Why Dutch policy fails before implementation begins

On September 15, the Jetten cabinet is due to present its first Prinsjesdag budget, the annual statement of what the government intends to do and how it proposes to pay for it. The more awkward question is what happens after the promise. In May, the Algemene Rekenkamer, the independent body that audits national public spending and policy, reported that many short-term objectives for 2025 had been missed and important longer-term goals were moving out of reach. Housing and criminal justice were among the troubled fields.

Those are not Jetten’s 2025 failures. His cabinet took office on February 23, 2026. It has also done something more useful than insist that its predecessors lacked determination. Through Slagvaardige Overheid, its program for a more effective government, it has identified complicated rules, divided responsibilities, difficult data exchange, and organizational barriers as problems requiring reform. A June package combined simplification and digital modernization with the ambition to operate a smaller state apparatus. Six months is too soon to demand that an inherited system be repaired. It is not too soon to ask whether the repair plan fits the machinery.

The easy diagnosis is that Dutch government has become unable to execute its own decisions. It is also wrong in an important way. The same state that misses conspicuous policy targets maintains highly regular financial administration, and Dutch users report comparatively high satisfaction with everyday administrative services. Some targets also describe outcomes no ministry can directly produce. A completed disability assessment is an administrative output. A completed housing development needs public decisions, but also land, infrastructure, finance, construction capacity, and somebody willing to build.

The harder problem is conversion: turning a political objective into a system that can actually produce it. Law, money, information, specialist judgment, organizational change, and private behavior may all have to line up. If the rules require more insurance-physician time than exists, legislation cannot summon another physician. If several organizations control different parts of a criminal case, making one stage faster can merely move the queue. If a housing target depends on sites, grid connections, permits, finance, and private investment, the target is not a production order.

That is why some implementation failures begin before implementation. The policy can assume an instrument too weak for the promised outcome, assign responsibility without matching authority, or set a timetable that ignores the work needed to make the law executable. The Jetten agenda recognizes much of this. Its real test is whether the government can give its reforms enough authority, money, and transition capacity to change the mechanisms it has correctly identified.

A state that mostly works

Routine administrative competence matters because it rules out the laziest explanation. In its audit of 2025, the Rekenkamer found that 99.46% of central-government expenditure and 99.98% of receipts met its standards of lawfulness and numerical correctness. Its approval of the accounts came with qualifications, not a declaration that every part of financial management was sound. But these figures are not what a wholesale inability to administer public money looks like.

A different measure points in the same direction. The OECD’s 2025 country note, drawing on its 2023 survey, reported that 78% of Dutch users were satisfied with the administrative services they had used, compared with an OECD average of 66%. That is a measure of users’ experience, not a score for housing policy or the entire state. It can coexist with severe failures affecting particular groups. Its value here is narrower: routine Dutch administration is not generally dysfunctional.

Established administration and policy transformation ask different things of the same state. A mature service can have a settled legal basis, trained staff, known workflows, and information systems built around repeated decisions. Transformation changes those elements while the old service often has to continue. A new system may need to be built, data migrated, staff retrained, and several organizations moved onto compatible timetables before any promised efficiency appears.

Transformation therefore needs temporary capacity that the steady-state organization may not need later. The people who know the old process are often the same people needed to redesign it, test the replacement, train colleagues, and keep today’s cases moving. A reform can promise lower running costs and still require more effort during the changeover. That is not an argument against modernization. It is why the timing of savings matters as much as their eventual size.

That distinction also changes what counts as evidence of failure. Faster processing inside one organization may leave the citizen waiting somewhere else. A missed national target may depend on choices or constraints outside the civil service altogether. High satisfaction with routine services does not tell us whether Dutch cross-government reform is unusually good or bad by international standards. It tells us something simpler and still important: the problem to explain is selective, not general.

Failure before the file reaches the executor

A political objective is not an operating instruction. Between a promise and an outcome sits a chain of choices: the instrument, the legal rules, the money, the people, the information, the systems, the organizations that must act, and finally the response of citizens, firms, or other governments. A weakness at any link can emerge later as an “implementation problem”.

Take a new benefit with carefully differentiated eligibility. Parliament can describe the distinctions in a few sentences. Applying them might require medical evidence, household data held elsewhere, a software change, and a judgment by a scarce professional. An exception may make the entitlement fairer while increasing the work needed for every decision. That does not make the exception a mistake. It makes its administrative cost part of the policy choice.

