Weighing module request provisional assessment (RSA) and business rules first provisional assessment (EVA)

Dutch Tax and Customs Administration

This page provides information on the 'VVA' and 'EVA' algorithms for provisional assessments of income tax/national insurance contributions (hereinafter income tax).

Last change on 3rd of June 2025, at 9:48 (CET) | Publication Standard 1.0

• Publication category: Impactful algorithms
• Impact assessment: Field not filled in.
• Status: In use

General information

Theme

Public finance

Begin date

Field not filled in.

Contact information

[e-mail]

Link to publication website

https://over-ons.belastingdienst.nl/onderwerpen/omgaan-met-gegevens/algoritmeregister/

Link to source registration

https://over-ons.belastingdienst.nl/onderwerpen/omgaan-met-gegevens/algoritmeregister/

Responsible use

Goal and impact

Income tax is settled in the final assessment after the end of the relevant calendar year. A provisional assessment allows citizens to pay or receive monthly income tax during the calendar year. This is an estimate of the expected income tax for a current calendar year.

In an RSA, the citizen requests a provisional assessment. Or wishes to adjust a current provisional assessment. The Tax Administration must assess whether a provisional assessment can be imposed in accordance with the request. The Tax and Customs Administration uses an algorithm when assessing whether the request should be processed automatically or manually. To do so, the algorithm uses relevant data known to the Tax Administration.

With the EFA, citizens automatically receive a provisional assessment because a provisional assessment was imposed the previous year or the final assessment gives cause to impose a provisional assessment. The Tax Administration estimates the amount for which the provisional assessment should be set. The Tax Administration uses an algorithm to estimate this amount as accurately as possible based on the most recent final assessment or provisional assessment. Income and deductions that are one-off or no longer applicable are no longer taken into account. In addition, changes in the citizen's private circumstances are taken into account. As a result, the Tax Administration may also not impose an automatic provisional assessment and the citizen will be notified that he or she must reapply for a provisional assessment themselves if desired.

Considerations

For citizens, it is important that the provisional assessment is imposed quickly and just as just as possible. The algorithm supports this interest and the Tax Administration employee.

The algorithm contributes to the systematic and accurate checking of requests. By deploying an algorithm, requests can be processed more efficiently, resulting in citizens receiving a better provisional assessment. The algorithm determines the amount of an automatic provisional assessment. The most recent provisional or final assessment is used for this purpose. If it is not possible for the Tax Administration to make a good estimate, citizens are given the choice to submit a new request.

The algorithm determines the advice based on the relevant data. The alternative is that an employee would have to manually collect and assess the relevant data for all provisional assessments. This would make the process more error-prone and less efficient.

Human intervention

Human intervention in the Tax Administration context implies that a competent and knowledgeable employee plays a substantial role in decision-making.

The operation of the algorithm involves human intervention, but decisions are also made by the algorithm RSA. The algorithm detects, selects and makes decisions. If, based on the selection rules, the algorithm assesses that the request can be accepted directly, a provisional assessment follows automatically according to the data in the request. If the algorithm ejects a request for further assessment, the request can be handled by an employee. Human intervention by a clerk occurs if the processing results in a deviation from the request.

Risk management

The Inland Revenue is taking several measures to manage the risks of using the algorithm.

• 1. Transparency

The Tax and Customs Administration publishes information about the algorithm in the Tax and Customs Administration's algorithm register and the Central Government's algorithm register.

This allows citizens to understand the use and operation of the algorithm.

• 2. Automated decision-making

The description of an algorithm explains this and also details the aspect of human intervention.

• 3. Privacy protection and legality

The Tax Authority ensures that the use of the data is assessed against the General Data Protection Regulation (GDPR).

The description of the algorithm also links to the applicable laws and regulations.

• 4. Responsibility

On the basis of the Algorithm Register Policy Framework, the responsibilities and tasks in the use of algorithms are clearly described and secured.

• 5. Monitoring and evaluation

The operation of the algorithms is periodically tested. Based on this, selection rules can be adjusted.

Legal basis

1. General State Tax Act:
2. General Administrative Law Act:
3. General Data Protection Regulation:
4. General Data Protection Regulation Implementation Act:
5. Legal Restoration Act/ Box 3 Bridging Act
6. Income Tax Act 2001:
7. Citizen Service Number (General Provisions) Act:
8. Archives Act 1995:
9. Healthcare Insurance Act
10. Double Taxation Avoidance Decree 2001
11. International (tax) treaties

Links to legal bases

• Algemene wet inzake rijksbelastingen: : https://wetten.overheid.nl/BWBR0002320/
• Algemene wet bestuursrecht: : https://wetten.overheid.nl/BWBR0005537/
• Algemene verordening gegevensbescherming: : https://eur-lex.europa.eu/legal-content/NL/TXT/HTML/?uri=CELEX:32016R0679
• Uitvoeringswet algemene verordening gegevensbescherming: : https://wetten.overheid.nl/BWBR0040940/
• Wet Rechtsherstel/ Overbruggingswet box 3:: https://zoek.officielebekendmakingen.nl/stb-2022-534.html
• Wet Inkomstenbelasting 2001:: https://wetten.overheid.nl/BWBR0011353/
• Wet algemene bepalingen burgerservicenummer:: https://wetten.overheid.nl/BWBR0022428/
• Archiefwet 1995: : Archiefwet 1995: https://wetten.overheid.nl/BWBR0007376/
• Zorgverzekeringswet: https://wetten.overheid.nl/BWBR0018450/
• Besluit voorkoming dubbele belasting 2001: https://wetten.overheid.nl/BWBR0012095/
• International (belasting)verdragen: https://www.rijksoverheid.nl/documenten/circulaires/2023/03/09/verdragen-op-het-gebied-van-directe-belastingen

Elaboration on impact assessments

The use of the data should be assessed against the AVG.

The AVG prescribes that no more data should be used than necessary. This is called data minimisation. The Inland Revenue regularly reviews whether the data used is still necessary and therefore may be used.

Operations

Data

1. Wage, pension and benefit data
2. Personal data
3. Property data

Links to data sources

• Loon-, pensioen- en uitkeringsgegevens: Werkgevers, uitkeringsinstanties en pensioenuitvoerders
• Persoonsgegevens: Persoonsgegevens Basisregistratie Personen (BRP)
• Vastgoedgegevens: Kadaster en Landelijke voorziening WOZ

Technical design

The RSA algorithm consists of selection rules drawn up by content experts based on laws and regulations and tax expertise.

Based on the selection rules, the algorithm assesses whether the request can be accepted immediately. If not, the request is ejected and can be handled manually by an employee. With the ejected request, the algorithm mentions the topics the employee should pay attention to. Read more about this under the heading: 'Human intervention'.

The EVA algorithm consists of business rules created by content experts based on laws, regulations and tax expertise.

The algorithm calculates the amount of a provisional assessment based on business rules. The most recent final or provisional assessment is used as the basis. The business rules translate the current laws and regulations for the relevant year into a provisional assessment. This includes removing income and deductions that are one-off or no longer applicable. If the algorithm is unable to make a good estimate, the citizen is automatically informed that a request is needed to obtain a provisional assessment. This does not involve human intervention.

The algorithms are not self-learning. This means that the algorithms do not develop themselves while being used.