The Netherlands: What Regulation Delivered, and What It Did Not

Every argument about whether paid intimate work should be licensed, criminalised or left alone arrives, sooner or later, at the Netherlands. It is the case both sides reach for. A quarter of a century is long enough that a reader is entitled to expect a verdict, and one is usually supplied: the Dutch experiment worked, or the Dutch experiment failed. Neither sentence survives contact with the Dutch record, which is longer, more administrative and considerably more self-critical than either camp tends to report.

Before any of that, the thing a reader needs first. This house sells paid intimacy. It is not the trade the Dutch statute regulates — the arrangement here is companionship and touch, conducted under Japanese notification requirements — but it is close enough that we are not a disinterested narrator, and close enough that a finding in one direction would be commercially convenient for us. That is not a footnote to be parked in a closing disclaimer. It is the central problem of this essay, and stating it late would be a way of getting the argument accepted before the reader knew who was making it. The rule this chamber works to is that a party with a stake declares the stake before it argues, and then argues in a form that can be checked against sources rather than against our good faith. Every factual claim below is attributed to a document a reader can find.

The finding this essay defends is narrow, and it is not a recommendation. Regulation is not the opposite of harm. It is a redistribution of who bears it. Licensing in the Netherlands delivered real and checkable things — inspection, a legal footing, premises that could lose a permit, a formal counterparty to complain to — to people who could reach the licensed sector. It also created, as any licensing regime does, a remainder outside the licence, and the state repeatedly found that its own enforcement capacity was drawn toward the part that was already visible. The Dutch experiment therefore answered a narrower question than either side wants it to have answered. It did not test whether regulation as such reduces harm. It tested what happens when a country removes a prohibition, hands the design to its municipal councils, settles the labour question by an arrangement with the tax authority, and then spends twenty-five years failing to pass the national rules it never stopped drafting.

What follows reports what is documented and names what is disputed. Where the record is contested, the contest is described along with the reason for it. Where the essay is reading rather than counting, it says so.

The statute is short and it is worth being exact about, because a great deal of what is said about the Netherlands attributes to it powers it does not contain. The Act of 28 October 1999, published in the Staatsblad of 9 November 1999 and brought into force on 1 October 2000, removed from the Penal Code the general brothel ban and the offence of living off the earnings of another. In their place it inserted a new article making the exploitation of prostitution punishable, a provision that was subsequently tightened, widened and renumbered into the human trafficking article that stands today.

The second half of the change is the half that gets forgotten. The same reform inserted into the Municipalities Act a provision stating that a council may adopt a by-law laying down rules concerning the commercial provision of the opportunity to perform sexual acts with or for a third party against payment. That single sentence is the whole of the national licensing regime. Parliament did not build a system. It lifted a prohibition and handed the design of whatever replaced it to roughly three hundred and fifty municipal councils, each free to license, to cap, to zone, or to permit nothing at all.

This chamber has set out elsewhere, in its essay on the districts and what every country decided to do about this, that there are five distinct containment decisions in the world rather than one, and that the licensing answer buys the worker a formal status at the price of an unlicensed remainder that darkens as the licensed part gets cleaner. The Dutch case is where that trade can be watched in official documents over twenty-five years rather than asserted.

One finding from the first serious evaluation deserves to be placed immediately, because it inverts the popular story. The research centre of the Ministry of Justice reported in 2007 that the legalisation had taken all the attention, but that what had actually produced most of the change was the stricter enforcement of the criminal forms. Workers and operators experienced the new situation as more strictly regulated than the old one, not less, because a tolerated arrangement had been replaced by an enforced one. A country that had abolished a prohibition had, in the experience of the people inside the trade, become harder to work in.

It delivered a legal footing, and the footing was tested at European level almost immediately. In November 2001 the Court of Justice held, in a case brought by Polish and Czech nationals refused residence permits to work in the Netherlands, that the activity of a prostitute is ordinarily that of a self-employed person, pursued outside any relationship of subordination as to the choice of the activity, the conditions of work and of remuneration, under the person’s own responsibility and for remuneration paid directly and in full; and that it therefore falls within economic activities as a self-employed person for the purposes of the association agreements then in force. A Dutch court had asked whether this was work. The answer was that it was.