The same applies to time. A commencement date is not credible merely because legislation contains one. If the responsible body needs eighteen months to change a system safely, a twelve-month start date does not create six months of extra capacity. It creates a choice: change the timetable, change the design, supply more transition capacity, or accept a higher risk of failure.

Dutch policy preparation already contains tools intended to expose these dependencies. The government’s Beleidskompas guidance distinguishes an uitvoeringsanalyse, an early look at how a policy might be carried out while alternatives remain open, from an uitvoeringstoets, a later test of whether a developed proposal can be executed and under what conditions. Where significant implementation consequences are expected, the guidance calls for testing before the proposal goes to the ministerraad, the cabinet’s decision-making meeting.

The sequence matters. An early analysis can still change the instrument. A late test may be left asking how to make an already favored instrument work. The test is not a statutory veto for the implementing body. Its practical force comes from making constraints visible early enough that ministers can choose what to do about them.

The Rekenkamer’s findings show why that information cannot be assumed. Across two years of audit work covering selected ministries, it examined 269 execution assessments. In 38%, there was no statement, or no clear statement, about the personnel consequences of the proposed law or regulation. Among 81 assessments involving inspectorates or supervisory bodies, 59% contained no staffing statement and 17% an unclear one, producing the 76% figure.

Those percentages describe the reviewed documents, not policy failure rates. They do not show that 38% of policies were understaffed, that 76% of supervisory proposals were unexecutable, or that ministers had no other information through budgets, correspondence, or oral briefings. They show that formal execution assessments often failed to record a clear staffing judgment. That increases the risk of a political decision being made without one of the basic facts needed to judge whether the work fits the available capacity.

Silence is not a harmless substitute for uncertainty. If the staffing effect is negligible, say so. If it is unclear, identify the uncertainty and decide whether the proposal needs further analysis, contingency money, or a phased start. An execution test need not manufacture precision. It should make the assumption visible.

Individual tests also cannot solve a portfolio problem. Two proposals can each look executable if both assume access to the same scarce specialists or the same narrow window for a major IT change. Ministers then have to decide which work takes priority. Capacity is not only an attribute of an agency; it is something government allocates among competing commitments.

There is useful contrary evidence. The Rekenkamer assessed Finance’s testing practice relatively positively. Tests at Social Affairs and Employment generally gave concrete judgments about executability, although staffing information could still be incomplete. At Health, Welfare and Sport, the Rekenkamer found both deficiencies and proposals changed after adverse findings. The instrument can work. The problem is that its completeness and political force are uneven.

Late changes create another gap. A parliamentary amendment can alter eligibility, add a duty, or accelerate commencement after the original proposal has been tested. Rodrique Engering, a program leader interviewed in the 2025 progress report of Werk aan Uitvoering, said quick tests of amendments still did not happen automatically. He pointed to both time pressure and hesitation within the administration. Amendments can be tested. They are not tested systematically enough to assume that the final policy carries the same capacity assumptions as the proposal that entered Parliament.

An execution warning should therefore force a political choice, not end a political choice. Supply the missing capacity, change the design or timetable, or accept the risk explicitly. The democratic decision remains with ministers and Parliament. So should responsibility for the conditions that make their decision executable.

The problem between the boxes

The person responsible for an outcome often does not control everything needed to produce it. A caseworker can see that an applicant is stuck but lack the legal authority to obtain a record. A ministry can fund a program while a municipality controls the site. One organization can pay for a digital change whose savings appear in another organization’s budget. The problem is not that Dutch government has “silos”. It is that law, money, data, and operational control can sit in different places without a reliable way to resolve the dependency between them.

Staat van de Uitvoering 2024, a synthesis of problems reported by public-service bodies, describes progress on individual rules and services alongside more persistent problems that cross organizations and policy fields. Complicated rules and data exchange recur throughout. The report is based heavily on organizations’ own accounts, so it does not prove a single national cause. It does show a repeated pattern that is also visible in Werk aan Uitvoering and in the cases below.

Funding can preserve the boundary even when everyone agrees the chain should improve. In the WaU progress report, an interviewed service leader argued that routing improvement money through departments encouraged proposals inside individual organizations rather than projects covering a whole chain. The logic is easy to see. A change can cost one body money while saving another body staff time. If nobody can finance the joint investment, both can rationally protect their own budget while the citizen keeps carrying the cost of the handoff.

