Methodology Study

Comparison of the Three Most Prevalent
Pay Equity Analysis and Remediation Methods

Study Overview
Zev J. Eigen, Allison Adams & Andrea Palmiter

fairpay-page-a-chart

Not All Pay Equity Assessments are Created Equal

Employers are wise to review pay practices to ensure compliance with state and federal laws. The laws have nothing to do with intent. They are all about the math and methods. That is, even employers with the best intentions may run afoul of the law if they are paying people performing comparable work inequitably net of relevant job-related factors like years of experience or other measures of output or effort exerted. And, some employers are more interested in reaping the benefits of declaring a pay equity victory, than they are in truly putting in the work to achieve it, and most importantly sustain it over time.

The challenge for well-intentioned employers seeking to get ahead of risk, and to fairly and equitably compensate employees is that most employers are not data scientists / labor economists / comp and benefit experts / employment lawyers. So, they are left to rely on other experts to help them navigate these waters. From what we have observed, there are several different approaches employed by experts in this space. The Fair Pay Workplace whitepaper shines a much-needed spotlight on the choices and challenges faced by organizations that want to shore up their commitment to this important social and legal responsibility. Instead of speaking in hypotheticals, we use real-world data, anonymized with permission, to explore the similarities and differences across the three most common pay equity approaches in the market. We think this will help employers assess which approach is best for them.

Key Challenges to Pay Equity Analysis

Buy-in is no longer the main roadblock in the pay equity space—parity is top of mind among organizations that wish to hire and retain top talent today. Spurred by the desire to do the right thing for their employees, organizations are equally concerned about achieving equity the right way, but may face pressures to present a certain outcome that could involve a suboptimal approach that may even exacerbate risk and disparities.

Unequal Measurement and Methodology

There are several ways to identify and remedy pay equity concerns, and the plain fact is that some approaches are flawed. Some even make things worse for employees and employers. That can erode brand confidence, negatively impact culture and decrease employees' earning power over the course of their careers.

Staking a claim to workplace fairness that is built on faulty methodology and assumptions is a net-negative for both businesses and employees, eroding brand confidence, negatively impacting culture, and decreasing an employee’s earning power over the course of their career.

Ineffective Organizational Grouping

How organizations group their employees is critical to successful analysis. Common mistakes and resulting issues include:

The faulty scale: By dividing an organization into an excessive number of groups, it becomes impossible to make meaningful comparisons for the majority of employees. Say, for example, all marketing VPs are subdivided by skill set and title. They would then be considered too specialized to compare and would be excluded from analyses.

The mountain: Conversely, a similar mistake arises when looking at the entire organization without breaking it up into meaningful, intentional groupings to measure equity. By looking at the organization as one large group, there is no way to examine parity between parties.

Instead, we propose an organizational approach centered around the Substantially Similar Groups (SSGs) methodology, which shifts away from the division of groups by job titles, which are too granular or too broad, and often arbitrary. SSGs are comprised of employees who perform substantially similar work based on skill, effort, accountability, and circumstance. By adopting a SSG methodology paired with an objective approach, organizations are primed to examine and discover disparities and avoid compounding pay gaps over time.

Analysis of Common Pay Equity Methods

The dataset for this representative study consists of 496 employees across six substantially similar groups (“SSGs”). We applied the three most common methods of identifying whether there are pay equity concerns from a statistical vantage to these data, and then applied different versions of the remediation recommendations associated with each. The main goals are to compare: (1) which groups are flagged for review?; (2) to whom is a remedy proposed? And, (3) after monetary adjustments are in place, does the method succeed in reducing the statistical evidence of gaps occurring “because of” individuals’ protected class status?

Method 1: One Overall Fixed Effect Model

This approach identifies and adjusts outliers using a model that spans the entire organization, and proposes remedies for flagged disparities by either a) adjusting negative outliers regardless of gender, or b) only adjusting negative outliers identified in the disadvantaged protected category.

The main positive of this approach is the convenience and ease of running one regression model. However, that positive is balanced by three core concerns:

Remediation costs: as this model is almost certain to both under- and over-correct, the cost savings of running a single model is eclipsed by the cost of unnecessary remediation.

