Mean vs median in pay gap — why the median is the number that matters

The directive requires both. Article 9 lists mean and median gender pay gaps as separate mandatory metrics. The directive is right to require both. But for an HR Director assembling the report, only one of these numbers is operationally useful — and it is not the one most companies instinctively reach for.

The mean is what people are familiar with from school. The median is what statisticians use for compensation. The difference is not academic. It changes the picture the report shows, and it changes the conversation with the works council.

Two definitions, two different sensitivities

Both are valid. The difference is what they are sensitive to. The mean is sensitive to the magnitude of every salary. The median is sensitive only to the position of the middle person. For compensation distributions — which are nearly always right-skewed by a small number of senior earners — these produce different numbers.

A worked example — 20 people, one Director

A 20-person team. 18 employees earn between €40k and €70k. One Senior Manager earns €90k. One Director earns €200k. Half the team is female, half male. The Director is male. The Senior Manager is female. Everyone else is mixed.

Same 20 people. Same data. Two reports. The mean gap says 22.3%; the median gap says 3.5%. A regulator reading only the mean would conclude there is a substantial structural problem. A regulator reading only the median would conclude there isn't.

Neither is wrong. They are answering different questions. The mean is asking: "what does the average woman earn vs the average man, given the current organisational shape?" The median is asking: "what does the typical woman earn vs the typical man?" The first picks up the fact that the Director happens to be male. The second doesn't.

Why the median is the operational number

When a regulator examines a pay gap report, the median is harder to dismiss. The mean can be explained with "we have one senior man earning a lot." The median requires no such explanation — it is the typical experience, by construction.

The median also generalises better. If the company grows by 50 people, the mean shifts with every hire. The median moves only when the middle position changes — far more stable, easier to compare across time.

For the HR Director, the median is the number that surfaces structural patterns. If the median is 0% but the mean is 15%, the structural pattern is "men hold senior positions." If the median is 8% and the mean is 9%, the structural pattern is "men are paid more at every level." These are different problems with different responses.

The quartile distribution as the bridge

Article 9 also requires reporting the distribution of women and men across the four quartile pay bands. This is the bridge between mean and median. The quartile breakdown shows where the men and women actually sit in the pay distribution — which is what the mean is sensitive to and what the median ignores.

Together, the three views compose the picture: mean shows the magnitude-weighted gap, median shows the typical gap, quartile distribution shows the structural composition. Reading them in isolation produces partial conclusions. The report is the combination.

The mean tells you what the organisation pays on average. The median tells you what the average person experiences.

Where the diagnostic starts

The mean and median can be calculated from a clean payroll export. What they reveal depends on whether the job categories are defined first — without that, the per-category metrics required by Article 9 cannot be calculated, and the overall numbers float without context.

The ReadinessCheck™ surfaces whether the category structure is in place, by axis of the directive's principal obligations. It does not produce the report. It tells you whether you can.