What should Box 3 actually tax?

The Netherlands has spent years preparing to tax much of the annual rise in the value of people’s investments, whether or not those investments were sold. The bill intended to do this passed the Tweede Kamer in February 2026. It remains before the Eerste Kamer. In June, that chamber held over its final vote pending consideration of an announced amending bill, or novelle. The official plan still says the new system should begin on 1 January 2028.

Then the politics appeared to turn around. On 1 September, NOS reported, citing leaked Prinsjesdag papers, that the coalition parties wanted to leave the pending bill parked and work instead toward a full vermogenswinstbelasting: a tax on gains when they are realized. That report was consequential, but it was not legislation or even a published cabinet decision. On 3 September, the Senate dossier still showed the bill awaiting a final vote, no novelle had been published, and the government’s public timeline still described the existing bill as the basis for 2028.

The apparent reversal exposes the question that matters. Suppose a share bought for €100 is worth €110 at the end of the year. The investor is €10 richer in an economic sense, but has received no sale proceeds. Should the Netherlands tax that €10 now, wait until the share is sold, or apply one rule to shares and another to assets such as property and private companies?

For a listed share, annual taxation has a formidable case. The price is visible. A broker can report it. The owner can normally sell a small part of the portfolio to pay the tax. Taxing the gain every year prevents the owner from choosing the most convenient year in which to recognize it. It also avoids asking taxpayers and financial institutions to preserve acquisition records for decades.

Yet Box 3 is not only a measurement exercise. It is a tax system that must operate across assets, legal forms, family events, political cycles, and ordinary human understanding. Once some gains are taxed annually while economically similar gains are taxed only at sale, the boundary between those categories acquires a price. Funds, companies, debt, derivatives, and other wrappers can then change not merely how an investment is held, but when its return is taxed. The system gains annual precision for the easiest assets and creates a permanent classification problem around them.

The better permanent rule is probably the more uniform one: tax interest, dividends, rent, and other cash returns when they are received, and tax appreciation when a sale or another defined event crystallizes the gain. That rule sacrifices some annual accuracy for listed portfolios. In return, it offers a transaction price, usually a source of cash, a timing concept closer to Box 2, and a principle taxpayers can understand without first working out which statutory clock applies to their investment.

That conclusion does not make the route from today’s stopgap cheap, quick, or obvious. Nor does it make realization simple. A serious realization tax needs basis records, loss rules, treatment of gifts and death, controls on tax-motivated selling, rules for transfers between tax boxes, and an answer to indefinite deferral. Choosing the destination is one decision. Deciding what to do in 2028 is another.

The bargain the courts broke

Box 3 is the part of the Dutch income tax that covers savings and investments not assigned to Box 1 or Box 2. It is often called a wealth tax because wealth has long driven the calculation. Legally, however, it has been structured as a tax on income from wealth.

For most of its modern history, the system avoided measuring each taxpayer’s actual interest, dividends, rent, and capital gains. It applied a forfaitair rendement, an assumed return, to the relevant assets and taxed the resulting amount. The attraction was plain. A bank balance or property value could support a largely standardized calculation. The Belastingdienst did not need to reconstruct every purchase, sale, cost, dividend, renovation, or loss. Tax receipts were also less exposed to the mood of financial markets.

That administrative bargain became harder to defend when the assumed return diverged sharply from individual taxpayers’ actual results. The 2017 redesign tried to make the forfait more realistic by assuming that people with more wealth held a larger share in higher-return investments. It still taxed modeled portfolios rather than individual outcomes. A cautious saver could therefore be assessed as if part of the money had earned investment returns that never existed.

On 24 December 2021, the Hoge Raad’s Kerstarrest held that this design violated the prohibition on discrimination, read together with the protection of property, in the cases before it. The Court granted rights-based relief. The judgment exposed more than an unfortunate parameter. It broke the legitimacy of taxing one person’s presumed investment strategy when that person had followed another.

The government responded with restoration rules and then the Overbruggingswet box 3, the bridging law. These distinguished bank deposits, other assets, and debts and assigned different forfaitary returns to them. That was closer to people’s actual asset mix, but it was still not a calculation of each person’s actual return.

