The absorbed debt without integration — Statecraft

Jacob Huibers — Statecraft, April 2026

Context for readers outside the Netherlands

A brief primer on Dutch institutions and dossiers is useful before reading the paper proper.

The childcare benefits scandal (toeslagenaffaire) — between roughly 2005 and 2019, the Dutch Tax Authority unlawfully accused tens of thousands of parents of fraud in claiming childcare benefits, with catastrophic financial and psychological consequences. The Mark Rutte government resigned in January 2021. Recovery Operation for Benefits (UHT, Uitvoeringsorganisatie Herstel Toeslagen) is the dedicated organisation, with around two thousand staff in 2026, that processes the resulting compensation claims. Benefits Service (Dienst Toeslagen) is the wider organisation hived off from the Tax Authority in 2020 in the wake of the scandal; UHT sits within it.

The Groningen gas extraction damage: half a century of natural gas extraction by the Nederlandse Aardolie Maatschappij (NAM, a fifty-fifty joint venture of Shell and Exxon) in the northern province of Groningen produced earthquakes that damaged tens of thousands of homes. Until 2017, NAM itself handled compensation claims — making the cause of damage and the assessor of damage one and the same party. The institutional response moved through several iterations: Centre Safe Living (CVW, 2015), the Temporary Commission Mining Damage Groningen (TCMG, 2018), and from July 2020 the Institute Mining Damage Groningen (IMG) as an independent administrative body. Throughout these iterations the underlying liability question — between private operator and Dutch state, which had authorised the extraction and collected the revenue — has remained organisationally muted.

Box 3 is the Dutch personal income-tax category covering wealth. The Dutch Supreme Court ruled in 2021 that the standing imputed-yield levy in Box 3 violated the European Convention on Human Rights. Recovery of the seventeen years of overcharged tax is being executed by the same Tax Authority that previously collected and judicially defended the levy.

Court of Audit (Algemene Rekenkamer) is the Dutch national audit office, charged with annually examining the legality and effectiveness of central-government spending.

Commission Van Dam: a 2024 commission that reviewed the structure of compensation arrangements in the childcare benefits recovery operation and found the system, in its own words, “unmanageable”.

A recovery location within the same architecture

For anyone who wishes to understand the childcare benefits scandal, one institutional fact is illustrative. The Recovery Operation for Benefits (UHT), where in 2026 just over two thousand staff work on repairing the harm the Tax Authority inflicted on citizens, is a separate part of Benefits Service. Benefits Service was hived off from the Tax Authority in 2021 but remained inside the same Ministry of Finance, with partly the same IT systems, the same statutory frameworks and personnel drawn largely from the same labour market. The compensation is delivered by an organisation that institutionally is a near-relative of the organisation that caused the harm. That is not a design fault and not a temporary imperfection. It is what arises when a country builds recovery operations without bringing the original architecture into question.

The same pattern can be identified in Groningen, in a different register and on a different scale. The Nederlandse Aardolie Maatschappij itself handled, until 2017, the damage claims of owners whose homes had been disrupted by gas extraction. Cause and assessor were one and the same party. In 2015, the Centre Safe Living was set up as a cooperation between NAM and third parties; in 2018 the Temporary Commission Mining Damage Groningen took over the work; and in July 2020 the Institute Mining Damage Groningen was instituted as an independent administrative body. On paper, damage settlement is now organisationally detached from the cause. In practice, the IMG handles the civil-law liability of the same NAM, operates in the same geographical space, with residents whose long experience of the dynamic of the old system shapes their dealings with the new, and organisationally builds the same relationship between authority and aggrieved party that fed the original anger. A hiving-off of legal liability is not the same as a break with administrative logic.

Box 3 displays a third variant. The Dutch Supreme Court ruled in 2021 that the imputed-yield levy violated the European Convention on Human Rights. Recovery of the years of overcollected tax is being carried out by the same Tax Authority that had levied the tax for seventeen years and defended it in court. The apparatus that collected the error is the apparatus that must undo it.