This is why a chain can be full of individually defensible decisions and still produce a bad service. Each organization can protect its statutory task, budget, and performance indicators while an unfunded handoff remains between them. Solving that problem does not always require a new super-agency. It does require someone to see the complete transaction, identify where the cost and benefit fall, and bring the resulting choice to an actor with authority over the relevant law or money.

Data problems have the same need for diagnosis. Sometimes the law genuinely prevents a proposed use. Sometimes two systems cannot communicate. Sometimes nobody owns the decision or wants to accept the risk. WaU’s data-sharing taskforce found that difficulties in the cases it examined were often problems of cooperation and implementation rather than strictly legal or technical impossibility. That does not make privacy law imaginary. It makes “privacy” an inadequate diagnosis until the actual legal rule, system limitation, or institutional disagreement has been identified.

The citizen often becomes the integration layer. If one office tells an applicant to obtain evidence from another, both organizations can meet their own processing standards while the complete service remains slow and burdensome. A reform that saves staff time by shifting paperwork to people outside government has not necessarily reduced work. It has moved it.

Digital change can make this worse before it makes it better. The old service must continue while data are migrated, staff are trained, and new standards are adopted. A shared system can eventually reduce duplication, but only if organizations can align budgets, migration schedules, and technical choices. A common standard nobody can afford to adopt is still an implementation problem.

Some boundaries are also there for good reasons. Under the Kaderwet zelfstandige bestuursorganen, a zelfstandig bestuursorgaan or ZBO is not hierarchically subordinate to a minister, even though ministers retain specified information, policy, and supervisory powers. The precise relationship depends on the applicable law. Municipal democratic authority and judicial independence are different arrangements again. None should be collapsed into a generic complaint that the minister is not “in charge”.

The right question is therefore not how to abolish every boundary. It is what happens when the actor accountable for an outcome discovers that a necessary decision sits elsewhere. Is there a legal route to obtain the information? Can a joint investment be financed? Can a conflict be escalated to someone who actually has authority to decide it? If the answer is no, responsibility without the means to act becomes a promise to coordinate by goodwill.

Rules deserve the same discrimination. An exception may protect a group that a broad rule would treat unfairly; a review procedure may protect against a consequential error. Removing either can simplify administration by transferring risk to the person affected. Other requirements merely force the same fact to be proved twice. “Simplification” is useful only after government has decided which work protects something worth keeping and which work does not.

Three failures, three mechanisms

The three cases that follow are deliberately different. WIA shows that capacity is partly made by the rules: the law and assessment design determine how much scarce professional judgment each claim consumes. Criminal justice shows that moving a queue is not the same as clearing the chain. Housing shows that a political target is not a production order. Treating all three as the same implementation problem would lose the point.

WIA: capacity is partly made by the rules

The WIA disability system makes the capacity problem unusually tangible. UWV must decide whether people who have been ill for two years qualify for income protection linked to their remaining ability to work and earn. That decision often depends on an insurance physician. If the law requires a professional judgment, UWV needs the professional.

UWV’s 2025 annual report recorded approximately 96,700 WIA applications and 74,500 completed claim assessments during the year. At year-end, 7,900 people had been waiting more than six months for an assessment. These are different measures: incoming applications, completed assessments, and the stock of long-waiting cases. Subtracting one annual flow from another does not produce the backlog. In the first four months of 2026, UWV completed more assessments than in the same period of 2025, yet the number waiting more than six months had risen to 12,400.

That is what a specialist bottleneck looks like. Insurance physicians are not interchangeable administrative staff. Their availability limits the number of judgments UWV can complete, while the assessment rules determine how much physician time each claim requires. More money can help recruit and retain specialists. It cannot create trained doctors on the commencement date of a new obligation.

The simplified route for claimants aged 60 and over shows that the required quantity of professional judgment is itself a policy variable. The measure, which allows many participating claimants to be assessed without an insurance physician, resumed on September 1, 2025. UWV identified the interruption of the earlier route as one reason assessment output fell in 2025. Changing the route therefore released scarce physician capacity for other work.

But this was not merely workflow optimization. In its policy framework for the 2025 extension, SZW itself described the temporary route as operating outside the ordinary statutory assessment method and acknowledged that some people could receive a higher benefit than a normal legal assessment would produce. UWV’s evaluation of the earlier 2022-2024 scheme found changes in applications, benefit awards, and employment. More awards are not automatically erroneous awards. The point is that changing how much professional judgment the system requires can also change substantive entitlement outcomes.