In June 2024, in judgments led by ECLI:NL:HR:2024:704, the Hoge Raad held that the restoration and bridging systems still failed where the statutory return exceeded the taxpayer’s actual return. The Court’s official summary of the judgments also explains what “actual return” means for this judicial remedy. It includes direct income and changes in the value of the taxpayer’s Box 3 assets, whether those changes were realized through sale or remained unrealized at year-end.

That holding is sometimes treated as if the Court selected annual accrual taxation as the permanent system. It did not. The Court had to decide how much relief was required when a forfaitary assessment exceeded the return relevant to the rights violation. Parliament faces a different question: which tax base should apply prospectively, with a complete set of loss, valuation, payment, anti-avoidance, and transition rules? The judicial measure of relief sets a legal floor. It does not write the future tax code.

Since July 2025, the statutory tegenbewijsregeling, or counter-evidence mechanism, has allowed taxpayers to establish a lower actual return. The enacted law is one-sided in an important respect. A taxpayer can move down from the forfait to a lower actual return, but a higher actual return does not normally replace the forfait. That protects taxpayers from the rights violation exposed by the judgments, yet it also gives them an option between methods whose value depends on the year’s outcome.

The resulting machinery is neither as simple nor as inert as a forfait sounds. In 2025, the government had to close an accrued-interest bond structure that produced a loss under the actual-return calculation in one year and allowed the taxpayer to fall back on the forfait when the corresponding income appeared later. Once a simplified default sits beside a one-way actual-return election, the tax authority needs both systems and rules governing the route between them.

Some legal room for a forfait may remain, but it is narrow. The Afdeling advisering van de Raad van State, the Council of State’s legislative advisory division, discussed a low-risk-return benchmark and rebuttable approximations in its advice on the pending bill. That was legislative advice, not a judicial blessing for any particular safe harbor. A deliberately low forfait would avoid much overtaxation but systematically undertax high returns. A rebuttable forfait recreates the need to measure actual return whenever the taxpayer challenges it. The old bargain cannot simply be restored by choosing a friendlier percentage.

The judgments therefore made material redesign, or continued individualized relief against excessive forfaitary assessments, necessary. They did not decide what the permanent architecture should be. The Netherlands still has to choose when appreciation becomes taxable income.

Why taxing paper gains looked so sensible

Consider the easiest case for annual accrual taxation: an ordinary portfolio of listed shares or exchange-traded funds held through a Dutch broker. The account begins the year at €100,000 and ends at €110,000, with no deposits or withdrawals. The annual gain is observable. If the tax law counts the €10,000 increase immediately, the broker can report the relevant values and the taxpayer does not decide when the gain enters the tax base.

This solves several problems at once. There is no need to know whether the investor bought one lot in 2008, another in 2015, and a third after a corporate split. There is no tax advantage in selling only the high-basis shares while leaving low-basis shares untouched. A taxpayer cannot postpone tax by retaining the winners, nor generate a special immediate deduction by selling losers while keeping the economic portfolio largely unchanged. Each year’s opening and closing values do most of the work.

The pending Wet werkelijk rendement box 3 uses this approach as its general rule. Under the text passed by the Tweede Kamer, direct income enters the calculation and annual changes in the value of most financial assets are taxed through a vermogensaanwasbelasting, an accrual tax. Appreciation on real estate and qualifying startup interests is generally deferred until realization.

The information advantage is substantial. The government’s parliamentary implementation account estimated that about 2.5 million of roughly 3.9 million Box 3 taxpayers in 2028 would have only bank and savings balances and investment accounts at Dutch financial institutions that could be prefilled in the income-tax return. For them, accrual does not mean hiring an appraiser or preparing a miniature set of business accounts. It can mean checking values already supplied by a bank or broker. That is a far stronger administrative case than the phrase “taxing paper gains” suggests.

Listed assets also weaken the familiar liquidity objection. A house cannot be sold one brick at a time, but a diversified securities portfolio is divisible. An investor who owes tax can normally sell a small portion. That does not make the cash-flow concern imaginary. A forced sale may come at an unwelcome moment, some shares are restricted or thinly traded, and the investor may reasonably object that the state has turned an unsold price movement into a current bill. Still, the practical difference between a €500,000 ETF portfolio and a €500,000 private-company interest is real.

