Want more expert insights on the topic of pay equity? Access our entire Pay Equity Deep Dive Series

One topic we hear too little about is evaluating the effectiveness of pay equity remediation efforts. For example, if an organization spends $X on remedial pay adjustments, will pay disparities fall by $X, by more than $X, or by less than $X?

To find out the answer to this question, let’s first review the typical pay equity remediation process.

Typical Pay Equity Remediation Process

Most organizations focus their remediation efforts on Pay Analysis Groups (PAGs) in which there are statistically significant pay disparities based on gender, race/ethnicity, or another demographic characteristic. Statistical significance is typically measured using a 5% significance level (i.e., p-value ≤ 0.05) and indicates that there is a 5% chance or less that the result is due to chance.

Once these areas of concern have been identified, the next step is to select a remediation strategy. We examined the remediation strategy selection process in an earlier blog. As we noted, the most common remediation strategy is what we call the “Class + Individual Effect Focused” strategy. Under this strategy, employees who are part of an impacted class (i.e., a class with a pay disparity) and are paid less than the analysis predicts they should be, are eligible for a pay adjustment.

Achieve Authentic Pay Equity With Software 

Further, in most instances organizations do not have sufficient budget to cover the full remediation cost associated with their selected remediation strategy. As detailed in a previous blog, there are numerous considerations when deciding on a remediation budget. A common option is to set the budget to the amount needed to reduce the size of the class disparities to the point where they are no longer statistically significant at the 10% level. Class effect pay disparities would still exist but would not be considered statistically significant.

We call this the “10% Minimum Remediation” budget.

Calculating the ROI of Remediation Efforts

Assuming we’re using the typical remediation process outlined above, the table below shows an illustrative example of how to calculate the ROI of remediation efforts. We discuss each column in turn below the table.

Calculating the ROI of Remediation Efforts

(1) Pay Analysis Group

The first column is a list of the ten PAGs with statistically significant pay disparities.

(2) Statistically Significant Annualized Pay Disparity

The second column is the annualized dollar value of the statistically significant pay disparities within each PAG. It is the dollar value of the class effect we discussed in our earlier blog. This figure reflects:

  1. Magnitude of statistically significant class disparities (e.g., Hispanic women earn 94 cents-on-the-dollar compared to White men; Black men earn 97 cents-on-the-dollar compared to White men).
  2. Number of people in impacted classes (e.g., number of Hispanic women; number of Black men).
  3. Average compensation of those in impacted classes (e.g., average base pay of Hispanic women; average base pay of Black men).

(3) “10% Minimum Remediation” Budget

The third column is the “10% Minimum Remediation” budget. As noted earlier, these values are estimates of the budgets needed to reduce the size of the class disparities in each PAG to the point where they are no longer statistically significant at the 10% level. Across the ten PAGs, the dollar value of the class disparities totals $2.6 million, and the estimated budget to reduce the class disparities in each PAG to the point where they are no longer statistically significant at the 10% level totals $840,000.

As discussed earlier, if we assume the most common remediation strategy, the $840,000 remediation budget would be allocated to employees who are part of an impacted class and whose actual pay is less than their predicted pay (i.e., “Class + Individual Effect Focused” strategy). There are a variety of ways this budget can be allocated to these impacted employees.

For our example here, we assume that the budget is allocated to those who are the most below their predicted pay. Specifically, we start with the individual who is the furthest away from their predicted pay, based on standard deviations. This individual receives an adjustment to bring their pay up to the standard deviation of the next lowest paid individual. These two individuals then receive adjustments to bring their pay up to the standard deviation of the next lowest paid individual. This process continues until the budget is used up.

(4) Post-Adjustment Disparity

In my experience, it’s commonly assumed that spending $840,000 on remediation as outlined in the table above will reduce class disparities in these PAGs such that they are no longer statistically significant at the 10% level. Given that the budget was selected with this objective in mind, it seems like a reasonable assumption. However, this may not be the case.

