Your pay gap report stops at your payroll. Article 19 lets a comparison reach another legal entity, a predecessor who had already left, or a person who never existed.

Every pay gap report has an edge. It ends where the payroll file ends: one legal entity, one reference year, the people who were on the books during it. That edge feels like the boundary of the exercise, and for reporting purposes it is.

Article 19 of the Pay Transparency Directive draws a different edge, and it sits further out. It governs how a worker proves that their work is equal, or of equal value, to someone else's. It does not accept the employment contract as the limit of that question, it does not require the two people to have been employed at the same time, and in its third paragraph it does not require a second person at all.

This is not new law in substance. Article 19 codifies rules the Court of Justice has been applying for four decades. What is new is that those rules now sit alongside a reporting duty that will hand workers and their representatives structured pay data every year. The evidence problem that used to protect employers is being solved for the claimant.

1. The comparison follows the pay decision, not the contract

The boundary that matters legally is not the employer. It is whatever sets the pay conditions.

Article 19(1) states that when assessing whether workers are in a comparable situation, the assessment is not limited to situations in which the female and male workers work for the same employer. It extends to a single source establishing the pay conditions.

The Court of Justice built that test in Lawrence (C-320/00) and applied it directly in K and Others v Tesco Stores (C-624/19), decided on 3 June 2021. In Tesco, shop workers, predominantly women, compared themselves with distribution centre workers, predominantly men, in entirely different establishments. The Court held that where the pay conditions can be attributed to a single source, the two groups can be compared under Article 157 TFEU even though they work in different places. Lawrence supplies the mirror image: where no single source can be identified, there is no body capable of restoring equal treatment, and the comparison does not run.

For a mid-sized group this is the practical consequence: a comparison does not stop at the entity that issues the payslip. If salary ranges, the grading structure, the bonus scheme, or the sign-off on salary decisions are set centrally, that central point is the source, and everyone whose pay it governs sits inside one comparison pool.

2. A comparator who has already left is still a comparator

A departure does not close the exposure. Article 19(2) provides that the comparison is not limited to workers employed at the same time as the person bringing the claim.

The origin of that rule is Macarthys Ltd v Smith (129/79), decided on 27 March 1980. Wendy Smith took over a warehouse manager role four months after her male predecessor left. He had been paid 60 pounds a week and she was paid 50. The Court held that the equal pay principle cannot be confined to situations in which men and women do equal work for the same employer at the same time. The four-month gap did not matter.

Two things follow for an employer preparing for the Directive. The first is that leavers belong in the analysis, not only in the archive. A role whose pay dropped when it changed hands is a documented pay difference between a man and a woman doing the same work, and the fact that one of them has gone does not resolve it.

The second is about retention. Article 9(9) already requires that the per-category figure under Article 9(1)(g) be made available to workers and their representatives, and that the four preceding reporting years be provided on request. Historical pay data will surface in any case. It is worth having the reasoning that goes with it, not only the numbers.

3. Where there is no comparator, statistics take their place

This is the paragraph that reaches the roles most employers assume are out of scope.

Article 19(3) provides that where no real comparator can be established, any other evidence may be used to prove alleged pay discrimination, including statistics or a comparison of how a worker would be treated in a comparable situation. The preamble to the Directive is explicit that a hypothetical comparator should be available where no real-life comparator exists, precisely so that workers in strongly gender-segregated occupations are not left without a route.

The Court has been here before. In Enderby v Frenchay Health Authority (C-127/92) a speech therapist, in a profession that was almost entirely female, compared herself with clinical psychologists and hospital pharmacists, professions that were predominantly male, and was paid substantially less. The Court held that where the two jobs are of equal value and the statistics describing the situation are valid, there is a prima facie case of sex discrimination, and it is then for the employer to justify the difference. The fact that the two pay rates came out of separate collective bargaining processes, each of them non-discriminatory internally, did not prevent that finding. Enderby herself had real comparators. What the case establishes is that statistics can carry the argument. Article 19(3) goes one step further and admits the comparison where no real comparator exists at all.

