Pay Equity: Grouping Employees, Coverage, and Best Practices

Resolution Economics recently released a Webinar titled: Analyzing Compensation by Job. The following is a summary of the topics covered in this webinar, as well as a number of supplementing points on these topics. Berkshire is a Division of Resolution Economics.

Pay Equity is an important and trending topic in the nation right now. Employees are keen on determining whether they are being paid fairly for their work, and employers are attempting to ensure that they are compliant and up to date with the best practices in compensation equity. This is a difficult question to answer because there are many factors that can be relevant to pay depending on the job and industry—things like an individual’s experience, education, certifications, and performance as well as a job’s market rate, to name just a few. To complicate the matter, the best way to group employees and measure pay equity can differ depending on which government entity you are interacting with. The California Equal Pay Act requires equal pay for “substantially similar work”, while Mississippi requires equal pay for “equal work on a job”, and then the OFCCP has a different explanation for the best way to group employees. For an organization that just wants to do right by its employees, these distinct requirements can provide a major headache, and it is true that the process is made more complex by the fact that there isn’t a prescribed formula for how to group employees, account for the relevant variables, and measure pay equity. However, there are guidelines and steps that a responsible employer can and should take to accurately measure equity within their organization.

The first step in a Proactive Pay Equity study is determining how to group employees—often referred to as Pay Analysis Groupings (PAGs). The ultimate goal of PAGs is to put groups of employees together who are as similar to one another as possible. Again, the exact specifications for how to build PAGs can differ depending on the states in which an organization operates and whether or not it is subject to OFCCP directives; however, some of the same principles will always remain intact. Whether the employees that can be compared must be considered ‘substantially similar’, ‘situationally similar’, or ‘substantially equal’, it is certain that organizations should seek to group employees who have similar roles and responsibilities to make apples-to-apples comparisons. A common initial approach to sufficing this goal is to group employees by job title. As many employers set pay ranges based on the market rate for each job, it stands to reason that these employers should build PAGs by title because differences in pay at the organization are primarily determined by title. As we will see, however, this approach won’t work for every employer.

The primary reason that this approach can fail for many organizations is a concept called Coverage. Coverage refers to the percentage of employees at an organization being analyzed, or covered, in a statistical analysis. The obvious goal for employers should be to include as many employees as possible in their pay equity analyses, however, this isn’t always feasible if grouping by job title. The primary reason being the statistical technique that is often used to measure pay equity: Multiple Linear Regression. Multiple Linear Regression is a statistical technique that allows us to predict a specific outcome variable (in this case, compensation) while controlling for multiple relevant factors (e.g., experience, education, certifications, etc.). Regression is considered ‘King’ in the world of pay equity analysis because it allows an employer to control for each and every job-related factor that’s relevant to pay before assessing whether there are differences between demographic groups. That way, an organization isn’t finding disparity where it doesn’t exist. So why does Regression sometimes lead to difficulty in attaining sufficient Coverage in a pay equity analysis? Because Regression requires at least 30 employees in each PAG, and at least 5 employees in each demographic group that is being compared to one another. So, if an employer is attempting to group employees by title and examine the differences between men & women, Regression will not be the recommended technique for any title that has fewer than 30 employees or fewer than 5 males/females.

So, what is an acceptable level of Coverage in a pay equity analysis? Unfortunately, there is not a hard and fast rule here. The OFCCP has a stated goal of covering at least 80% of employees in their audits; the DE&I group at your organization will likely have a goal of including 100% of employees in a pay equity study. This points to the fact that it depends on the specific goals and requirements of an individual organization; however, “as many as possible” can certainly be considered a guideline. If an employer is able to study 90% of employees when creating PAGs based on Job Title, then this is often a good starting point. Then, the remaining 10% can be grouped in a different manner or studied via techniques built for smaller groupings, such as Mann Whitney Rank-Sum or Cohort Analysis. However, if an employer is only able to review, say, 30% of employees in their pay equity study when grouping by Job Title, then a new approach is almost certainly recommended.

The best alternatives to grouping by Job Title are dependent upon the organization and its structure. Again, the ultimate goal here is to group together situationally similar employees while covering as many employees as possible. Sometimes, an organization can use a ‘Hybrid’ approach where they group by Job Title for Titles that have over 30 employees and then attempt to put similar titles together to build the remaining PAGs. Alternatively, grouping by Job Family, Function, Division, and Department are just a few of the other groups that organizations use to build PAGs when unable to achieve sufficient coverage via Job Title. Oftentimes, an organization might group by something like Job Family and then control for the specific pay differences between different jobs in that Family by including Job Title as a predictor variable. Including Title as a predictor can be essential for organizations that pay primarily according to market rates; however, caution must also be utilized here. If there are 100 people within a Job Family PAG and 70 unique Job Titles within that Family, you will often run into a ‘onesie-twosie’ issue. This refers to a situation where Job Title is a nearly perfect predictor of compensation because there aren’t typically more than 1 or 2 people within each title—Bob is predicted to be paid exactly what Bob gets paid because he’s in a title that only Bob is in. If this is the case, job title should not be used as a predictor to adhere to best practices.

Of course, there are other concerns when conducting a pay equity study as well. In order to determine how to group jobs together, it is imperative that every job’s functions are understood and well-defined. Additionally, it is important to be able to build a Regression model that predicts compensation well—the variables you’re accounting for should have high validity in predicting employee compensation (r-Squared). Data quality is also a significant hurdle for many organizations. For example, an employer might pay their employees more based on their education levels, but not track each employee’s education in their HRIS. This makes measuring how much education matters in predicting compensation very difficult! All of these complexities and dependencies are why it is always recommended that organizations looking to have their analyses completed according to best practices work with a pay equity expert. An experienced pay equity expert will have seen many of these issues and questions play out and will be able to cater a pay equity solution to each client’s specific needs. Peace of mind regarding compensation practices can be achieved—it just requires experience, effort, and expertise.