That is a policy trade-off, not an engineering tweak. A more demanding assessment may distinguish more precisely among claimants, but it also consumes the scarce time that determines how long other people wait. A simplified route can release capacity, but its effects on entitlement and work incentives belong in the choice. Implementation design is policy design.

Management still matters. UWV’s early-2026 report describes difficulties applying priorities consistently across districts and redirecting scarce medical capacity toward the work given priority. The queue therefore cannot be assigned a single percentage to law, demand, specialist scarcity, or management. WIA is useful because none of those explanations makes the others disappear.

For the person waiting, the distinction is not academic. An advance can provide money while a decision is pending, but it does not provide the certainty of a completed assessment. If an entitlement can only be delivered through professional time that the system cannot supply, the shortage belongs in the policy decision long before the waiting list becomes intolerable.

Criminal justice: moving the queue is not clearing it

A criminal case is not processed by one organization. Police investigate; forensic services may examine evidence; the Openbaar Ministerie, the public prosecution service, decides how to proceed; courts adjudicate. A victim or defendant experiences the whole sequence. Faster work at one stage can simply deliver more files to the next queue.

The Rekenkamer’s May 2026 assessment found persistent failure to meet the chain’s own timeliness standards, including for categories such as juvenile and sexual-offense cases. It also criticized the minister’s improvement approach: the plan it examined lacked a sufficiently developed diagnosis of causes, concrete measures, and a credible timetable, while information about where delays arose across the chain remained inadequate.

The criticism has not gone unanswered. On June 22, the government announced a TNO examination of how the whole system works, a joint chain-wide dashboard, and a deltaplan to be sent to Parliament in the first quarter of 2027, after the underlying work is expected by the end of 2026. That is recognition of the Rekenkamer’s diagnosis, not proof that the diagnosis has been solved. The test has moved from whether government sees the chain problem to whether the promised chain-wide response changes it.

The mechanism is straightforward. More completed police investigations can mean more work waiting for prosecutors. Faster prosecutorial decisions can increase pressure on court schedules. A new priority can displace another class of cases rather than expand total capacity. Without end-to-end information, a minister can fund visible activity at one stage without knowing whether the total delay will fall.

Handoffs can nevertheless improve. WaU’s 2025 report describes an example in the police unit Midden-Nederland in which forensic traces from sexual-offense investigations were sent on directly rather than held pending a separate request. The relevant forensic-research interval fell from roughly 50 to 34 days. Process coordinators and administrative staff were part of the wider effort. This was one regional subprocess, not the duration of a criminal case and not proof of cost-free efficiency. It shows something narrower and useful: a queue created by the handoff itself can be shortened.

Coordination also has constitutional limits. In its response to the audit, the Raad voor de rechtspraak, speaking as the institutional administration of the courts rather than as a court issuing a judgment, warned that ministerial instructions about matters such as scheduling and deployment could indirectly determine which cases receive priority. It accepted the performance problem while resisting the idea that better chain management requires executive control of judicial work.

That distinction matters. Shared information, realistic planning, and agreed handoffs are compatible with judicial and prosecutorial independence. Political direction of individual proceedings is not. Criminal justice therefore shows why coordination is a form of capacity in its own right, but also why “put someone in charge” is not an institutional design.

Housing: a target is not a production order

Housing is the strongest challenge to an administration-centred thesis. The Rekenkamer reported about 79,900 homes added in 2025, against the political ambition of roughly 100,000 a year. The additions include new construction and other routes such as conversion and dwelling splits; they are not a count of new-build completions alone. More importantly, a gap of about 20,000 homes is not a 20% civil-service failure rate. The ministry was never the national housebuilder.

National government shapes finance, regulation, infrastructure, and parts of spatial policy. Municipalities control important planning and site decisions. Housing associations and private developers must assemble viable projects. Land, environmental permission, electricity connections, construction capacity, and finance have to become available together. The housing audit identifies nitrogen constraints, grid congestion, and shortages of administrative capacity among the obstacles. Removing one does not guarantee that a project can move through the others.

This is what political overpromising looks like when the target is mistaken for the mechanism. An efficient ministry cannot create a suitable site by memorandum, install all the missing grid capacity, or force a private project to become financially viable. Nor can environmental law simply be scheduled away. A government can be responsible for building a credible policy without controlling the exact number of homes that emerge from it.