Note where this lands in a pay gap report. A value group with no members of the other sex produces no gap figure. It is also the group that will be suppressed for data protection reasons, because a figure covering three people can identify them. The absence of a number in the report is not the absence of exposure. Under Article 19(3) that group is precisely where the comparison switches from a person to a statistic.

4. Article 18(2) removes the first hurdle

A wide comparator rule only bites if a claimant can get a case started. Article 18 is what makes that easy.

Article 18(1) is the familiar shift: once the worker establishes facts from which discrimination may be presumed, the employer must prove there was none. Article 18(2) goes considerably further. Where an employer has not complied with its transparency obligations under Articles 5, 6, 7, 9, and 10, the burden of proof lies with the employer without the worker having to establish that presumption first. The only carve-out is a breach that is manifestly unintentional and minor.

Read the two provisions together and the picture is uncomfortable. The comparison can reach across entities, back in time, and into a hypothetical. The evidence can be statistical. And if your pay criteria under Article 6 were never written down, or your Article 7 answers went out late, you are the one explaining the difference from the first day of the proceedings.

5. The German position today is narrower, and that is the point

German law currently sets a much tighter comparison than Article 19 does, which is why transposition is likely to be a widening rather than a tidy-up.

Under section 10(1) of the Entgelttransparenzgesetz, a worker exercising the individual right to information has to name a comparable activity themselves. Section 12(1) restricts the right to establishments with more than 200 workers. Section 12(3) sentence 2 then removes the comparison figure altogether where the comparable activity is performed by fewer than six workers of the other sex.

Stack those three and a familiar outcome appears. A woman in a function that is almost entirely female asks what a comparable male role earns, and receives nothing, because there are not six men doing it. That is a dead end created by the design of the German provision, not by the underlying principle. Article 19(3) operates at a different stage. It does not repair the information right, but it means the same worker is not left without evidence in proceedings, because it requires neither six comparators nor one. Whether the information right itself widens depends on how section 12 is transposed.

As at 24 August 2026 Germany has not published a transposition draft. Commentators expect national legislation in 2027 at the earliest. The Directive's transposition deadline of 7 June 2026 has passed.

6. What this means for the next twelve months

The work here is mostly mapping and documentation, and none of it depends on the German draft appearing.

1. Write down where pay is actually decided. For each entity, record who sets the salary ranges, who approves offers above them, and which schemes are group-wide. That document is your single source map.
2. Test the group boundary before someone else does. If one grading structure covers several entities, run the value group analysis across that population as well as per entity. You want to know the difference between the two results.
3. Bring leavers into the analysis. Check roles that changed hands between a man and a woman in the last three years and look at what happened to the pay attached to the role.
4. Identify the segregated functions. List the value groups with no members of the other sex. These produce no reportable figure and carry the statistical comparison risk under Article 19(3).
5. Compare those functions upwards, not sideways. A female-dominated group is compared with a differently composed group of equal value on the Article 4(4) criteria of skills, effort, responsibility, and working conditions. That comparison is the one Enderby describes.
6. Close the Article 18(2) exposure first. Written pay and progression criteria under Article 6, a working answer process under Article 7, and a report you can stand behind under Article 9 are what keep the burden of proof where it normally sits.
7. Record the justification at the time, not afterwards. Whatever objective, gender-neutral reason explains a difference, it is far easier to evidence in the year the decision was made than three years later in a proceeding.

The instinct behind most pay gap preparation is to define the population, calculate the metrics, and file the report. Article 19 says the population is not yours to define. If one grading structure governs four entities, a comparison can cross all four. If a role was better paid under its previous holder, that difference survives their departure. And if a function has no comparator at all, the comparison is made with statistics instead of with a person. The report is the visible part of the obligation. The comparison is the part that decides what the report has to survive.