It delivered inspection, and the inspection functioned. The 2007 evaluation found that licensing was practically everywhere in order, that controls took place everywhere to a greater or lesser degree, and that the lack of uniformity noted in the first evaluation had diminished. Premises-based businesses operating without a licence were a clear minority. The researchers who studied working conditions in the licensed sector reported few genuine abuses, though they also recorded large differences between businesses, and noted that workers responded to bad conditions in part by moving to another business.

These are not nothing. Against a regime in which the entire arrangement is criminal, a sector in which an operator holds a permit that an authority can withdraw, in which an inspector may appear, and in which a person can move to a different premises without becoming a criminal by doing so, is a materially different place to work. Any account of the Dutch case that does not concede this is not describing the Dutch case.

The 2007 evaluation, however, put a second sentence directly beside the first, and the second sentence is what the rest of this essay is about.

The second sentence was this. Labour relations inside the licensed businesses had hardly changed in the years since the reform, and there was no meaningful improvement to speak of. The legal position of the workers was poor. Operators and workers maintained to the tax authority that the workers were self-employed while, at the same time and on a large scale, operators involved themselves in the work in ways that in fact constituted relationships of authority. The consequence the evaluators singled out is the one that matters: notwithstanding the factual existence of those relationships of authority, the risk of not being able to work was shifted entirely onto the workers. They added that improvement was not to be expected to arise from the sector by itself, and suggested that minimum requirements on labour relations might have to be attached to the licence. They also recorded that only six per cent of municipalities reported that their policy addressed exit options at all, and that the measured emotional wellbeing of the workers interviewed was lower on every aspect than it had been at the first evaluation, with use of tranquillisers up.

What happened next is the most instructive single episode in twenty-five years, and it is documented by the state that did it. In 2008 the tax administration, together with the finance and social affairs ministries and the benefits agency, negotiated with the industry an alternative to employment. Operators objected to employment contracts on constitutional grounds relating to bodily integrity and on business grounds, notwithstanding case law pointing the other way. Workers, for their part, preferred to work independently, did not see themselves as employees, and placed a high value on anonymity. The result, in force from 2009, was the opting-in arrangement: a fictive employment relationship for tax purposes, under which the operator withholds and remits, and the worker is not an employee. At its start in January 2009, three hundred of the five hundred and ninety operators of clubs and escort businesses signed the agreement with the tax authority.

The government’s own evaluation of that arrangement, conducted a few years later, is unusually candid. It records that the workers criticised the arrangement for providing no safety net in the event of illness or incapacity for work, and then observes drily that this is precisely the consequence of opting-in: they cannot call on one. It records that it is the operator who chooses the form of the relationship and the worker who chooses an operator, so the choice the workers were said to prefer was not theirs to make. It records that the workers named the percentage of their earnings handed to the operator as producing a false self-employment, and that they did not experience themselves as independent of the operator. And it concludes that whether the second objective of the arrangement — a more independent position for the workers — had been achieved was an open question.

Set that against the argument most often made for licensing anywhere, which is that bringing a trade inside the law brings its workers inside labour law. In the Netherlands the workable arrangement was purchased by trading that away, deliberately, with the tax authority in the room, and the state wrote down what had been traded. This is the essay’s thesis in its cleanest documentary form. A real legal footing was delivered. A real protection was not. The two were not independent of one another.

A licence divides a trade into a licensed part and a remainder. What the Dutch documents show, repeatedly and from different directions, is that the state’s attention flows toward the part that already holds a permit.

The 2007 evaluation found that the police remained the principal inspecting party in the licensed sector, carrying out controls that could have been performed by the labour inspectorate, and that a large share of available police capacity went on those controls, which limited the capacity available for a controlling and investigative role in the unlicensed part. It recorded the perception inside the trade that licensed businesses were checked more often than unlicensed ones, and observed that this undermined the willingness of licensed operators to comply and complicated the effort against trafficking. It found that businesses not tied to premises could simply move to a municipality where no licence was required or enforcement was lighter, and that escort work in particular retained that option. And it identified categories in which enforcement had no purchase at all: at swingers’ and sauna clubs, officially nothing is sold, everyone enters as a paying visitor, the operator officially has no involvement in what visitors arrange between themselves, and an inspector therefore has no ground on which to check anybody’s papers.

