Opinion: the adjusted gap has become a way of explaining the problem away

I am closer to your position than I used to be. We dropped performance rating from our model in 2023 after finding that women received the top rating at 14% against 19% for men, with no difference in any output measure we could find. The adjusted gap rose from 1.2% to 2.7% overnight. Nothing about pay had changed; we had just stopped laundering one decision through another.

The answer is surely to test each control rather than abandon the method. Before a variable goes in our model we check whether it differs by gender and whether that difference can be justified. Grade stays in, but we run a separate analysis of time to promotion. Tenure stays. Rating is out. That is more work than either extreme, and it is the only version I can defend to both a lawyer and a sceptical employee.

Practical point: whatever you believe, you will be reporting gaps by category of equal-value work under the directive, which is a form of adjustment whether we like the word or not. The debate will shift to whether the categories themselves are fair, which is your argument about grade in new clothes.

From a smaller employer, 600 people in Scotland: we report three numbers side by side and refuse to give any of them alone. Raw median, the gap controlling only for grade, and the fully adjusted figure. Showing the middle one is what makes the point you are making. Ours go 16%, 3.8%, 0.9%, and the drop from the first to the second is the conversation that matters.

Fair challenges, and I accept I overstated it. I do not want the adjusted figure gone. What I want is for every control to arrive with its own equity test attached, as described above, and for the board paper to lead with the raw number. I am going to propose the three-figure format for our next report and see whether it survives the committee.