July 28, 2026
Category:
Pay Equity LegislationAuthor:
Adam Seoudi
/
Head of CX
The Dutch pay transparency consultation has moved well beyond broad legal principles. The government has now published draft definitions, calculation rules, a web form, a reporting template and a data specification showing how employers may have to turn payroll data into a formal gender pay gap report.
The most important takeaway is simple: this is not a seven-number report.
The seven statutory reporting headings are translated into 26 public percentage fields for an employer's own employees. If supplied workers are in scope, the same 26 fields are completed again for that separate population. Employers must also calculate four additional measures for every category of workers, with the draft template allowing up to 200 categories in each of the two category sections.
The methodology is detailed, payroll-driven and, in several places, counterintuitive. It also raises a few questions that should be clarified before the rules and technical specification are finalised.
At the time of writing, on 27 July 2026, the Dutch government is consulting on two connected packages:
The underlying Implementation Bill 36 949 is still before the House of Representatives, and both lower-level packages are drafts. The parliamentary tracker records a target commencement date of 1 January 2027, but the commencement provisions in the draft decree and regulation still contain placeholders. The reporting portal is not yet live, and the methodology or template may change after consultation. The official consultation page says that additional employer guidance is planned for autumn 2026.
The draft template itself is dated June 2026 and marked version 2026 v0.1.
The template divides the report into public elements, corresponding to reporting items (a) to (f), and a non-public category-level element, corresponding to item (g).
For each relevant workforce population, the public part contains the following 26 percentage fields:
The annual and hourly quartiles are separate calculations. Employees are first ranked by gross annual pay and divided into four parts, and then ranked again by gross hourly pay and divided into four parts.
Every result is submitted as a percentage rounded to one decimal place. Although employers need the underlying euro values to perform the calculations, the draft reporting template does not ask them to submit the male and female averages or medians in euros.
The sign convention is consistent across the draft:
Pay gap = (male value - female value) / male value x 100%
A positive result therefore means that the male value is higher. A negative result means that the female value is higher.
If an employer has both its own employees and supplied workers, the template repeats the entire public section. This means up to 52 public percentage fields: 26 for own employees and 26 for supplied workers.
The category-level part of the report is particularly important. For each category of workers, the employer must provide four measures:
There is no separate category-level field for the gender pay gap in reportable gross pay. In other words, at category level the Dutch draft requires employers to split the analysis between basic pay and additional or variable pay instead of reporting one overall gross-pay gap. Employers should not substitute a total-cash gap for these four prescribed measures.
The template allows a maximum of 200 categories, with a description of up to 100 characters for each category. It says that the same category structure should be used for own employees and supplied workers.
This creates a potentially large reporting dataset:
These category results are not intended to be published by the monitoring body, but they will be submitted to it and may be used in its analysis and reporting to the European Commission. The underlying bill also requires the employer to provide item (g) to its employees. "Non-public" therefore does not mean that the category results remain unavailable to employees. By contrast, the 26 organisation-level fields are intended to be public.
For an employer's own employees, the methodology starts with loon LB/PH, defined as the income on which wage tax and national insurance contributions are calculated for the payroll filing period.
However, reportable gross pay is not simply the amount taken from that payroll field. The draft decree applies the following adjustments:
Reportable gross pay = loon LB/PH - holiday allowance paid from a previously accrued entitlement - amounts paid from an accrued employment conditions amount + holiday allowance accrued during the period + employment conditions amount accrued during the period.
The purpose is to separate pay from the timing of a payout. Holiday allowance is therefore not excluded from pay. Instead, the payout from a previously accrued entitlement is replaced by the amount accrued during the reporting period.
The decree defines an employment conditions amount (arbeidsvoorwaardenbedrag) as a future pay component expressed in money and accrued under an individual or collective employment agreement, excluding a separately accrued holiday allowance, insofar as it can result in wage within Article 16 of the Dutch Social Insurance Financing Act. The explanatory note gives examples such as accrued year-end payments, individual choice budgets, personal choice budgets and arrangements labelled "employee benefits". The amount is recognised as it accrues rather than when the employee later takes it as money, time or another benefit.