By the time of the most recent national baseline study, carried out for the same research centre and published at the end of 2021, the licensed part had visibly contracted. Of the municipalities that responded, just over half said they had a sex work policy at all, against three quarters in the previous evaluation, with the commonest stated reason for having none being that sex work barely occurs there and researchers noting that some councils were waiting for the national bill. Licensed sex work occurred in a third of the municipalities studied. Extrapolating from the licences counted, the researchers estimated three hundred and seventy-three licensed businesses in the country, against eight hundred and thirty-three in earlier research — while stating plainly, in the same paragraph, that the licence counts for a number of larger municipalities were not included in their study, which is the kind of caveat that determines what the figure may be used for.

Over the same year the same study monitored public advertising sites and recorded more than twenty-seven thousand unique advertisements posted, with between six and a half and seven thousand live in any given week. Those advertising overwhelmingly stated that they were women, overwhelmingly stated a non-Dutch origin, mainly received clients at home or offered escort work, and in only a very limited number of advertisements was there any indication that the work being offered was licensed at all.

Read carefully, that is not the story of a regime collapsing. It is the story of a regime whose covered part shrank while the activity did not, and whose remainder migrated to a medium that no municipal by-law was built for.

The clearest documentation of the redistribution is municipal, and it was written by an auditor rather than an advocate.

From 2007 Amsterdam ran a programme to reduce the window brothels and coffee shops in its oldest postcode district and transform the area’s economy. One hundred and eighty-six windows were designated for closure; the target was later reduced to one hundred and forty when the budget for acquiring the buildings was cut. By December 2017 one hundred and three of the designated windows had gone, together with nine exchange windows, and thirty-seven remained to be closed.

In June 2018 the city’s own audit office reported. It found that the state had become more visibly present in the area and that window prostitution had become more manageable, and that criminal activity had not been broken but had changed character and become more covert. On the workers it was blunt. Although closing the windows had produced a more orderly sector, the broadly shared view was that it had worsened the position of sex workers relative to the window operators; the principal reason was that it was above all the sheltered windows that had disappeared, which made a good workplace harder to find and to keep, and left the workers more dependent on the operators. And it added the sentence that makes this a case study rather than an anecdote: when the areas in which brothels had to close were chosen, it was not possible to weigh those interests, because of the rules of the zoning plan.

There is the mechanism, in a municipal audit report. The instrument was chosen for the neighbourhood. The cost landed on the workers. And the procedure that selected the streets made it structurally impossible to count that cost at the moment of deciding. Nobody in that story acted in bad faith, which is the point: the redistribution was a property of the instrument, not of anyone’s intention.

Utrecht is the other half of the same lesson. In 2013 the city revoked the licences of the operators of the window premises on the Zandpad and in the Hardebollenstraat on the ground that there was a reasonable prospect of trafficking victims working there in future; the highest administrative court later held that it had been entitled to do so. A replacement zone was planned, redesigned and tendered. The municipality’s own current information states that there is at present no place for window sex work and no street zone in Utrecht; that since 1 January 2025 a person may work from their own home, one to a dwelling; and that the city is developing one or two small locations offering at most fifteen workspaces to rent. That is the interval between closing a licensed sector and having a licensed replacement, and it is now into its thirteenth year. This essay makes no claim about where the people who had worked there went, because nobody counted, and a guess would be exactly the kind of sentence this piece exists to avoid.

Trafficking is part of the documented record and it is the most hostile terrain in it for figures. The reason is not that researchers are careless. It is that the quantity being estimated is defined by a criminal statute whose contents have changed, observed only through registrations by agencies that do not use identical criteria, and reported by organisations on both sides of a long argument who have an interest in its size. The responsible thing an essay can do here is report the contest and its causes rather than choose a number, and that is what follows.

The most careful Dutch attempt is a multiple systems estimation applied to the national registration dataset, published for the ministry’s research centre in November 2021 by researchers at Tilburg, Utrecht and Southampton. Its method is to model the overlap between six independent lists of observed presumed victims in order to estimate how many were never on any list. For the last two years it covered, it put the estimated number of presumed victims of all forms of trafficking at around five thousand a year, of whom about one thousand three hundred had been reported to the national coordination centre by any agency — a ratio of roughly one observed to four estimated, close to the ratio found in the previous Dutch study and in comparable studies in other destination countries.

