You’re Missing Compensation Data for Your Pay Ratio Calculation. How Do You Find the Median Employee?

As 2017 approaches, bringing with it the reality of pay ratio disclosure, companies are taking the time to draft their calculations and statements. But a problem many of them have run into is that they simply don’t have data for every business division in every country. This is especially true of multinational conglomerates that are not highly centralized.

Unfortunately, ignoring the problem won’t work. The final pay ratio rule says that companies have to consider every sizable employee population in their pay ratio. That means a country with a production plant employing 10% of your workforce, for example, has to be part of the equation.

But all is not lost! Here are a few ways to work around, rather than through, this difficult and potentially costly issue.

Shortcut the population.

To use this expedient, we need to be able to say with confidence that all of the employees missing data would fall in one half of the distribution, likely below the median. For example, assume you have a global population with a median pay above $40,000, but there are employees from a jurisdiction for which you don’t have salary data. If the highest-paid employee in this jurisdiction makes $20,000, the median employee won’t change no matter what they are making. Those missing salary data could make $20,000 or $0—for a median, unlike a mean, only the actual middle observation matters. In this case, you can just account for all of the employees from this jurisdiction as below median, perhaps by assigning a $1 value, and move on.

Focus on the key compensation range.

This approach involves using a preliminary analysis to narrow down a key range—perhaps $5,000 wide—where you believe the median will fall. If you can’t get full data from a jurisdiction, get a basic count of employees who are above, below, or within the target range instead. You can then increase accuracy by gathering detailed data on employees in this range only.

Sample selectively.

For key employee groups who may be in this near-median range, use statistical sampling to understand this population and its parameters. While sampling, especially simple random sampling, may have drawbacks as an overall approach, the pros of a more thoughtful sampling technique may well outweigh the cons.

Develop a hypothetical distribution.

Perhaps the most unique and useful methodology, this involves gathering some summary data—for example, the total pay, number of employees, and types of roles in a given jurisdiction—and then sampling from a group of “simulated” employees. You can then develop summary data from:

1. Comparison to other known groups
2. Financial reports available to the company
3. Statistical sampling within the group
4. Limited available data gathered from unit heads

Analyze your approach

When using any of these methods, be sure to ask the following:

1. How much known data and how much assumption-driven data are you using in your calculation?
2. What level of variation are you exposed to, based on differences that could arise in this calculation approach? Would you expect changes to arise due only to noise in the methodology itself?
3. How do you maintain a consistent definition of pay over any group where a simplification is used?
4. Is this methodology repeatable, and are the results consistent in future periods?

Answering these questions will help you tighten up your process ahead of crunch time in late 2017 and early 2018. Details matter greatly here, and every company’s facts and circumstances are unique. Got questions about your particular situation? Let us know.