That does not make the target meaningless. Government can invest in infrastructure, change financial incentives, coordinate planning, and make political choices about land and environmental capacity. Some constraints that are physical today respond to policy over time. Others persist because solving them imposes costs someone must bear. The relevant question is whether the instruments plausibly move the constraints, not whether the target is socially desirable.

The distinction also changes accountability. Ministers should be judged on whether they chose instruments capable of moving the bottlenecks they can influence, whether they funded the public inputs they promised, and whether they revised the plan when a constraint became binding. They should not be credited with every home the market would have built anyway, or treated as if every missing home were a file their department failed to process.

Flex housing illustrates the difference. These are often factory-built, relocatable homes intended for temporary sites. Faster construction can remove one delay, but the home still needs a first site, connections, permission, and finance, and later may need somewhere else to go. The earlier government ambition was to grow toward 15,000 flex homes annually. In its detailed audit, the Rekenkamer identified 17,665 homes associated with three principal support measures over 2022-2025. That figure includes supported projects not yet completed and does not cover every route by which flex housing can be built. It is not a national completion count and cannot be converted into a clean percentage of the national target achieved.

The business case explains why the instrument can stall without anybody simply failing to process a file. A housing provider has to judge whether the period of use, future relocation prospects, and expected rent justify the investment. A subsidy can improve that calculation. It cannot guarantee a second site when the first temporary permission ends. If that risk makes the project unattractive, the design of the instrument matters as much as the speed of administration.

Better municipal capacity and stronger national coordination could still produce more housing. The evidence does not assign credible causal shares to the different constraints. It therefore supports neither “bureaucracy caused the shortfall” nor “administration is irrelevant”. The useful counterfactual is simpler: if ministerial administration were flawless tomorrow, would the land, grid capacity, permits, finance, and willing builders necessarily appear? No.

For prospective tenants and buyers, that does not make scarcity less painful. It makes the government’s obligation more concrete. A serious housing promise should say which constraints government can remove, which depend on other actors, and which trade-offs it is prepared to finance or impose. A target can discipline policy. It cannot manufacture the inputs.

What improvement looks like

Groningen is the useful counterexample to fatalism because the system improved without becoming simple. The damage and strengthening operation created after earthquakes caused by gas extraction combines technical assessment, compensation, construction, individual property decisions, and badly damaged trust. It is exactly the kind of public task in which “process more files” is too thin a definition of success.

The Rekenkamer found material improvement in 2025: almost twice as much physical damage was handled as in 2024, and more than two-thirds of the addresses in the strengthening operation were classified as safe by year-end. “Safe” does not mean every one of those addresses had completed physical strengthening, and the operation was far from finished. But the direction was real.

Those gains came after sustained political attention, additional resources, and substantial changes to the organizations and processes doing the work. The chronology is not a causal accounting exercise. Earlier changes under Nij Begun had themselves caused delay while new arrangements were worked out, and the evidence does not establish whether, or how much, each later change contributed to the 2025 gains. What it does establish is that a deeply troubled delivery system can improve when its rules, resources, and working arrangements are changed rather than treated as fixed.

At the same time, the government deliberately pursued a different kind of improvement. It wanted the process to become milder, menselijker en makkelijker: more generous, more humane, and easier for residents. The Rekenkamer says the minister maintained that choice even where it came at the expense of speed, and reported that residents were not yet more satisfied. That is not an embarrassing footnote to better execution. It is the point. A public administration can rationally choose a slower process if the faster one imposes unacceptable burdens on the people it serves.

Groningen therefore widens the definition of capacity. Throughput matters, especially after years of delay. So do accuracy, treatment, predictability, and the distribution of effort between the state and residents. An implementation reform that makes a spreadsheet improve by making citizens do more work has not necessarily improved the service.

The trade-off also explains why a reform cannot be judged by one speed target. A system can become faster by standardizing more cases, while deliberately reserving slower treatment for situations where individual circumstances matter. The policy question is where that boundary belongs. Capacity determines which choices are feasible; it does not decide which balance between speed and care is legitimate.

Werk aan Uitvoering offers the same warning at a smaller scale. Its program period runs from 2021 to 2031, and its progress report records local improvements in services, processes, cooperation, and digital systems while acknowledging that combined funding often makes clean attribution impossible. The lesson is neither that the long reform program has failed nor that another local success proves the system is fixed. Improvement is possible. The harder task is reproducing it where no single organization controls the necessary changes.

