Looking at any large company's published pay equity disclosure, one pattern repeats. The reported pay difference — calculated from base pay — sits in single digits. The reported bonus difference, calculated from the same workforce, often sits in the high twenties or thirties. The two numbers describe the same people, paid by the same company, in the same year. They are not the same metric.
Article 9 of Directive 2023/970 requires both. The median pay gap is one metric. The median bonus differential is a separate, parallel metric. The arithmetic that produces them is identical; the inputs are not. The structural difference between base pay and variable pay is what makes the bonus reading consistently wider.
Two compounding mechanics drive it. Neither is hidden. Both can be quantified before a report is filed.
The formula, sitting in plain sight
The single subtlety: the median is calculated across the whole population, not across the population that actually received a bonus. An employee with €0 bonus is a data point at €0, not an excluded row. The "participation" effect — who is in the bonus-eligible pool, and who actually received an award — flows directly into the median.
Mechanic 1 — participation difference
Bonus eligibility is rarely uniform across roles. Sales has a commission plan; engineering does not. Senior management has an annual bonus tied to company performance; entry-level operations does not. The participation rate — the share of the population eligible for a bonus, and the share that actually received one — varies by function.
If men and women are not evenly distributed across bonus-paying functions, the participation rate differs by gender. Even when the bonus amount paid to a recipient is identical, a population-wide median calculated across many zero-receivers produces a wider differential than the base-pay median.
A toy workforce of six. Three men, three women. Base medians: €64,000 men, €52,000 women — a 19% base difference, mostly explained by role mix (sales vs operations). Bonus medians: €12,000 men, €0 women — a 100% bonus differential, driven entirely by which function each gender is concentrated in.
The bonus difference is wider not because women in sales were paid less than men in sales. It is wider because women are not in the function with the bonus plan. The same role-mix produces a moderate base difference and an enormous bonus differential.
Mechanic 2 — magnitude variance
Among employees who do receive a bonus, the amounts vary more than base pay does. A team has base salaries spread across a 1.3× to 1.5× range — the highest base is roughly 30–50% above the lowest. Bonuses for the same team often span a 3× to 5× range — top performer's bonus can be five times the floor.
That variance multiplier matters because medians read magnitude as well as ordering. Wider variance means small allocation differences produce visible median movements. A bonus pool of €1.2M distributed across 40 people with discretion produces more median sensitivity than the same €1.2M added to base.
Two structural reasons drive this. First, base pay is anchored to a band — most people sit inside it. Second, bonus is anchored to a target — but the multiplier above or below target is at the manager's discretion. Discretion produces variance. Variance amplifies any underlying allocation pattern.
Combined effect, plotted
The pattern above is illustrative, but the shape is recognisable. The bonus number can read 5–7× the base number in the same company. Both are accurate. Both come from the same data. They answer different questions.
How the two numbers read together
The base-pay difference is a question about pricing inside the band. The bonus differential is a question about which population gets allocated discretionary cash, and how that allocation distributes.
A clean base-pay reading with a wide bonus differential is usually one of three patterns:
- Role mix. One gender concentrated in functions without bonus plans. Visible at the workforce-composition level.
- Eligibility threshold. Bonus plans extend down to senior roles only; the lower bands are over-represented by one gender, producing a participation gap that reads as a magnitude gap.
- Discretionary variance. Among recipients, the upper-tail awards skew toward one demographic — visible in the magnitude distribution, not the participation rate.
The first two are explainable by structure (the architecture of plans). The third requires looking at the discretionary allocation directly — and is where Article 18 record-keeping becomes relevant. The rationale for the multiplier above target is exactly the kind of decision-level record that the burden-of-proof reversal expects to be contemporaneous.
The "supplementary pay" metric under Article 9(1)(c) covers more than annual bonuses. It includes commission, signing bonuses, retention awards, equity vesting events, and any non-base component awarded in the reporting year. The variance reads even wider on this denominator. Reading bonus and supplementary separately can be more diagnostic than aggregating them.
Where compensation design enters
The two mechanics — participation and magnitude — both flow from design choices that were made before the reporting year began. Three design choices have the largest downstream effect on what the bonus differential reads:
Eligibility breadth. A bonus plan that extends to all employees collapses the participation effect. A bonus plan that extends only to certain functions — or only to certain levels — amplifies it. Eligibility breadth is a binary lever per plan.
Target consistency by band. If the bonus target as a percentage of base is consistent within each band, the bonus inherits the base distribution. If the target varies inside the band (some senior engineers have 10% target, others have 25%), the bonus introduces variance the base does not contain.
Discretionary multiplier governance. A bonus where 0.5–1.5× of target is allowed at manager discretion produces less variance than one where 0.0–3.0× is allowed. The wider the discretionary corridor, the more allocation patterns surface in the median.
The bonus differential is what compensation design looks like under a median lens. The number is not the diagnosis — the design upstream of it is.
Where the diagnostic starts
The most useful first read is to compute both differentials separately — base, then bonus — and to identify which of the three patterns (role mix, eligibility threshold, discretionary variance) the difference between the two numbers maps to.
A clean PayGapCheck™ report produces both metrics, broken down by category and quartile, with the structural decomposition that distinguishes design effects from allocation effects.
Two metrics. Same workforce. Very different stories.
PayGapCheck™ produces both differentials — base and bonus — broken down by category and quartile, with the structural decomposition that separates design effects from allocation effects.
See your numbers →