Inside that total is the finding this essay wants the reader to hold. The estimated numbers for sexual exploitation peaked at six thousand five hundred and eighty in 2012 and five thousand seven hundred and thirty in 2013, and then fell to around two thousand a year across the following four years. The authors offer one possible partial explanation for the stability of that decline: a 2016 judgment of the Supreme Court holding that a conviction for cross-border trafficking for sexual exploitation of non-Dutch nationals also required coercion. If that explanation carries any weight at all, then a substantial part of a fall in a trafficking figure is a change in what the word counts, and any comparison across that boundary is comparing two different quantities. This is the single most important thing to understand about the numbers in this field, and it is why a figure quoted without its definition and its year is not evidence.

The same study supplies a second reason for caution against itself. From 2018 several private organisations stopped supplying case-level data to the registration centre under a strict reading of the new European data protection rules, supplying aggregate totals instead, which are unusable for this method; the researchers reconstructed the missing years by imputation across twenty-five datasets. The estimates for those years are therefore modelled twice over, and the authors say so.

At municipal level the picture is worse, and the Amsterdam audit office said so in the plainest available language. There is still no shared picture of the extent to which trafficking occurs in Amsterdam window prostitution; objective official information about its scale is not available; and, as at the start of the programme, estimates of the share of workers not standing behind a window voluntarily diverge sharply. The estimates the auditors heard in interviews for that study ranged from about five per cent to about seventy per cent. A spread of that width is not a measurement of anything; it is a record of what people believe. The auditors also noted that the absence of an agreed picture produced a second harm, in that operators and workers experienced the official portrayal as stigmatising, and that operators reported it remained difficult or impossible to obtain ordinary bank financing or insurance.

The 2007 evaluation had already said the same thing from the other end. Involuntariness, it recorded, is often very difficult for third parties to establish, and it is therefore practically impossible to say anything about developments in the number of people working under some degree of coercion. In the same study, of the workers interviewed in the licensed sector, eight per cent reported having entered prostitution under some form of coercion, and the fieldwork in the non-licensed part encountered signals of involuntary prostitution, though to a limited extent. For the most recent registration year available, the national rapporteur recorded nine hundred and forty-four victims of all forms of trafficking reported to the coordination centre in 2024, seventy-six more than the year before, while stating that under-registration is substantial and that certain categories are presumed to be considerably larger than the figures show.

The position this essay takes is therefore the uncomfortable one. Nobody can establish from this record whether licensing increased or reduced trafficking in the Netherlands. The estimate exists, its ratio to observation is stable, its internal trend is entangled with a change in the legal test, its recent years are imputed, and no design in the record isolates the effect of the 2000 reform from everything else that happened in twenty-five years. Anyone who tells a reader otherwise, in either direction, is filling a gap with a preference.

Secondary accounts of Dutch policy routinely garble what happened next, usually by reporting a bill as law. The parliamentary record is unambiguous and it is worth setting out.

In November 2009 the government introduced a bill to regulate prostitution and combat abuses in the sex industry. It proposed a uniform national licensing system, a registration duty for every sex worker, and an offence of using an illegal offer. The lower house passed it on 29 March 2011. In the upper house it stalled, principally over the registration of workers and the corresponding duty on clients to check.

In March 2014 the government introduced an amending bill that dropped the registration duty and the client’s duty to verify, and set a minimum age of twenty-one for sex work. The lower house passed that amending bill on 21 June 2016. The package still did not complete its passage.

On 26 January 2021 a fresh bill, the regulation of sex work bill, was introduced, described in its own explanatory memorandum as based on the two earlier bills, adjusted and supplemented, the reasoning being that further amendment would have produced a confusing and fragmented whole. It was declared controversial within days of introduction, and again subsequently, and was never brought to a final vote.

By a letter to parliament of 3 April 2026 the government stated that it would not continue with that bill: it still endorsed the aims, but the bill contained too many contested choices. In its place it announced that it would shortly submit a narrower bill on municipal supervision of sex businesses, intended chiefly to give municipalities a firmer legal basis for the processing of data, alongside separate tracks on raising the minimum age to twenty-one and on a prohibition directed at pimping, and an identity check on new advertisements introduced on the advertising platforms from 1 April 2026.

So the answer to the question what the Dutch regime is, in 2026, is: the removal of a prohibition in 2000, a one-sentence municipal by-law power, a tax arrangement from 2009, several hundred different local policies of which roughly half are now unrenewed, and seventeen years of failed attempts to write the national rules. That is not a model anyone chose. It is what remains when nothing passes.