False positives: accuracy is compromised by both flagging for remediation groups that aren’t statistically related to gender differences, and the inability to be flexible and tailor controls to groups.

Questionable legality: in a court of law, the appropriate regression model used by the agency, plaintiff, or defendant would likely be to regress compensation on gender not for the entire employee population, but for the specific group or groups at issue in the suit, which is not in line with the method applied here.

Method 2: Group by Group Model—Standard Deviation/Outlier Approach

Method 2 applies multiple regression models for each SSG large enough to be statistically viable, remediating disparities by either a) adjusting flagged negative outliers regardless of gender (like Method 1, but with models siloed by SSG, or b) adjusting the negative outliers identified in the disadvantaged protected category within the SSG.

The advantages of Method 2 lie in avoiding the issues inherent in trying to fit a single model to a complex organization, also avoiding the problem of applying controls to groups to which they should not be applied. That positive is counteracted by a higher average per person cost for remediation, though allocated to a smaller number of people. As with Method 1, there is also a risk that this approach will only improve the pay for a relatively small subset of statistically disadvantaged individuals.

Though there is a well intentioned focus on fairness with Methods 1(a) and 2(a) in which both men and women are eligible for remediation adjustments if their compensation is less than expected by the applicable models, correcting for all individuals who deviate significantly from the mean organizational pay, the argument could be made that this approach to remediation is explicitly unfair. Treating men and women (or people of different races or ethnicities) equally only makes sense if there were a level playing field to start, and does little or nothing to address problems if women or people of color are underpaid because of their gender, race or ethnic background.

Method 3: Group by Group—Statistically Significant Gaps (p < .05)

Method 3 is explicit in examining data based around SSGs, using parametric tests to evaluate whether men and women are paid differently because of their gender. Disregarding the method of identifying outliers, Method 3 instead focuses on looking at predicted values derived from regression modeling as a guide for remediation. This approach takes into account not just whether individuals are being paid less because of gender, but the degree to which they are. It provides a more targeted method that will ultimately be less costly to companies while maximizing employees’ earning potential. With Method 3, there is no option to apply economic remediation to both genders where there is statistical evidence that one gender is paid less because of that protected class status. Remedies are suggested in that instance to first level the playing field.

Methodologies Matter.
It’s Personal.

Not all methods are created equal—the future of fair pay depends on the fairness of the approach. Most importantly, after each method’s remediation recommendations are implemented, only Method 3 succeeds in reducing the distributional gap to levels below the standard accepted legal criterion.

FairPay_Infographic_FINAL

The resulting remediation varies widely depending on the methodology applied. For example, looking at one woman from our data set who is just starting her career, a pay equity adjustment can pay dividends in the years to come. Assuming a 3% cost of living increase every year for the remainder of a 45-year career and no other pay changes, a woman in the fourth year of her career would see substantial impact over time regardless of approach. However, changes in lifetime earnings range widely from method to method—an increase of $383,319 with Method 1(b) to nearly $2MM more (a 31% increase) with Method 3.

As the methods studied also differed widely around how many individuals are actually recommended a remedy, we found examples of other women who are flagged for remediation by method 3 but left out with the alternate approaches. In this scenario, the organization declares achievement of pay equity having met their own criteria for that outcome and it’s uncertain whether the individual will ever receive her disparity corrected. Organizations interested in reducing legal risk should take note that statutory damage periods can be four to six years depending on the jurisdiction, so even employers with no altruistic motivation should regard these comparisons as presenting a sobering choice of methods when thinking of potential class-wide pay equity damages calculations.

Conclusion

Method 3 is the only approach that yields true, meaningful and lasting pay equity. For this reason, the Fair Pay Workplace relies upon Methodology 3 as the required approach for pay equity assessments as described in the certification program’s Standards & Rules. The pay gap doesn’t have to last for the next 200 years—employees and organizations both win if we can guide organizations to adopt a wise and appropriate approach such as Method 3, proven to address disparities with measured and uniform fairness.