Two to ten times

In Recovery State Netherlands I have shown the financial signature of the pattern. The childcare benefits scandal cost just over eight billion euros in recovery operations to the end of 2023, while the original fiscal interest in the disputed claims was a fraction of that. Box 3 recovery is expected to cost over eleven billion euros. The Groningen damage and reinforcement operation is estimated at at least fifty billion euros across its full life, many times the cumulative fiscal benefit gas extraction yielded the Dutch state in its final decade. The reassessments at the employee-insurance agency UWV are conservatively estimated at ten billion euros, the nitrogen recovery operation at twenty-five billion. The total over a decade is at least seventy-eight billion euros, and on a generous count one hundred billion. These are direct government expenditures. The societal costs — the loss of trust, the foregone life chances of affected citizens — are not included.

The ratio is comparable in each dossier. Recovery costs two to ten times what prevention or early intervention would have cost. The childcare benefits scandal would have been a fraction of its eventual price had the signals from 2013 onwards been institutionally heard. The Box 3 problem was technically solvable in 2017, when fiscal lawyers had identified the legal untenability. The Groningen damage would have been limited had the seismological warnings in NAM’s archives in fact led to production restrictions. It is not the figures that tell the story. It is the regularity.

This paper adds to that observation something that in Recovery State Netherlands was only identified, not developed. Recovery spending in the Netherlands does not produce a learning impulse in the organisations that caused the damage. It is institutionally absorbed, in parallel organisations or in hived-off successors, while the underlying structure remains intact. The debt is taken in. It is not digested.

Single-loop versus double-loop

The administrative-science literature knows this phenomenon under a precise name. Chris Argyris and Donald Schön distinguished, in their work on organisational learning, two levels of learning at which an organisation can respond to signals that its action is not working.¹ The first level, single-loop learning, is the correction of actions within existing assumptions about how the work is supposed to happen. The thermostat that registers too warm and adjusts is the classical example. The second level, double-loop learning, is the questioning of the assumptions themselves. The thermostat that asks whether the desired temperature was well chosen, and where necessary adjusts its reference frame.

Argyris worked through, into the 1990s, the observation that organisations confronted with dysfunction systematically remain in single-loop. They create defensive routines — predictable patterns by which individuals and groups protect themselves and the organisation from the painful information that the assumptions themselves are wobbling.² Compensation is organised, procedures are tightened, the personnel base is adjusted, but the values, the culture and the underlying theory of what the organisation actually does remain untouched. It is precisely the defensive routines, Argyris wrote, that prevent the most competent staff from breaking through to the place where learning genuinely hurts and where the real work would lie.

The Dutch recovery operations are, in this terminology, an extensive documentary record of single-loop on an unprecedented scale. With the childcare benefits scandal, the fraud hunt was ended, Benefits Service was hived off from the Tax Authority, UHT was set up, compensation arrangements were designed, and the Recovery Operation Benefits Act was brought into force. That is much movement at the level of action. What has not happened is that the Tax Authority, the Ministry of Finance and the House of Representatives have together made a double-loop on the assumption that an enforcement apparatus capable of setting aside a constitutional framework is an apparatus that procedural measures alone cannot correct. The apparatus has been patched. The framework within which the apparatus operates is intact. Whoever looks for the redesign of the relationship between Tax Authority, citizen, court and legislator that would prevent a next form of mass harm finds that redesign nowhere as an operational project.

In Groningen the same structure is visible. The chain NAM, CVW, TCMG, IMG, NCG is a sequence of organisational renewals, each as a single-loop response to the fact that the previous body could not carry trust. What is missing is the double-loop on the assumption that a country in which civil-law liability for hundreds of thousands of damage owners belongs to a private party, while the State authorised the extraction and collected the revenue, is a country that needs to bring its own liability allocation into question, both legally and administratively. That conversation has been held in fragments. It is not the drawing board on which the IMG stands.

Sensemaking that does not close

Karl Weick added a complementary concept.³ Sensemaking, the term he introduced in his 1995 standard work on the subject, is the process by which individuals and organisations retrospectively bring coherence to what has happened to them, in order to be able to act anew. When sensemaking fails, an organisation is left without a coherent story of what has occurred, and so without a foundation for learning.