The only way to know how remediation actions impact pay disparities is to re-run the analysis with the adjusted pay amounts. This exercise uses the same data we started with, except for incorporating updated pay information for those employees who received a remedial pay adjustment. When we re-run this analysis, we are re-estimating the regression weights associated with the Wage Influencing Factors (WIFs) included in the model and re-estimating the class effects (i.e., pay disparities) associated with the demographic characteristics.

The fourth column in the table above represents the post-adjustment disparity in each PAG, after implementing the remediation actions. As an example, in the first PAG, the initial pay disparity is $635,000. After spending $138,000 on remediation and re-running the analysis to take these higher wages into account, the resulting post-adjustment disparity is $348,000. This post-adjustment disparity is not statistically significant at the 10% level, which was the objective of the pay adjustments.

We’ve also noted which disparities are statistically significant at the 10% level (PAGs 3 & 9) and which are statistically significant at the 5% level (PAGs 4 & 10), even after spending the corresponding estimated remediation budgets. This example illustrates that it’s possible for statistically meaningful disparities to remain even after making remedial pay adjustments.

(5) Disparity Reduction from Remediation

The fifth column is the reduction in pay disparities after remediation. This figure is calculated by subtracting the post-adjustment disparity in the fourth column from the original pay disparity in the second column. Looking across all ten PAGs, the total disparity reduction is about $1.1 million. For the first eight PAGs, disparity reductions exceed the budgets spent adjusting pay. The last two PAGs illustrate that while disparity reductions are usually larger than remediation budgets, this need not be the case.

(6) Disparity Reduction Per Dollar Spent on Remediation (“ROI”)

Finally, we’re ready to calculate the ROI of our remediation efforts! ROI is measured by comparing the disparity reduction in the fifth column to the remediation budget in the third column. Essentially, it measures the reduction in disparity for every dollar spent on remediation. You’ll note that the table is sorted in descending order of ROI. In eight of the ten PAGs, each dollar spent on remediation reduces the pay disparity by at least a dollar. Looking across PAGs, each dollar spent on remediation yields a $1.31 disparity reduction. Note that in some PAGs, the ROI exceeds a dollar, yet the post-adjustment disparity remains statistically significant (e.g., PAG 4). In the PAGs in which statistically significant disparities remain, additional remediation may be warranted.

For the last two PAGs, the ROI is less than $1, reflecting that the remediation costs exceed the disparity reductions they produce. In these circumstances, it may be worthwhile to consider the Class Effect strategy, since it ensures that pay adjustments are made to all members of an impacted class, rather than just those who are paid less than their prediction.

Trusaic’s R.O.S.A. Approach

As we saw in the illustrative example above, there is no guarantee that remediation activities will remove the statistical significance of pay disparities. In fact, in working with clients, we noticed that statistically significant pay disparities sometimes remained, despite remediation efforts.

Our data science team, headed by Mark Dwyer, decided to tackle this issue. The goal was to generate an optimized set of pay adjustments that removes the statistical significance of pay disparities, after applying the adjustments. We call the approach Remediation Optimization Spend Analysis (R.O.S.A.).

R.O.S.A. performs remediation simulation runs iteratively to identify the pay adjustments that will address a pay disparity. The simulation runs are applied to each PAG, until statistically significant disparities are removed from each one. To the extent that budgets are limited, R.O.S.A. prioritizes remediation in PAGs with the greatest ROI.

Although most often used in tandem with a “Class + Individual Effect” remediation strategy, R.O.S.A. can be applied to a Class Effect remediation strategy. Moreover, R.O.S.A. can target a statistical significance threshold (e.g., below 10% statistical significance) or the size of the disparity (e.g., all disparities should be within 98 cents-on-the-dollar).

By design, R.O.S.A. ensures that remediation budgets are distributed to groups and individuals where they will have the greatest impact toward achieving pay equity.