Q&A: Stay in Front of Pay Bias Claims

Pay Audits_Dollar Bill

It will be interesting to see if pay audits on the “outliers” within an organization and a systemic approach to verify factors affecting salaries will emerge as topics of discussion at an upcoming hearing on the Equal Employment Opportunity Commission’s controversial proposal requiring employers to disclose compensation data.

On Feb. 16, the EEOC announced it will hold a public hearing on March 16 at its headquarters in Washington to solicit comments and information from stakeholders on a proposal to revise the annual Employer Information (EEO-1) Report to seek aggregate data on pay ranges and hours worked. The proposed changes are aimed at addressing “discriminatory pay practices,” the EEOC said.

The measure also replaces the Office of Federal Contract Compliance Programs’ Equal Pay Report proposal, requiring federal contractors to submit summary pay data on their employees, classified by race, sex and ethnicity. Public comments on the proposal are due by April 4. If everything goes as planned, employers will start to submit compensation data in EEO-1 Reports for the 2017 filing cycle.

Downtime Ripe for Review 

Meanwhile, as the commission gathers and hears public comments on changes to the report, companies may want to brush up on the “hows and whys” of their compensation practices and systems.

“Employers are best able to defend themselves against [pay discrimination claims] if they have a comprehensive understanding of the compensation decision-making process and if they have accurate and reliable data on the factors that affect compensation,” labor economist Paul F. White of Resolution Economics, LLC told Bloomberg BNA. In February, he weighed in on the proposed revisions to the EE0-1 Report shortly after they were published.

In the last installment to a three-part blog Q&A series, White, a specialist in statistical analysis of employment practices, discusses compensation reviews focusing on “outliers” and the importance of clean data for compensation analyses.  

A Partial Deep Dive May Avert Claims

Bloomberg BNA: Federal contractors can perform a self-audit under attorney-client privilege on their compensation practices before facing a government audit. You have said in the past that certain contractors may want to first conduct an analysis of pay data on the “outliers” who are similarly-situated employees with similar jobs, instead of launching a full-scale pay audit on large groups of employees. How would this type of audit work?

White: While it is certainly no substitute for a statistical analysis, employers can gain a lot of useful information about potential compensation issues by examining both “positive outliers” and “negative outliers” in their employee ranks. 

As an example, this can start with the employer looking at its data for people who have unusually high or unusually low compensation measures among others with similar positions in the company. What explains their pay levels and why are they so different from the others who are supposedly similarly-situated? 

The research into their compensation levels often leads to a discussion about the factors that affect compensation, [such as] starting salary, performance evaluations, years of service, special training or skills, and offers from competitors. If it turns out that those paid at the lower levels tend to be a protected group of employees, and if the employer cannot explain why this is the case, then there may be an underlying systemic issue. 

The same can be said if those paid toward the top of the pay range are disproportionately a non-protected group of employees.  Assuming the data fields are available and reliable, this kind of data check can be done by first limiting the data to a group of people reasonably thought to be comparable to each other, and then sorting their data by compensation levels.

It should be noted that the use of the term “outliers” takes on a different meaning in the context of a compensation regression analysis.  Here, the actual compensation levels are compared to the compensation amounts predicted by the regression model, and outliers are those who are paid statistically significantly more or less than the amounts predicted by the model. These outliers should also be examined to determine whether an adverse regression result is generated by inaccurate data, factors not included in the model, or a legitimate underlying compensation issue.

Employers conducting self-audits should do so only under the guidance of legal counsel, and they should not conduct an audit if they’re not willing or able to address and correct any adverse findings that arise from the analysis. 

Clean Up Your Compensation Data

Bloomberg BNA: A common pet peeve among some consultants is cleaning up the contractor’s pay data before running a statistical analysis. You have also stressed the importance of employers vetting their pay data for errors before submitting the information to consultants. It saves time and money, so why is submitting clean pay data a challenge for certain employers?

White: You’re correct that much of our time is spent reviewing employers’ data, identifying issues to research with our clients, and then writing computer code that addresses the issues and creates data files suitable for the statistical analyses.  Once the data files are constructed, the analyses themselves don’t take much time to run. 

My experience is that data fields related to pay amounts are usually very reliable, because if there was an issue with someone’s paycheck amount, employees quickly pick up on the issue and corrections are made. 

Other data fields, however, are not always so complete and accurate. My belief is that employers usually maintain reliable data for areas that are directly related to their core business; for example, prices, sales volume, profitability, production measures, and costs.

However, because the data fields needed for pay analyses often do not fall into this category, there is less emphasis on maintaining accurate and reliable data. There are some employers who maintain immaculate employee data, but there are many who don’t collect and maintain enough employee data that allows for a sufficient pay analysis. 

A good example is education. Many employers have education data as of the time the employee was hired because that was included in the application. However, these employers don’t always know if the employee earned another degree after their initial hire unless the additional degree explicitly impacts the compensation level.

Employers who are inspecting and prepping pay data may want to look into the following areas as they get started: • determine the factors that set the parameters on who is similarly situated to one another; • find out if the electronic pay data includes those factors; • check to see if these data elements are free of missing or unrealistic values, for example, the date of birth in the year 1900; • look for and correct internal inconsistencies across data fields; • determine if there are people who are unique to their positions and are thus not comparable to anyone and should be excluded; and • understand the compensation decision-making process, which is crucial for the analyst to understand in building the model, and which may vary from one job grouping to another.

Take advantage of an authoritative resource to address the ever-evolving issues in labor and employment law with a free trial to the Labor & Employment Law Resource Center.