Skip to main content
Pay equity analysis

Unadjusted vs adjusted pay gap

What the two gender pay gap figures mean, how Evenpay calculates them and how to judge the model's quality.

Written By Karl

Last updated About 9 hours ago

Evenpay shows two gender pay gap figures side by side. The unadjusted gap compares the pay of women and men across the whole organisation. The adjusted gap compares women and men after accounting for the pay factors you have chosen, such as job grade, experience or location. Reporting needs the first; explaining pay differences needs the second.

Where: Pay Equity → Overview → Pay Gap Analysis

The unadjusted gap

The unadjusted gap is calculated in one of two ways. You choose which one is shown in Settings → Pay equity & transparency → Pay gap calculation, and both are calculated and stored either way.

  • Mean-based: the difference between the average (or median) pay of the two groups.

  • Regression-based: a gender-only regression on the logarithm of salary.

The figure is labelled with its direction, for example (men earn more), (women earn more) or Gender neutral.

The adjusted gap

The adjusted gap uses a Blinder-Oaxaca decomposition on the logarithm of salary, with your active pay factors as the explaining variables. Each factor's share of the explained difference is calculated with Shapley values, and the results are converted back to percentages. The note How is this calculated? on the card describes the method.

Anything you don't include as a pay factor becomes part of the unexplained gap. See Choosing pay factors.

The adjusted panel has two modes:

  • Statistical: the model result.

  • Explanation Status: the gap split into acknowledged and unacknowledged parts, based on your pay rationale tags.

Minimum data

An analysis needs at least 20 employees, and the adjusted model also needs at least 20. If the model can't run, the card shows Adjusted gap unavailable.

Checking the model's quality

Select Show model transparency to see:

  • R² and Adjusted R²: how much of the pay variation the factors explain. 70% or more is high, 50% or more is acceptable.

  • The p-value and the number of observations by gender.

  • Obs / PIF Ratio: observations per pay-influencing factor. Too many factors for too few people makes the model unstable.

  • Max VIF: how strongly factors overlap with each other.

  • Factor Statistics: each factor's coefficient, p-value, impact, contribution and data quality.

Good to know

  • The reference group, sign and headline statistic are set in Pay gap calculation settings. Changing them changes how figures are displayed, not the underlying run.

  • Edit factors on the card shows how many factors are used, for example 6/16.

  • The Compare to toggle (Median/Mean) changes only your own view.

Was this helpful?

Still need help? Share an idea