Does Slagvaardige Overheid move the constraint?

The Jetten government’s reform agenda should now be judged against three questions. Has it diagnosed the right problems? Does its machinery have enough authority and resources to solve them? And is the transition sequenced so that capacity disappears only after the workload does?

Has the government diagnosed the right problems?

On diagnosis, the answer is substantially yes. The government’s coalition commitments call for earlier involvement of delivery organizations, simpler rules, cross-organizational teams, common digital standards, and more specialist expertise inside government. The June action agenda added data sharing and changes to supervision. Those priorities track the mechanisms that recur in the Rekenkamer, Staat van de Uitvoering, and Werk aan Uitvoering material.

That is more significant than it sounds. The central problem is no longer that government has failed to notice that late policy changes, fragmented responsibility, awkward data exchange, or thin specialist capacity can wreck delivery. The program has moved beyond diagnosis and announcement in several areas. The question is whether the institutional answer is strong enough.

Simplification is one example. The government set a target of removing or simplifying at least 500 rules a year, and on September 2 SZW opened the Vereenvoudigingswet tweede editie SZW for public consultation. It is a five-measure departmental package, not an enacted government-wide simplification law, with intended commencement in summer 2027. That is real movement, but it also shows why counting rules is a weak outcome measure. Removing one requirement that consumes thousands of professional hours can matter more than deleting dozens of obsolete provisions. The relevant measure is work actually removed, decisions made more reliably, and burdens reduced without quietly discarding protections that matter.

The staffing agenda has a similar logic. The coalition proposed reducing external hiring from the 15.4% share of personnel expenditure it cited toward 10%, while building more specialist expertise in-house. Retaining knowledge can strengthen the state. But replacing a contractor with a permanent employee is different from eliminating the function, and a spending ratio says little about whether the people needed for a difficult migration or legal reform are available at the moment they are needed.

Does the machinery have enough authority and resources?

The March taskforce assignment gives ministers and state secretaries a shared program and a route for bringing cross-departmental problems into collective political decision-making. It does not itself create a general legal power to override another minister’s statutory responsibilities, a municipality, a court, or every ZBO. That is not necessarily a defect. Many of those boundaries are legitimate. The test is whether a problem discovered by the taskforce reaches somebody who can actually make the missing legal, budgetary, or organizational decision.

The Nederlandse Digitale Dienst, or NDD, makes that test concrete. The coalition promised a compact expert service with doorzettingsmacht, authority intended to push common decisions through rather than merely advise. By July, the government had specified a staged start rather than an empty box. A 10 July parliamentary letter said the NDD would begin during the summer with three lines of work: a breakthrough role on priority projects, promoting and setting common standards, and guidelines for modern IT development. It would start with existing temporary financing and capacity from BZK-linked organizations, while formal decisions on its final tasks, mandate, and structural financing were deferred to 2027.

That same letter made a consequential choice about transition costs. The NDS-Raad had advised giving the service a role in financing friction costs, migrations, and the retirement of organization-specific legacy systems. The cabinet did not make that an NDD task, leaving ownership of legacy reduction in principle with the individual organizations. This is not evidence that those costs are unfunded or that central coordination is impossible. It makes the financing question sharper: the service can press for a common standard while the organization being asked to migrate still owns much of the cost of getting there.

There is also evidence of practical progress. A July ministerial letter had identified a sovereign-government-cloud proof of concept as early work, and by August 31 the NDS cloud program reported that a technical design had been published for open review and realized in a proof of concept. Governance, financing, and procurement remained for later stages. That is precisely why the NDD is now a better test of the article’s argument than it was in June: parts of the wider digital agenda have moved from planning into technical work, so questions about mandate, migration, and who pays are no longer speculative abstractions.

The right evaluation is therefore neither “the NDD has no authority” nor “the government has solved digital fragmentation”. The initial design assigns the service work, technical staff, and temporary resources. Some common standards may be made more binding within existing Rijksdienst arrangements. But final NDD tasks, structural financing, and the reach of its own mandate are still being decided, and no general power has been established here to command municipalities, courts, or every autonomous body. A useful central service needs enough leverage to make genuinely common choices stick without turning every local digital decision into a new central bottleneck.

