Prioritisation of recovery work

Dutch Tax and Customs Administration

The algorithms Dynamic Monitoring (DM), Calling After Dunning (BNA) and Willing Can Quadrant-GG (CHP-GG) help Tax Administration staff keep track of outstanding tax debts. The algorithms also support in tracking agreements made on those tax debts.

Last change on 26th of June 2024, at 7:33 (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/dynamisch-monitoren-dm-en-bellen-na-aanmaning-bna/

Responsible use

Goal and impact

The Tax Administration uses an algorithm to actively track (monitor) outstanding debts and agreements made. We call that algorithm 'Dynamic Monitoring' (DM). DM works with data from various data sources and with signals from the algorithms 'Willen Kan Kwadrant-GG' (CHP-GG) and 'Call after demand' (BNA).

The CHP-GG and BNA algorithms can provide a signal to a Tax Administration employee if calling a tax debtor:

• contributes to the outstanding claim being paid, or
• that recovery arrangements are made.

The CHP-GG and BNA algorithms give signals only after the demand letter and before the enforcement order. CHP-GG does this in the case where someone has received a reminder for the first time in a long time.

After the enforcement order, the DM algorithm checks whether someone with a tax debt has sufficient (new) income or property that the Tax Administration can seize.

This is done for all outstanding claims where:

• an enforcement order has been unsuccessful and
• where no attachment has yet been levied and no agreements (such as a payment arrangement) have been made.

Signals are put in the most efficient order. This allows the Tax Administration to collect most of the outstanding debt with the least cost. Does a person with a tax debt expect to have sufficient (new) income or property? Then a signal is sent to a Tax Administration employee who assesses the situation. The algorithm DM monitors the progress of the seizure and the agreements made when deferred payment or a payment schedule has been agreed.

Considerations

These algorithms allow the Tax Administration to make the most efficient use of data known to us. This makes the collection of overdue debts better and more successful than when this was done entirely manually. When recovering debts, care is taken to ensure that a taxpayer's livelihood is not affected.

Human intervention

The decision whether or not to follow a signal is always made by a Tax Administration employee. So not by the algorithms.

Risk management

For the development of algorithms, the Tax Administration has drawn up conditions, a quality framework. This contains rules and agreements that are followed when developing the algorithm. The conditions of the Audit Service Rijk (ADR) are leading in this respect. When changes are made to one of the algorithms, the Tax and Customs Administration checks whether 'Dynamic Monitoring' and 'Calling after Dunning' still meet the quality requirements. The use of the relevant data is checked against relevant legislation. The AVG prescribes that we must not use more data than necessary. This is called data minimisation. The Tax Authority regularly examines whether the data used is still necessary and therefore may be used. If this is not the case, the algorithm is adjusted and this data is also no longer used. BNA and CHP-GG divide taxpayers into groups. This is done to maximise the chances of only calling tax debtors where there is a high chance that the outstanding claim will be paid (sooner) as a result. The algorithms use only the data listed in the table above for this purpose. The use of data is tested to avoid profiling that is not allowed.

Legal basis

The collection and use of the data described below is regulated by the:

• Collection Act 1990
• General Act on State Taxes
• Regulation implementing the General Act on State Taxes
• Act simplifying attachment free feet (and related regulations)
• General Administrative Law Act
• General Data Protection Regulation
• General Data Protection Regulation Implementation Act
• European Convention on Human Rights
• Charter of Fundamental Rights of the European Union
• Constitution
• General provisions Citizen Service Number Act
• Collection Guidelines 2008
• Code of Civil Procedure
• Archives Act 1995
• Tax and Customs Administration Selection Lists

Operations

Data

• Status of outstanding receivables | source: Tax Authorities | used in: BNA, CHP-GG and DM
• Payment details | source: Inland Revenue | used in: BNA, CHP-GG and DM
• Identifying data (BSN) | source: Basisregistratie Personen (BRP) | used in: BNA, CHP-GG and DM
• Financial products | source: Banks | used in: BNA, CHP-GG and DM
• Vehicle data | source: RDW | used in: DM
• Real estate data | source: Land Registry | used in: DM
• Wage data | source: UWV | used in: BNA, CHP-GG and DM
• Objection data | source: Belastingdienst | used in: BNA, CHP-GG and DM
• Turnover tax return data | source: Belastingdienst | used in: BNA, CHP-GG and DM
• Corporate income tax return data | source: Belastingdienst | used in: DMA
• Wage tax return data | source: Belastingdienst | used in: DM
• Income tax return data | source: Belastingdienst | used in: DM
• Company data | source: Chamber of Commerce | used in: BNA, CHP-GG and DM
• Result of previously handled signals from Dynamic Monitoring | source: Belastingdienst | used in: CHP-GG and BNA
• Result of previous telephone contact | source: Belastingdienst | used in: CHP-GG and BNA

Technical design

All three algorithms consist of decision rules created in collaboration with content experts and lawyers. These decision rules provide:

• estimating in which cases telephone contact will be effective
• monitoring possible (new) income and property of debtors
• monitoring the agreements made.

The algorithms are not self-learning. This means that they do not evolve while being used.