Arguments both for and against the Securities and Exchange Commission’s pay ratio rule probably won’t go away any time soon. The debates, however, don’t change the fact that, for now, the SEC expects companies to comply.

The good news is that as far back as 2015, the SEC began offering concessions intended to make compliance easier.

Companies need only identify the median employee once every three years, not annually. They can also select any date within the last three months of the most recently completed fiscal year for the calculations.

Registrants can exclude up to 5 percent of non-U.S. employees when determining their median employee. Employees in a foreign jurisdiction with data privacy laws that prohibit pay information from being transmitted can be removed from the metric. Employees acquired through a merger or acquisition can also be excluded from the median employee calculation until the next fiscal year.

Among the most meaningful of concessions was that the SEC will allow statistical sampling for determining the median employee. The Commission also provided for geography-based cost-of-living adjustments—essentially, pay differentials to let companies equalize what the wages of an employee in a low-income country would be in the CEO’s home jurisdiction.

In a recent comment letter to the SEC as part of a reconsideration of the rule by Acting Chairman Michael Piwowar, Willis Towers Watson—a multinational risk adviser and insurance brokerage—offered six additional suggestions for simplifying the calculation of the CEO to worker pay ratio. They include:

Narrowing the definition of “employee” and “employee of the registrant.” Having a broad rule in place that requires companies to make a determination as to what involvement they might have in setting the pay of independent contractors is unwieldy and creates an enormous amount of extra work that provides very little benefit, Willis Towers Watson says. It recommends that the Commission adopt an interpretation that permits companies to determine the workforce based on whether a worker is treated as an employee in his or her home country.

Clarifying that prior-year data can be used to identify the median employee. The most straightforward pay information for companies to gather in performing a 2017 calculation of the “median employee” would be using 2016 calendar year-end data. “The current [compliance and disclosure interpretations] are vague about whether that was permissible” the comment letter says, recommending that guidance be amended to permit use of prior fiscal year data as long as there has not been a change in the registrant’s employee population or employee compensation arrangements that would result in a significant change in its pay distribution to the workforce.

Clarifying that companies can use base pay as a consistently applied compensation measure. The SEC’s recent guidance on this issue, with references to companies that grant equity compensation, “muddied the waters a bit and called into question whether a company could choose base pay as its CACM,” Willis Towers Watson wrote. “While we understand that some companies read the regulations to permit use of base pay, we recommended that the SEC clarify that base pay would be an appropriate CACM for companies that have workforce segments that receive different elements of pay, such as those that have both hourly and salaried employees.”

“The SEC’s rule was rather vague and potentially overbroad in that it asked companies to do a test of every employee to see how it as that person’s pay was determined.”
Steve Seelig, Compensation Counsel, Willis Towers

Making it clear that the disclosure can be a “reasonable estimate.” When determining the median employee, Willis Towers Watson wrote, companies are permitted to take an approach that is highly likely to yield a different pay ratio than they would obtain if they were required to calculate the Summary Compensation Table total compensation number for every employee, and that their approach will be different from that of the next company. “We recommended that the SEC make it clear that companies can disclose that their CEO pay ratio is a reasonable estimate calculated in a manner consistent with the regulations,” the firm wrote.

Allowing registrants to use readily accessible records to determine employee classifications. Many global companies find it difficult, if not impossible, to independently verify the information provided by their HR and payroll representatives in foreign jurisdictions that identifies full-time, part-time, seasonal, or temporary workers. As a result, such companies are forced to rely on the information provided in performing their median pay determinations, the firm’s comment letter explains. Without being able to independently verify the accuracy of this information, their median employee calculation would be considered inaccurate.

The recommendation: that the SEC permit companies to rely on the employee demographic information provided by their various payroll and/or HR information systems to the extent they reasonably believe that the data is accurate.

