the employer. A common starting point for investigation is interviews and discussions with managers. In some cases, these conversations will bring to light additional legitimate factors that were not originally included in the analysis. If the employer has information readily available regarding these factors, the model can be re-estimated incorporating these factors, and the results from the two models can be compared. If the employer does not have information regaring these factors, these conversations may serve as the impetus to begin collection of this necessary data.
While it is generally considered “best-practice” to involve legal counsel in the follow-up process, any remediation should only be done under the auspices of legal counsel. The implications of making remedies are significant, and legal counsel is best positioned to understand these implications and to advise the organization on appropriate remedies and implementation.
Standard #5: contemporaneous creation and retention of required dataThe guidelines require that all data used in the self-evaluation must be retained for a period of two years from the date of the analysis. Examples of the data that must be retained include:
• documentation and justification of SSEG construction
• documentation relating to the structure and form of the statistical analysis
• data used in the statistical analysis
• results of the statistical analysis
• employees excluded from the statistial analysis and reason(s) for exclusion
• data and documents used in the non-statistical analysis
• results of the non-statistical analysis
• documentation of the follow-up into any statistically significant disparities
• documentation of conclusions reached from follow-up
• documentation of any pay adjustments made to remedy any compensation disparities
Once again it should be noted that involvement of legal counsel is imperative. Legal counsel will be able to assist in document and data retention, as well as provide valuable guidance on work product, privilege, and discoverability issues.
ConclusionThe five standards outlined by the OFCCP serve as a solid foundation around which to construct a compensation self-evaluation system. When designing such a system, legal counsel should be involved in the process from its inception to protect the interests of the employer and the interests of the employee. A well-designed and well-executed compensation self-evaluation system will not only allow federal contractors to satisfy requirements of federal affirmative action regulations; it can provide the employer with a deeper understanding of how and why employees are paid. The information gained from a compensation self-audit can provide valuable insight into the organziation, illuminating the policies and procedures – both formal and de facto – used in the compensation decision-making process.