Annual valuation is not confined to stock exchanges either. Dutch municipalities already issue a yearly WOZ-waarde, the statutory assessed value used for property taxation and other purposes, based on a valuation date one year earlier. The Waarderingskamer’s figures show objections for about 2.5 percent of homes in 2026; its 2025 data indicate that reconsideration changed the value for about 0.9 percent of all homes. Those counts are not an accuracy audit, and a low objection rate is not proof that every assessment is right. They do show that nationwide recurring residential valuation is an existing administrative system, not a technology the Netherlands would have to invent.

WOZ does not settle the case for taxing annual property gains. The valuation date lags the tax year. Renovations, unusual properties, local market shifts, foreign real estate, and non-residential property complicate the link between an assessed value and the owner’s annual investment return. Most importantly, an annual assessment says nothing about whether the owner has cash to pay tax on the increase. Valuation and liquidity are separate questions.

Losses pose the harder problem for accrual. Suppose the €100,000 portfolio rises to €110,000 in year one, is taxed on the €10,000 gain, and then falls to €90,000 in year two. Over the two years the investor has lost money. A coherent income tax must eventually recognize that result.

Unlimited loss carryforward helps, but it does not refund the earlier tax unless future gains arise. If the taxpayer dies, emigrates, or leaves Box 3 before using the loss, the correction can disappear. Carryback or current refunds improve symmetry but make revenue more volatile and create another verification task.

Even perfect loss relief cannot reverse the order of the cash flows. The taxpayer paid after the rise and receives relief only after the fall. For liquid securities this is a manageable objection to a technically accurate annual tax. For an indivisible asset whose value is debatable, the same sequence can become severe.

None of this defeats accrual for listed investments. On its strongest terrain, accrual measures nominal annual income more directly, limits voluntary timing, uses third-party data, and avoids lifetime basis records. A knowledgeable supporter should look at that case and wonder why the Netherlands would willingly give those advantages away.

The answer begins outside the brokerage account.

The line that creates a second tax system

The pending bill’s hybrid has an appealing rule of thumb. Tax gains annually where prices and liquidity are good; wait for realization where annual valuation or payment is difficult. A listed fund and an unusual parcel of land are not administratively equivalent. Treating them differently can look less like compromise than common sense.

The difficulty is that a statutory asset category is not the same thing as an economic exposure. A listed share can be locked up. An unlisted fund can offer regular redemption. A property company can deliver exposure similar to direct real estate. A debt instrument can reproduce part of an equity return. An investor can own an asset directly, through a fund, through a private company, or through contracts linked to its price. Each legal form carries its own reporting and regulatory consequences, but the underlying investment risk may be close.

Once annual accrual applies on one side of that line and realization on the other, timing becomes part of the product. A euro of appreciation taxed today is worth less to the investor than the same euro taxed years later. Taxpayers and financial firms therefore have a reason to move gains toward the realization side and losses toward the annual side. The boundary is no longer a neutral description of assets. It is a tax preference.

Box 2 makes the point visible without requiring exotic engineering. A person with a substantial interest, normally at least 5 percent pays Box 2 tax on dividends and gains on shares. An owner-controlled company can retain profits, postponing shareholder-level taxation until money is distributed or the shares are sold. Corporate tax is paid inside the company, and Box 2 has its own rates, loss rules, extraction rules, compliance costs, and anti-abuse provisions. Direct Box 3 ownership and a company are not interchangeable.

Yet the timing contrast matters: annual Box 3 taxation of a financial investment can make the corporate wrapper more attractive to an investor willing to leave returns inside it.

The Raad van State highlighted this interaction when criticizing the proposed system. A realization rule in Box 3 would not erase the difference between Box 2 and Box 3, but it would remove one conspicuous reason to change legal form merely to obtain a different tax clock.

Derivatives show why partial mark-to-market designs need careful safeguards, but the relevant example is narrower than a general attack on hybrids. In his 2013 paper on partial mark-to-market taxation, tax scholar Thomas Brennan examined the Camp proposal, which he interpreted as marking derivatives, including short positions, to market and treating their gains and losses as ordinary income while directly held corporate shares remained realization-taxed and generally received capital treatment.

His strategy exploited that combination of rate, character, loss-treatment, and recognition differences. Under the paper’s assumptions, it could produce negative effective tax rates; the historical examples were counterfactual back-tests, not evidence of observed taxpayer use.