In his famous analysis of the Mann Gulch fire of 1949, in which thirteen American firefighters died because their routines no longer held in a rapidly changing crisis, Weick described how a group under pressure lost its shared meaning structure and disintegrated into separate individual reflexes. The catastrophe was not primarily the consequence of physical impossibility, but of the collapse of the shared reality within which the group had operated.

The Dutch recovery operations show a milder but related form of sensemaking failure. For the childcare benefits scandal there exist at this moment at least three different coherent stories, alongside many variations. In one story the affair was the result of an unlucky combination of a strict statutory framework, IT systems that algorithmically flagged fraud, and an administrative culture that suppressed signals from above. In another it was the result of a deliberately chosen enforcement line in which tens of thousands of parents were knowingly treated as fraudsters for political reasons. In a third it was an instance of a much broader pattern of erosion of the rule of law of which benefits is one episode. Each of these stories has empirical support, and each calls for a different redesign of what must change. What an organisation cannot do in this situation is learn. For learning requires a story which the majority of those involved at least workably accept, and that story does not exist for the childcare benefits scandal — not within the ministry and not within society.

For Groningen something comparable holds. Was the Groningen damage a disaster that flowed from private gas extraction which the State regulated too late, or a disaster the State itself caused as the principal beneficiary of gas revenue, hiding its own liability behind a private construction? The difference between those two stories determines whether the recovery operation is a correction within the system or a fundamental revision of what the State owes its citizens in mining regions. Both stories are told in fragments. Neither is institutionally anchored in an organisational learning process that runs through the entire chain.

What the Court of Audit sees

The Court of Audit documents the financial consequences of this not-learning in its accountability investigations of recent years with notable consistency. In the 2024 accountability investigation of the Ministry of Finance, published in May 2025, the Court establishes that the Box 3 recovery operation and the recovery of the childcare benefits scandal place a heavy burden on personnel and IT capacity, that other matters as a result remain longer in the pipeline, and that there is only limited room for the badly needed renewal and simplification of the tax system and the benefits system.⁴ The renewal of IT systems for the three large tax flows, through which over 182 billion euros of taxes are collected annually, was again delayed in 2024, partly through the capacity that the Box 3 recovery operation absorbs. The Court attaches to this observation a warning it has been repeating since its 2020 accountability investigation. Changes in fiscal policy, it wrote again in May 2024, must not once more place a heavy claim on a Tax Authority that already cannot manage its old problems.⁵

For the benefits operation itself, the Court reported in May 2025 that processing in 2024 had accelerated, that the first reviews had been completed, and that the target for integral assessments had been reached.⁶ But the same report warns that the acceleration is not enough to meet the cabinet target of completion in 2027, that the damage system, in the words of the Commission Van Dam, is unmanageable, that the differences between the damage routes lead to uncertainty, and that the procurement governance of Benefits has shown shortcomings since 2021 that the Court must again signal in 2024. The picture that emerges is not that of an organisation integrating its past. It is that of an organisation executing its past, and in that execution running aground on the same structural shortages — in personnel, in IT, in administrative weight — that caused the past.

This is what I mean in the title of this paper by absorption without integration. The recovery spending is processed by the organisation, booked, reported and accounted for. What it does not do is produce a learning impulse that brings the original architecture back to the drawing board. The apparatus absorbs its own debt — physically in the budget, organisationally in a parallel organisation, legally in a new act, communicatively in an altered narrative. The assumptions under which the debt arose remain intact.

What this does to the other symptoms

In the introductory paper of this series I identified four symptoms which the series develops separately. The absorbed debt without integration does not stand apart from the other three. It feeds and is fed.

The first symptom, the reputation architecture, touches this dossier directly. A recovery operation announced before the apparatus can carry it is a recovery operation that has been delivered front-stage before back-stage has built it. Each change of state secretary in the benefits dossier brought a new communication of what the recovery operation was, not always a new substantive advance in its execution. An apparatus of reputation is capable of announcing progress even where the substance lags behind, and that is precisely what typifies a dissociated organisation. Sensemaking failures of the kind Weick describes are not solved by reputation work but covered over.

