Causal Link: How the FSV Destroyed Entrepreneurs

Executive Summary

The Fraud Signalerings Voorziening (FSV) was an internal Belastingdienst system operational from 2007 to February 2020 that registered approximately 180,000 citizens as fraud suspects without verification, due process or notification. This report identifies three causal chains that connect the FSV to the systematic destruction of entrepreneurs, particularly MKB businesses, ZZP’ers and childcare agency owners. Chain 1: FSV registration leads to “fraud” label, triggering increased surveillance, tax assessments, OGS qualification, asset seizure and bankruptcy. Chain 2: FSV labels childcare agencies as “facilitators,” leading to collective punishment of all connected parents. Chain 3: Discriminatory profiling via nationality data creates structural disadvantage for entrepreneurs with migration backgrounds. The State systematically destroyed evidence — FSV registrations, 9,000 appeal files and 64 million data vault files — justifying a reversal of the burden of proof.

What Happened

The FSV functioned as both an original selection mechanism and an active damage amplifier. Once registered, an entrepreneur’s FSV label followed them through every future interaction with the Belastingdienst. All directorates (MKB, Toeslagen, Particulieren, GO) had access. FSV data fed into Deloitte’s risk classification models as source data and unfiltered exports were used for statistical profiling and shared with the FIOD.

Entrepreneurs entered the FSV through three routes: via their own fiscal behavior (deviant income patterns flagged by Project 1043), via childcare agency/facilitator constructions (CAF 80/20 approach assuming guilt) or via discriminatory profiling (nationality as input criterion, “allochtoon” as first selection criterion for MKB startup investigations).

The Belastingdienst reversed the causal relationship: instead of the State proving fraud, the entrepreneur had to prove innocence. The “stoppertje” procedure — 160 messages with this codeword triggering immediate benefit cessation without substantive review — exemplified this reversal.

Evidence

Chain 1 is documented through the tax chain (R07): FSV registration → intensified surveillance (4 follow-up investigations/year for “allochtoon” startups) → tax assessments → OGS qualification (automatic at debts exceeding €10,000) → full recovery (“all-or-nothing”) → asset seizure → WSNP exclusion → bankruptcy. Each step is condicio sine qua non, reasonably foreseeable and reproducible across hundreds of entrepreneurs.

Chain 2 is evidenced by the CAF’s 80/20 approach: of approximately 630 childcare agencies investigated, ~2,200 families were affected through collective punishment. The PwC document reveals conscious escalation: “[Facilitator] is not pleased with the tone of the letters. That was also the intention.”

Chain 3 is technically proven: BVR Nationaliteit was embedded in the SAS model data pipeline by Deloitte, 11,236 tax returns were flagged based on second nationality (2012-2014) and 71% of victims had a migration background (CBS). The model ran with insufficient training data — “100% correct and 100% incorrect” — which was known at launch.

Evidence destruction is documented across five categories: FSV registrations deleted, ~9,000 appeal files prematurely destroyed, 64 million data vault files unsorted, 36.7 million Q-drive files inaccessible and 100 uncontrolled locally developed applications.

Analysis

The three causal chains each independently satisfy BW 6:95 et seq. and BW 6:98 causation requirements. The burden of proof should be reversed on three independent grounds: (1) unlawful evidence destruction by the State (following HR NJ 1994/583 Vejlby/Rijnhaven), (2) structural opacity — the FSV, “stoppertje” procedure and selective document provision created an information asymmetry favoring the State — and (3) the mechanism was foreseeable, reproducible and predictable.

Upon reversal, the State must prove that FSV registration was lawful, played no role in tax assessments, OGS labels were justified and the data vault contains no relevant information. Given the destroyed evidence, this is in most cases impossible.

Compensation remains grossly inadequate: material damage covered at 25%, immaterial damage at 5-10%, business loss/goodwill at 0-4% and statutory interest at 0%. The Rechtbank Noord-Holland case (ECLI:NL:RBNHO:2025:8961) illustrates the gap: €654,159 claimed, €30,000 awarded, 4.6% coverage.

Frequently asked questions

How long was the FSV in operation and how many citizens did it register?

The Fraud Signalerings Voorziening operated within the Belastingdienst from 2007 to February 2020. During that period the system registered approximately 180,000 citizens as fraud suspects, without verification, due process or notification. In July 2020 the Autoriteit Persoonsgegevens ruled the practices unlawful, discriminatory and improper.

What was the outcome of the June 2025 case at the Rechtbank Noord-Holland?

In its ruling of 19 June 2025, ECLI:NL:RBNHO:2025:8961, the court awarded an affected entrepreneur 30,000 euros of the 654,159 euros claimed. That amounts to coverage of just 4.6 percent of the actual damages. According to the research, the case illustrates the gap between suffered and compensated damage.

What share of the affected entrepreneurs had a migration background?

According to CBS statistics, 71 percent of the victims had a migration background. Between 2012 and 2014, moreover, 11,236 tax returns were additionally checked on the basis of a second nationality. Nationality had been embedded by Deloitte as fixed source data in the SAS risk models.

How many childcare agencies did the Combiteam Aanpak Facilitators investigate?

Of approximately 630 childcare agencies investigated, some 2,200 families were affected through collective punishment. The team applied the 80/20 approach, which reversed the presumption of innocence. The PwC working document shows the harsh tone of the letters was intentional.

Read also

• The butcher as gatekeeper: the debate — why the cabinet keeps the prosecution service out
• Who knew what about the data vault — who knew what about the data vault, and who is still there
• The full data vault timeline — the full timeline, 2019-2026
• 64 million files out of sight — the discovery and the nine-month delay

Sources

• R07, MKB-Impact Toeslagenaffaire
• R08, PwC Working Document FSV Effects
• R09, Deloitte Risk Classification 2013
• R06, FSV Stuurgroep 25 juni 2020
• ECLI:NL:RBNHO:2025:8961, Rechtbank Noord-Holland 19-06-2025
• HR NJ 1994/583 (Vejlby/Rijnhaven); HR NJ 1996/688 (Kleyensteijn/Swinkels)
• AP Report July 2020
• Commissie-Donner March 2020; Parliamentary Inquiry December 2020
• Inspectie OE 22-04-2026
• CBS Statistics victims toeslagenaffaire