The absence of data has a similar shape, and this is where the essay’s counter-thesis begins to press on it. The authors of the estimation study recommended that the state create a stronger statutory basis for the supply of the data that makes estimation possible, as some other European countries have done, and that the obligation be written into the funding conditions of the relevant private bodies. The quality of the national estimate is, in other words, a policy variable. It was identified as such, in a report commissioned by the ministry, and it was left where it was.

Carefully, or not at all. Dutch outcomes are the outcomes of Dutch instruments operating in a Dutch administrative culture, inside a labour-mobility regime that shapes who arrives, and alongside a tax authority prepared to negotiate a bespoke arrangement with an industry. Change any of those and the same statute produces something else.

Japanese law is built on a different axis, and this chamber has already dissected its machinery at length in its essay on the woman holding up the fiction, so the argument is credited here rather than repeated. What can be stated from the statute itself is this. The Anti-Prostitution Act of 1956 defines prostitution as intercourse with an unspecified counterparty for, or on the promise of, compensation. Its third article states that no person shall engage in prostitution or be its counterparty, and attaches no penalty to the act itself. The penalties in the Act fall elsewhere: on soliciting in a manner visible to the public, on procuring, on knowingly providing a place, on making a business of providing a place, and on making a business of having a person reside in a place under one’s control or designation and engage in prostitution there. The architecture criminalises the surrounding business while leaving the act prohibited and unpunished. It is not a licensing regime and it is not a prohibition of the Swedish or the American kind; it is a fifth thing.

One Japanese change of the last few years is worth naming precisely, because it is the opposite half of the Dutch one. As enacted in 1956 the Act had a third chapter on protective rehabilitation, and its statement of purpose included the taking of measures of protective rehabilitation directed at women who, in the light of their conduct or their environment, were at risk of engaging in prostitution. The consolidated text of the Act today contains neither that clause nor that chapter; it has two chapters, general provisions and criminal dispositions. The supplementary provisions of the Act on Support for Women Facing Difficulties, promulgated on 25 May 2022 and in force from 1 April 2024, repealed the Women’s Guidance Home Act, and the support function was rebuilt in a statute whose stated object is the welfare of women facing difficulties rather than the prevention of prostitution.

That yields a comparison this essay is prepared to make, because it rests on two statutes rather than on an impression. The Netherlands changed the market and left the worker’s labour position to a tax arrangement. Japan changed the welfare frame, moving the support of women out of a prostitution statute and into a support statute, and left the market’s architecture where it was. Each country reformed the half the other did not. Neither half, on its own, settles what happens to a person standing inside the trade on an ordinary Tuesday.

What does cross is not a model but a mechanism, and it is not Dutch. Any state that brings part of a trade inside a permit and not the rest produces a remainder, and the state’s attention will tend to flow toward the part that already holds the permit, because that part is legible and can be inspected. The useful question for a reader here is therefore not which country to copy. It is where this country’s remainder is, who is standing in it, and whether anybody is counting.

Four arguments, stated at full strength, because the chamber requires that the case against a piece be put by the piece.

The first is that a house with this business model cannot be a fair narrator here, and that a declared interest does not cure the problem. Disclosure addresses the reader’s ability to discount; it does not address selection. Every essay chooses which documents to read and which findings to place first, and selection is precisely where an interest operates, invisibly and often without the author noticing. A reader who concludes that this essay should be weighted down for that reason is reasoning correctly. The only serviceable answer is that the sources are named specifically enough to be checked against the essay rather than through it, and that a reader who wants the argument without the narrator should read the audit report and the evaluations directly.

The second is that the redistribution frame can be used to excuse any regime whatever. If every arrangement merely moves harm around, then no arrangement is worse than another, and anybody defending a status quo can reach for the frame and stop the conversation. That conclusion is false and the objection is serious. The reply this essay offers is that the frame is a description of a mechanism, not a scoring rule, and that it becomes informative only when it is filled in: how large is the remainder, who is in it, and what happened to them. In the Dutch case that is exactly where the record is thinnest, which means the frame is doing less work here than its elegance suggests. A frame that cannot be filled in with quantities should be held as a hypothesis, not deployed as a verdict.

The third is that Dutch outcomes may not transfer to any other country, including Japan, and that a chamber which uses the world as a mirror is always at risk of importing a conclusion with the story. This is not a caveat; it is a live possibility that would invalidate most uses of this essay. The Netherlands has a municipal administrative tradition, a particular relationship between the police and local government, an internal European labour market, and a legal culture in which a ministry publishes research that contradicts its own policy. Remove the last of those alone and none of the findings above would be on the public record to be read.