The second symptom, the reproduction inwards, indicates why the redesign that double-loop would require does not in practice arise. Top officials change post every three to four years. A director who comes onto a dossier in 2026 has rarely lived through the errors of 2014 personally, knows the defensive routines of her predecessors more by hearsay than by experience, and on departure before 2030 has little incentive to question assumptions she did not herself help design. The institutional learning environment in which double-loop could take place has rotated apart.

The fourth symptom, performative maturity, indicates the form in which the single-loop materialises in the Netherlands. Codes, governance frameworks, integrity arrangements and compliance architectures form the structure with which organisations show themselves that they have learned, without having to touch their assumptions. The recovery operation gets its own act, its own organisation, its own governance, its own external supervisory chain. Each of those elements is single-loop. The double-loop, which would ask why the original framework permitted this in the first place, gets no comparable institutional carrier.

What would work

It is obvious to plead for prevention. Whoever takes the regularity of the pattern seriously presses for earlier signalling, for counter-power, for legal review at the administrative front end. That is right and is argued at length in Recovery State Netherlands. What this symptom paper adds is a different accent. The recovery operations themselves also need a design that makes double-loop possible, and that design is now missing.

Three design choices are not in themselves revolutionary. The first is the institutional integration of the learning impulse. A recovery operation should not only compensate the harm; it should feed the organisational learning back to the parent organisation through a formal commission and an independent audit. What UHT has learned in six years about the Tax Authority should not stay with UHT and evaporate with the wind-down of UHT. It should reach the Tax Authority itself, in its work processes, its IT architecture, its dealings with citizens and its accountability to the House. That requires a statutory link between recovery organisation and parent organisation that is now missing.

The second is an explicit sensemaking trajectory alongside the execution. Under the guidance of independent researchers — not the department itself — a coherent story of what has happened and which assumptions need to be questioned should be built up in a structured way. That trajectory should not arise only in a parliamentary inquiry or a commission Van Dam, but in parallel to the execution. A recovery operation without its own sensemaking layer delivers compensation, not learning.

The third is that the size of the recovery operation itself be used as a signal for redesign. When a recovery operation costs more than a multiple of the original fiscal, financial or economic benefit out of which the harm emerged, that should lead to an independent evaluation of the framework within which the harm could arise. Not as aftercare, but as a formal trigger condition. That is now missing. One hundred billion euros in recovery spending over a decade has, in the Netherlands, no institutional place where it must be collectively analysed. It is dealt with per dossier by whoever happens to be in charge.

None of these three choices requires radical constitutional revision. They require a different weight in the design of recovery operations. They require the recognition that a country that takes its debt in without digesting it is already building its next debt.

For interim management practice

For the interim executive or project lead this symptom is daily present in miniature. Whoever comes into a municipality processing a personal-care-budget fraud case, trying to clear a social-domain waiting list, or correcting an energy-allowance roll-out, operates in a micro-version of the macro patterns this paper describes. The question is then not only how the specific harm to residents is repaired. The question is whether the organisation that made the error is capable of questioning its assumptions, or whether the recovery operation will inadvertently work as a quarantine in which the original routines are protected.

Two disciplines help. The first is, in the final report of the assignment, to make an explicit distinction between what has been corrected and what has been learned. What the Strategic Triangle has yielded in operational capacity and public value is single-loop, and it is of value. What the Triangle has yielded in redesign of the assumptions beneath operational capacity and public value is double-loop, and it is rare. The final report should name both separately, so that the principal knows clearly what she does and does not have in hand after completion. The second is that the anchoring phase of the interim cycle must explicitly contain the learning impulse. What the interim executive has learned about the parent organisation in its dealings with the harm should return to the organisation itself — not as a communications product but as change in work processes, decisions and accountability. What in the forthcoming book is called transfer value is, in this symptom, the carrier of the learning that within the recovery operation itself does not arise spontaneously.