As a mathematical consequence of the current formulas, the current-period accruals for holiday allowance and the employment conditions amount enter reportable gross pay but are not added to additional or variable components. Because basic pay is the residual after subtracting additional or variable components from gross pay, those accruals flow into basic pay.
This is a significant data requirement. A single year-to-date taxable-pay field will not be enough. Employers will need reliable payroll fields for both:
For own employees, the draft defines additional or variable components as:
Basic pay is then calculated as:
Basic pay = reportable gross pay - additional or variable components.
This means that the classification is strongly influenced by the Dutch payroll-tax treatment of a component.
The decree and its explanatory note produce the following treatment:
The explanatory note lists a thirteenth-month payment as an example of a payment that may be taxed under the special remuneration table. It also lists accrued year-end payments as an example of an employment conditions amount. The formulas therefore make the payroll route important: a special-table payment is normally variable, but a payment made from an accrued employment conditions amount is removed and replaced with the current-period accrual.
The examples of excluded tax-free or zero-valued items include Christmas gifts, off-site company fitness, a company bicycle, staff events, qualifying homeworking and travel reimbursements, qualifying study costs, necessary equipment, the cost of a certificate of conduct, workplace refreshments, occupational health and safety facilities, and qualifying work clothing. Their exclusion applies to this standardised report and the corresponding pay definition used for the right to information. It does not mean that such a component can never be relevant to an individual equal-pay claim.
The resulting reporting definition is narrower than every benefit that might be relevant to a substantive equal-pay claim. The explanatory note expressly distinguishes between what is included in this standardised report and the broader concept of pay that may still matter when an employee alleges unequal pay.
The draft methodology requires both annual and hourly results, but the annual measures are not full-time-equivalent values.
Annual gross pay is the sum of reportable gross pay for an employment relationship during the reporting year. Annual additional or variable pay is calculated in the same way. The methodology says explicitly that the employer must not apply a part-time factor.
The explanatory note goes further: employers are not permitted to correct the figures for:
The timing of the payroll entry remains decisive.
The annual group means are also employment-relationship averages. The employer first calculates annual pay for each employment relationship, sums those values for men or women, and divides by the number of male or female employment relationships. Each relationship therefore receives equal weight in the annual mean; there is no full-time-equivalent adjustment.
The draft uses three distinct hour concepts:
The documents do not define verloonde uren as hours physically worked. The precise source is the number of hours reported in the payroll tax return. The safest English description is therefore payroll-reported paid hours.
The employer calculates an hourly value for each employment relationship:
Gross hourly pay = gross annual pay / annual paid hours
Hourly value of additional or variable components = annual value of those components / annual paid hours
It then calculates the male and female mean by adding the individual hourly values and dividing by the number of male or female employment relationships. This is an unweighted mean of relationship-level hourly rates. The draft does not instruct employers to divide total payroll by total hours. A relationship with 100 reported paid hours has the same weight in the group mean as a relationship with 2,000 reported paid hours.
The same sequence applies to the hourly value of additional or variable components. For supplied workers, only payroll-reported paid hours allocated to the relevant hirer and assignment are used.
This distinction is important when interpreting the public report. An annual gap may partly reflect differences in reported paid hours or employment for only part of the year. The hourly measure standardises each relationship by its own reported paid hours, while the employer must still report the unadjusted annual result.
The formulas for mean and median additional or variable pay do not contain a recipient-only filter. On a literal reading of the draft, all in-scope male or female employment relationships remain in those calculations, including relationships with a zero value for additional or variable components.
The greater-than-zero test appears only in the separate participation metrics that identify the share receiving additional or variable components.