The fourth is the strongest, and the essay cannot answer it. The absence of good data is sometimes not an accident but a choice. A state that does not fund the counting, or that allows the supply of the underlying data to lapse and does not legislate to restore it, has decided not to know; and the uncertainty that follows is then available to every party in the argument, including the ones who benefit from it. The Dutch record contains a recommendation from its own commissioned researchers that the statutory basis be strengthened, and it contains the fact that it was not. That is not evidence of intent, and this essay does not assert any. But it is enough to say that the phrase the data is unclear should never be received as a neutral statement about the world. Somebody decides how clear the data is.

Against its own interest first. The argument for licensing that a house like this one would most like to be true is that bringing an intimate service inside a formal frame improves the position of the person providing it. That is the argument the Dutch record supports least. The Dutch state brought the sector inside a frame, and its own evaluation then found the legal position poor, the risk of not being able to work shifted wholly onto the worker, the sickness safety net absent by design, and the measured emotional indicators worse than before the reform. If formalisation were sufficient, that would not have been the finding. The house should say so plainly, because the finding runs against what the house sells.

What this house actually sells is narrower and should not be inflated into a policy claim. It is an arrangement with a stated scope, a documented agreement about what is and is not wanted, a word that ends the evening without explanation, and a provider bound by the document who has no standing to reinterpret it. The Library has argued elsewhere, in the essay on why paid or unpaid is the wrong axis, that visible structure is what makes warmth non-coercive rather than what makes it false, and that undocumented obligation escapes scrutiny precisely because it is invisible. That argument concerns the frame between two people in one room. It is not an argument about statutes, and this essay declines to let it become one. A frame that works between two consenting parties who both hold the document tells you nothing about what a licence does at the scale of a national trade, where the person holding the document is a municipality and the person carrying the cost is not in the room.

What the house cannot claim follows. It cannot claim that its own arrangement is a model for a regulatory regime, and it has not recommended any regime here. It cannot claim that the Dutch record vindicates anything it sells. And it should notice, out loud, that this chamber’s earlier essay on Sweden and Germany reached the parallel finding from the opposite direction: two rich democracies ran opposite experiments on the same question for two decades, and neither could demonstrate the outcome its advocates had promised. Three countries, three instruments, and the same shape of result. That is either the most important pattern in this field or the clearest sign that the field is asking a question its data cannot answer, and the essay does not know which.

It takes no position on decriminalisation, on the Nordic model, on licensing or on prohibition, and nothing above should be read as recommending or opposing any of them in any country. It does not claim that the Dutch experiment succeeded or that it failed; it claims that the experiment tested a narrower thing than that question requires.

It makes no prevalence claim about trafficking in the Netherlands or anywhere else. The estimates, ranges and registration counts above are reported with the boundaries their own authors placed on them, and the essay does not combine them into a number of its own.

It narrates no individual. Nobody in this essay is a person, no worker is quoted or characterised, and no composite figure stands in for anyone. Where survey findings are reported they are reported as aggregates from named studies.

It makes no claim about where the people displaced by the Amsterdam or Utrecht closures went, because that was not counted. It makes no claim about the current status of the Amsterdam relocation proposal, which could not be established from the municipality’s own published record at the time of writing. It makes no claim about the condition of the Dutch licensed sector after the 2021 baseline study beyond what the legislative record shows. And it makes no claim about Japanese law beyond the text of the two statutes cited and the difference between the 1956 text and the consolidated text as it now stands.

It credits rather than restates the chamber’s earlier work: the five containment models in the essay on the districts, the opposite experiments in the essay on Sweden and Germany, the mechanics of the Japanese arrangement in the essay on the woman holding up the fiction, and the argument about visible structure in the essay on why paid or unpaid is the wrong axis.

What is left, when all of that is subtracted, is a small and durable finding. A country removed a prohibition, and the removal was real. It then discovered that the instrument it had chosen distributed its costs by its own internal logic rather than by anyone’s intention: toward the unlicensed, toward the undocumented, toward the person whose interests the zoning plan had no procedure for weighing. Twenty-five years of official documents record that discovery with a candour that is itself worth something. The Dutch experiment did not answer whether to regulate. It answered what a regulator will be able to see, and who will be standing where the light does not reach.