The same principle applies to organizational reform. The coalition is exploring whether some ZBOs, explicitly including UWV, could be brought under more direct ministerial responsibility. That remains an exploration, not an accomplished transfer. A reporting line may be the obstacle in a particular case. It is not an answer to a shortage of insurance physicians or a rule that consumes more physician time than exists. Institutional form should follow the diagnosed authority problem, not substitute for diagnosing one.

Is reform sequenced so that capacity disappears only after workload does?

This is where the savings program becomes consequential. The June announcement links Slagvaardige Overheid to an annual savings requirement rising to approximately €1 billion in 2030, alongside a separate efficiency requirement rising to €392 million, about €0.4 billion, from 2030. The two lines should not be collapsed into one number. Nor do they fall mechanically on every organization discussed here. The coalition annex protects or exempts several justice bodies and the execution costs of Groningen damage handling and strengthening from the relevant allocation, among other components; it does not establish a direct cut to UWV.

Saving money while improving government is not inherently contradictory. Removing unnecessary tasks, duplicated checks, or expensive external dependence can do both. The problem is timing. WaU’s report explicitly recognizes that digital renewal can create more work before it produces efficiency. Staff have to run the old service while building and learning the new one. Legacy systems do not disappear when a future saving is booked.

If capacity is removed before the workload disappears, the saving can become a queue. That is a transition risk, not evidence that the Jetten government has already produced one. The correct question is what happens if the expected productivity gain arrives late. Does the reduction wait? Does another budget absorb the transition? Does the operating service take the hit? Each answer allocates the risk somewhere.

The NDD design makes the same issue concrete from the other direction. Common standards can reduce duplication once organizations adopt them, but the July design leaves much of the migration and legacy burden with the organizations that have to change. That may be a sensible division of responsibility. It still means the fiscal plan has to leave enough room for them to carry out the transition before counting the savings from the new arrangement.

A convincing reform program should therefore connect savings to workload and milestones rather than to faith in eventual efficiency. Which task disappears? When is the replacement ready? Which scarce skills must be protected until then? What happens when the timetable slips? Those are not internal project-management questions. They determine whether political savings are real reductions in unnecessary work or simply costs displaced into waiting time and service quality.

The Commissie Toekomst Rijksdienst, established to advise on the future organization of the central civil service, is due to deliver its advice by the end of December. Its conclusions are not yet available to settle these choices. That should keep the structural design open to evidence without becoming an excuse to postpone practical changes that already have a clear mechanism.

The Jetten program therefore deserves more credit, and a more demanding test, than the cynical version of the story allows. It has identified several real constraints and has begun concrete work. The remaining question is no longer whether the program contains substance. It is whether the powers, resources, and sequencing are sufficient for the constraints the program itself says must be overcome.

An executable promise

The Dutch state can execute many things. The harder failures arise when politics promises an outcome before the law, authority, information, specialist capacity, organizational changes, and external inputs needed to produce it have been made mutually consistent.

WIA shows why “hire more staff” is too crude. The rules determine how much scarce professional judgment each case consumes, and changing the assessment route can change entitlement outcomes as well as capacity. Criminal justice shows why enough effort inside each organization is not enough when delay lives in the handoffs. Housing shows the boundary of the argument: even a well-run public chain cannot command land, grid capacity, finance, and private investment into existence.

That is why “implementation failure” is often a misleadingly late description. Some failures begin when an objective is attached to the wrong instrument, a timetable outruns the preparation, or responsibility is assigned without a way to resolve the dependencies. Others are not administrative failures at all. The useful discipline is to identify which kind of problem government is actually facing before prescribing more staff, fewer staff, more centralization, or another reorganization.

Slagvaardige Overheid understands more of this than a dismissive reading would suggest. The government has moved from diagnosis into experiments, legislative simplification, digital design, and new coordination machinery. Its unresolved test is harder: whether authority, transition money, and workload will change in the right order. A common standard that nobody can finance, or a saving booked before the old task disappears, recreates the problem inside the reform intended to solve it.

A political target should therefore be treated as a commitment to build an executable system, not merely to announce an outcome. Ministers and Parliament may still choose difficult goals, accept risk, preserve costly safeguards, or spend more slowly than impatient observers would like. But when the necessary conditions are missing, the choice belongs upstream. Leaving the promise intact and sending the unresolved law, queue, or financing problem to the executor does not postpone the political decision. It hides it.