Permitting broader use of “reasonable estimates” as part of statistical sampling to greatly reduce the effort required for companies with global workforces. “The current regulations don’t provide sufficient information on what’s meant by permitting companies to use ‘reasonable estimate’” in their calculations,” Willis Towers Watson wrote. “The issue we confront with clients using statistical sampling is that their data are difficult to obtain for overseas payrolls.”


As executive compensation counsel for the Center for Research and Innovation of Towers Watson in Arlington, VA., Steve Seelig serves as the firm’s expert on matters involving the taxation, accounting, and legal implications (including SEC disclosure requirements) of all forms of executive compensation and perquisite programs. He is a frequent public speaker on executive compensation matters and a frequent author. Follow Seelig and other Towers Watson consultants at the blog:

The firm recommended that the SEC permit companies with employees in multiple jurisdictions to use a sample of employees from a single jurisdiction to represent the employees from other jurisdictions, provided the company uses a statistically valid method to adjust for any known differences in average compensation between employees in the varying jurisdictions.

This, it says, “would greatly reduce the need to gather data in all jurisdictions and could smooth the process significantly for global companies.”

Additional advice for companies from Steve Seelig, a senior regulatory adviser for executive compensation at Willis Towers Watson, is to have the right pay ratio team in place. The talent mix may need to go beyond lawyers.

“You need to have the right talent in place to read the regulations,” he says. “They were written by lawyers and economists on the SEC side and that’s who you need to have analyze them.”

Questions of employee definition are causing “a huge amount of consternation for companies,” Seelig says. “The SEC’s rule was rather vague and potentially overbroad in that it asked companies to do a test of every employee to see how it as that person’s pay was determined.”

Many of the recommendations filed as part of the recent rule reconsideration argued that this definition should be limited to how those workers are treated for in-country tax or legal purposes.

“This notion that companies would be treating people, who otherwise would be employees, as independent contractors solely for purposes of doing their pay ratio calculation, is really something that is misguided,” Seelig says, agreeing that the definition should be narrowed.

“For some of our clients, we are going through tedious work to try to figure out how exactly these folks are paid, and who is determining the pay,” he adds. “It is a process unto itself and takes an inordinate amount of time.”

The result: Many companies are reaching the conclusion that instead of gathering separate compensation data for everyone, they are going to use base pay.

“The SEC’s regulations, however, are not clear on that issue and we suggested to them, as did many other commentators, that they need to specify that you can use base pay,” Seelig says. “Hopefully, they will take everybody up on that. It would be fairly easy to implement through just the additional questions and answers of a CD&I.”

“There is no reason for them not to,” he adds. “We’ve done calculations using base pay for some of our clients. We’ve done calculations using total compensation for some of our clients who can access it. The answer ends up being the same. You find roughly the same median employee anyway. It doesn’t yield a different answer,”

Seelig also weighed in on the problematic nature of pay ratio disclosures being “filed” versus “furnished.”

“The inherent problem here is that the rules themselves are so flexible, by permitting estimates and statistical sampling, that whatever number you come up with is going to be done in a different methodology than the next person,” he says. “That basically reflects that nobody has got the exact, precise, and correct summary compensation table median employee determined. So, no one is really following the statute based on what the SEC regulations say they can do.”

To call that data a filed and precise number is problematic. In response, most comment letters called for making it a furnished disclosure, not a filed one.

“What we said, is that if you feel it is necessary that this is a filed disclosure, then provide guidance that allows companies to disclose that their pay ratio is a reasonable estimate,” Seelig says. “Even though it might be a filed number, by showing it is a reasonable estimate and declaring that in your disclosure you have given yourself some wiggle room.”

What makes that essential to do, Seelig says, is that if you do statistical sampling you basically get to disclose it as an estimate. “You would disclose an error rate, ‘Here is my median employee’s pay give or take $1,000.’ By definition, if you use statistical sampling, you are disclosing a range. So why shouldn’t anybody else do that who isn’t using statistical sampling?”