For every median, the draft requires employers to:
The draft regulation calculates results using arbeidsverhoudingen, or employment relationships. For an employer's own employees, this corresponds to the payroll income relationship identifier.
One person may have more than one employment relationship with the same employer at the same time. Each relationship can therefore enter the calculation separately. A person may also be associated with more than one worker category where different employment relationships relate to different work.
For own employees, the draft instructs employers to select only relationships where:
The draft also specifies labour-relationship codes for the population:
Fictitious employment relationships are excluded from the own-employee population.
The population is territorially connected to the Netherlands through Dutch tax liability. For an employer established outside the Netherlands, the draft adds conditions connected to working in the Netherlands or having an employment contract governed by Dutch law.
The draft recognises four possible administrative gender entries: man, woman, X and not completed. The pay gap itself is calculated only between employment relationships recorded as male and female.
The methodology says that a non-binary person recorded as such may tell the employer whether they wish to be included as a man, included as a woman or left outside the male-female pay gap calculation.
There is, however, an important detail in the current formulas. For the variable-pay participation and quartile metrics, the denominator is drafted as the total number of employment relationships in the relevant reporting population, including male, female, X and unknown records.
Taken literally:
The same logic is used for men.
X and unknown records are also included when employees are ranked to create the annual and hourly quartiles. This is not the calculation many employers would infer from the labels in the template. If the wording is intentional, the figures describe each gender-and-quartile cell as a share of the whole relevant reporting population. If it is not intentional, the formulas need correction.
The draft also says to divide the ranked population into four equal parts but does not provide a detailed rule for remainders or ties at a quartile boundary. These are important points to clarify during consultation and ones that software providers should keep configurable.
The Dutch draft does not simply add supplied workers to the employer's own payroll population.
For this reporting obligation, the explanatory note states that only supplied workers with an employment contract, including a temporary agency work contract, are counted.
Where supplied workers are in scope:
The lender holds the payroll data, while the hirer has the reporting obligation. The draft regulation therefore requires the lender to provide the hirer, per person and male/female gender specification and as non-identifiably as possible:
The same paragraph says that the lender supplies the requested pay components only for male and female workers. The drafts do not explain how this restriction is meant to operate alongside the general formulas that include X and unknown records in participation denominators and quartile ranking. This is another point that needs clarification for the supplied-worker report.
The lender allocates the data to the relevant hirer and assignment, but the annex says that the hirer does not have to retain a link to a particular assignment in its reporting dataset. A supplied worker is identified by a citizen service number or, where that number is not known to the hirer, by the personnel number under which the worker is registered with the hirer.
Only compensation attributable to the particular hirer and assignment enters that hirer's calculations. The explanatory note says that an employee pension contribution and the taxable benefit of a company car are not included in the supplied-worker section because those components cannot be directly allocated to a hirer. This is a supplied-worker rule; for own employees, taxable pay provided in kind is included in additional or variable components.
The supplied-worker pay formulas differ from those for own employees:
The lender does not include workers it has lent out in its own report. It reports its own employees and any supplied workers that it has itself engaged. If a worker is temporarily not supplied to a client and works for the lender's own organisation, the draft treats that period as work by the lender's own employee.
The draft decree also uses the same gross-pay definition for the employee right to information and requires the information to be separated between own employees and supplied workers. A supplied worker wishing to exercise that right approaches the formal employer, meaning the lender.
For employers, the practical implication is clear: compliance will require an agreed, repeatable and privacy-conscious data exchange with every relevant staffing supplier. An HRIS or payroll-only project will not be sufficient.
The draft template draws a clear line between organisation-level and category-level reporting:
The public-facing data may also include:
Contact details are not published. The size class, collective agreement information and number of worker categories are also described as non-public, although they may be used for official analysis.
Where an employer submits a replacement, the most recent version is intended to be published. This makes version control and the ability to reproduce a submitted result particularly important.
The draft methodology also states that information for the preceding four years is to be provided on request where available. Employers should preserve the source data, calculation logic and submitted outputs rather than treating each filing as a stand-alone annual exercise.