Bill 36 748 draws a different line. Ordinary marketable shares and marketable derivatives generally sit on the annual-accrual side, while realization treatment is reserved mainly for direct real estate and qualifying startup interests. Brennan therefore provides neither a ready-made Dutch trade nor an estimate of Box 3 losses. His broader lesson is still useful: when economically related positions receive inconsistent rates, tax character, loss rules, or recognition clocks, the system needs consistent treatment or effective safeguards. Which Dutch boundaries would attract planning, and at what scale, remains unknown.

Depending on where the final boundaries fall, a hybrid may need matching or anti-conversion rules for straddles, synthetic exposure, related parties, matched debt, private vehicles, funds, and transfers between categories. It would also need information that lets the Belastingdienst see positions held across institutions and legal entities. A broker can report the year-end price of a share. It cannot necessarily know that a taxpayer has offset that share through a contract elsewhere, controls the private company on the other side of a transaction, or financed the position with debt receiving different treatment.

A hybrid is not two sensible methods meeting in the middle. It is both methods plus border control.

That does not make every distinction indefensible. Real estate can be genuinely illiquid. A private-company valuation can be expensive and contestable. Rules sometimes must recognize practical differences. The mistake is to treat those differences as if they naturally map onto durable legal categories and then stop the analysis. The longer the system lasts, the more products, wrappers, and ownership structures adapt to the line Parliament drew.

The democratic explanation also becomes strained. Under the hybrid, a rise in one taxpayer’s listed property fund can be current income while a closely related rise in directly held property waits for sale. A listed share in one company may be marked each year while an economically similar qualifying startup interest receives deferral. Tax specialists can explain the classification. The harder task is explaining why the resulting difference reflects equal treatment rather than the history of how the asset was packaged.

Comprehensibility is sometimes dismissed as a softer consideration than accurate annual measurement. It is not. Official reviews support a narrower claim: complexity makes taxpayer error more likely and increases demands on service, supervision, and voluntary compliance.

Sophisticated planning is a separate risk created by valuable differences in treatment, not by confusion alone. Political durability cannot be measured in advance, but a tax whose organizing principle is difficult to explain starts at a disadvantage. Maintaining two gain-timing systems, and policing the route between them, is not free merely because each method is defensible when viewed alone.

The Netherlands could still choose that complexity because annual marking offers substantial advantages for the roughly 2.5 million modeled taxpayers whose Dutch financial accounts could be prefilled. But the hybrid must be judged as a permanent institution, not as a clever solution to today’s known asset list. Its central weakness is the very feature that makes it look pragmatic: it assigns different clocks to gains that markets can make economically similar.

Why sale is the better clock

A realization system begins with a rule that can be stated without a flowchart. Interest, dividends, rent, and other cash income are taxed when received. Appreciation is taxed when the asset is sold or another event defined by law brings the gain into account. In an ordinary arm’s-length sale, the transaction supplies both a price and, usually, the cash from which tax can be paid.

“Usually” matters. A gift, migration, corporate reorganization, or transfer between tax boxes may trigger tax without producing cash. Some sales involve deferred payment or illiquid consideration. Parliament would need payment relief or continuity rules for such cases. The advantage is comparative, not absolute: realization ties the normal tax event to an observed transaction rather than to an annual valuation that may be provisional, contestable, or soon reversed.

That link has a legitimacy value that annual economic measurement does not automatically displace. The case for realization rests partly on an intuitive distinction between an unsold price rise and cash available to pay tax. Economists can answer that the owner is wealthier and can sell a fraction. For a liquid portfolio, that answer is correct. It may still leave an acceptance problem for a tax expected to last for decades. A system can be economically coherent and politically brittle at the same time.

Realization also provides one general timing principle for capital appreciation. It does not eliminate every boundary in Box 3, every difference between legal forms, or every incentive to plan. Cash yield must still be distinguished from appreciation. Exemptions, debt, foreign assets, and transfers remain. But it removes the statutory line between appreciation taxed every year and appreciation taxed only when realized. That is a significant simplification because the line otherwise attaches a financial reward to classification itself.

The timing alignment with Box 2 is similarly incomplete but useful. Box 2 taxes dividends and share gains at shareholder level, while profits retained in a company can postpone that shareholder tax. Taxing Box 3 appreciation on realization would narrow the contrast between holding an investment directly and holding it through a company. Corporate tax, control, the 5 percent threshold, compliance costs, and extraction rules would continue to shape the choice. Alignment is not identity. It is the removal of an avoidable mismatch.