The open question

The introductory paper closed with the question of what recovery means if the norm towards which recovery directs itself has itself shifted. For the symptom of this paper a related question can be posed. Can a country actually digest its debt as long as the recovery operation stands organisationally apart from the organisation that caused the harm, and as long as the recovery operation itself is organised within a single-loop architecture?

What I suspect, and cannot prove, is that the threshold here is not financial. One hundred billion euros of recovery operations is, for a Dutch budget, manageable across a decade — however painful. The threshold is that double-loop has political costs no cabinet and no senior civil service voluntarily bears. Questioning the assumptions under which the Tax Authority operated in the benefits years touches choices of four successive cabinets and the role of parliament itself. Questioning the liability allocation in Groningen touches five decades of gas-revenue politics and the question of what the State owes its citizens in a mining region. Single-loop is cheaper than double-loop, not in euros but in political-administrative pain.

That does not mean redesign is impossible. It does mean that redesign will not arise spontaneously from the organisation that caused the harm, as long as there is no external carrier capable of holding the learning impulse institutionally. The Court of Audit does that work to the extent its mandate reaches. An independent national evaluation function for recovery operations could help where the Court’s mandates do not by themselves reach. What in any case does not work is for a country that takes its debt in without digesting it to hope the next debt will not present itself. That is the hope the figures in Recovery State Netherlands have already refuted.

Colophon

“The absorbed debt without integration” is the third symptom paper in the Statecraft series Dissociated Organisations. The series builds on Navigating versus Planning, the paper on vacation real estate and the invisible policy, the paper on scarcity, and the architecture of silence. The introductory paper Dissociated Organisations appeared in April 2026, followed by the symptom papers on the reputation architecture and the reproduction inwards. This paper builds directly on the institutional analysis in Recovery State Netherlands.

Statecraft is the platform of Jacob Huibers for strategic reflection on public-service delivery. The content connects to the forthcoming book The Direction of Movement: Interim Management in the Public Sector (autumn 2026).

Jacob Huibers is an interim manager with more than twenty years’ experience in the Dutch public sector. He has worked as cluster manager, cluster director and project lead for municipalities ranging from fifty thousand to over two hundred thousand inhabitants and for regional inter-municipal partnerships.

Chris Argyris and Donald Schön, Organizational Learning: A Theory of Action Perspective, Addison-Wesley, 1978. The distinction between single-loop and double-loop learning was introduced in this work and developed in Argyris’s later work. ↩
Chris Argyris, Strategy, Change and Defensive Routines, Pitman, 1985, and Knowledge for Action: A Guide to Overcoming Barriers to Organizational Change, Jossey-Bass, 1993. The analysis of defensive routines as a barrier to double-loop learning is central in these works. ↩
Karl E. Weick, Sensemaking in Organizations, Sage, 1995. Weick developed his analysis of the Mann Gulch fire in “The Collapse of Sensemaking in Organizations: The Mann Gulch Disaster”, Administrative Science Quarterly, 38(4), 1993, pp. 628–652. ↩
Court of Audit, Results of the 2024 accountability investigation of the Ministry of Finance, May 2025. Online: rekenkamer.nl/publicaties/rapporten/2025/05/21/vo-2024-fin. The Court points expressly to the impact of the recovery operations on IT renewal and the limited room for structural simplification. ↩
Court of Audit, Results of the 2023 accountability investigation of the Ministry of Finance and National Debt, May 2024. Online: rekenkamer.nl/documenten/2024/05/15/resultaten-verantwoordingsonderzoek-2023-ministerie-van-financien-en-nationale-schuld. The warning about the claim fiscal policy places on the Tax Authority dates from the 2020 accountability investigation and has been repeated since. ↩
Court of Audit, Childcare benefits recovery operation advancing steadily, treatment of additional damage worrying, May 2025. Online: rekenkamer.nl/publicaties/rapporten/2025/05/21/hersteloperatie-toeslagen. The observations on the damage system are partly based on the report of the Commission Van Dam on damage processing. ↩