The report will be submitted through an online portal using eHerkenning at assurance level 3. The employer's board must confirm that the information gives a faithful representation, and the explanatory note says that the works council must be involved through information and consultation.
The draft decree also provides for the publication of Labour Inspectorate information for up to three years. This may include sanctions and warnings, but also confirmation that an inspection found no infringement.
The threshold is not necessarily a raw payroll headcount. The underlying bill states that:
The draft regulation defines an annual work unit by reference to the annual hours that a full-time worker would have to work under the employment contract. Employers should therefore not assume that the 100, 150 and 250 thresholds can be tested by simply counting payroll records or unique people.
The threshold weighting must not be carried into the pay metrics. Part-time work is fractional for the size test, but each in-scope employment relationship has equal weight in the annual and hourly group means described above.
The draft documents set out the following timetable:
Although the first filing date may appear distant, employers in the 150-plus groups would be reporting 2027 payroll data. Data design, category governance and supplier arrangements therefore need to be ready before the reporting deadline itself.
The drafts may still change, but they are detailed enough to identify the main workstreams.
Identify the fields required for loon LB/PH, the special remuneration table, non-cash taxable pay, paid hours, holiday allowance accruals and payments, and accruals and withdrawals from employment conditions amounts.
Do not rely on labels such as "allowance", "bonus" or "benefit". Test how each component is processed in Dutch payroll and how the draft adjustments change its final classification.
Build separate annual and hourly datasets. Do not normalise the annual result for full-time equivalence or partial-year employment. For hourly reporting, calculate each relationship-level rate from annual pay and annual payroll-reported paid hours before calculating the group mean.
Make sure payroll income relationships can be identified consistently across the reporting year. Do not deduplicate people with multiple concurrent relationships. Preserve zero variable-pay observations in the mean and median populations unless the final methodology introduces a different rule.
The reporting design assumes categories covering the same or equal-value work and one consistent structure for own and supplied workers. The current template has capacity for 200 categories. This is both a job-architecture and data-governance task.
The calculation engine needs to produce 26 public percentages per population and four additional percentages for every category. It should retain the underlying values and evidence even though the template requests the calculated metric outputs only as percentages.
Agree data content, frequency, identifiers, assignment allocation, privacy controls and correction procedures with staffing suppliers. The required data cannot be reconstructed from the hirer's own payroll.
Confirm the reporting entity and RSIN, especially in group structures. The draft says that a subsidiary normally reports where it is the employer, unless employment conditions are completely and mandatorily set at holding-company level.
Set up board confirmation, works-council involvement, eHerkenning access and a controlled approval process before submission.
Do not use raw headcount as a shortcut. Build a separate threshold calculation that combines own employees and supplied workers and applies the annual-work-unit treatment required for part-time work.
The public report consists mainly of percentages and includes unadjusted annual results. Employers should prepare a clear explanation of what the annual and hourly figures show, why they may differ and what action is being taken. The optional company website link creates a formal place for this explanation.
Do not hard-code the current denominator rules or the category-level variable-pay formula. The documents are still under consultation and these points may be clarified or amended. The future portal, spreadsheet and step-by-step employer guidance are also not yet final.
The drafts are detailed, but they are not complete technical specifications. In particular, they do not yet provide:
basisjaarloon and basis-uurloon, even though those terms are used in the category formulas; the decree defines basic pay as gross pay minus additional or variable components, but the annual and hourly implementation is left implicit;These are not alternative interpretations supplied by the employer. They are matters on which the current drafts are silent or internally unclear and which should remain configurable until final guidance is published.
The Dutch proposal is unusually concrete. It shows that pay transparency reporting will not be a single calculation added to an annual HR process.
It requires:
For many employers, the difficult part will not be the headline gender pay gap formula. It will be producing a complete, traceable and repeatable dataset that follows all of these rules at the same time.