Realization is also familiar internationally. The OECD’s comparative work on capital gains describes realization as the dominant approach among member countries, often combined with preferential rates, exemptions, or special rules. That prevalence does not prove superiority. It does show that a sale-based system is not an eccentric retreat from modern taxation. The recurring problems of realization taxation are well documented in countries that use it.

The largest problem is deferral. If a gain is taxed only when the owner sells, the amount that an accrual system would have collected remains invested. The longer the holding period, the more valuable that timing option becomes. A patient, well-capitalized household can obtain more deferral value than someone who must sell assets to finance consumption, although no Dutch incidence estimate establishes the scale of that distributional effect. At an identical statutory rate and without an offsetting charge, a realization tax therefore normally imposes a lower present burden than an annual accrual tax. An eventual larger nominal tax payment does not erase the advantage of retaining and compounding the amount an accrual system would have collected earlier.

That is why the relevant comparison is between approximately revenue-neutral mature systems, not between two systems carrying the same 36 percent label. Parliament could adjust the rate, exemption, inclusion fraction, loss treatment, terminal-event rules, or an interest charge on deferred gains. It could finance part of a transition elsewhere in the tax system. No official Dutch estimate yet identifies the package that would make a full realization system revenue-neutral, and pretending otherwise would convert a policy preference into fictional arithmetic.

Realization also changes behavior. Selling an appreciated asset accelerates tax, so investors may hold it longer than they otherwise would. Selling a loss can reduce tax, so they may realize losses while retaining gains. Research using US brokerage data by Ivković, Poterba, and Weisbenner finds both gain lock-in and tax-motivated loss selling, with stronger gain effects for larger transactions and longer-held positions.

Agersnap and Zidar estimate material but finite long-run realization responses and do not find capital-gains rate cuts self-financing in their US estimates. Hines and Schaffa add an important qualification: some dramatic short-run responses around tax changes may partly reflect anticipation of future rates rather than the ordinary long-run effect of realization taxation.

These studies establish the mechanisms, not a Dutch elasticity. A future Box 3 system would differ in rate, exemptions, terminal events, household portfolios, and interaction with pension wealth and companies. Lock-in deserves weight because it can impede rebalancing and delay revenue. It does not decide the choice by itself. The question is whether its cost, after mitigation, is greater than the permanent cost of maintaining different clocks across substitutable assets and legal forms.

Inflation adds another imperfection. Unless basis is indexed, a realization tax on a long-held asset includes nominal appreciation caused by rising prices. Accrual usually taxes nominal annual gains too, but the inflation component is less visible because it is collected year by year. Indexing basis could improve measurement while increasing records and complexity. This is a design choice, not an argument that one timing rule uniquely taxes real income.

The case for realization is not that sale is the only moment a gain exists. It is that sale is normally the best public event around which to organize taxation of appreciation: observable, comprehensible, commonly accompanied by cash, and capable of applying across asset categories without making legal packaging determine the tax year. Annual accrual wins the contest for precise yearly measurement of listed assets. Realization may still build the better tax system.

The simple rule’s complicated underside

“Tax the gain when sold” fits on a placard. A functioning realization tax does not.

Start with basis, the amount against which sale proceeds are compared. For new investments made after the system begins, brokers could report acquisition cost, transaction fees, corporate actions, and sale proceeds. Parliament would still have to standardize how lots are identified when only part of a holding is sold. First-in-first-out, average cost, and specific identification produce different timing. Transfers between brokers must carry the basis with the asset. Mergers, splits, distributions, inherited positions, and foreign accounts need rules that prevent basis from being lost, duplicated, or invented.

The transition is harder. An asset purchased years before the new law may contain gains accumulated under the old regime. Taxing the full historical gain at the first post-reform sale could reach appreciation Parliament never intended to include. Ignoring the past requires an opening value. The Netherlands could use the market value on a transition date for quoted assets, historical cost where records are reliable, or a mixed rule for older holdings. Each choice trades data burdens against valuation disputes and revenue. There is no official full-realization bill that has made that choice.

Losses require an equally serious design. A tax on gains over a lifetime should recognize losses over time, not merely within years that happen to be convenient for the Treasury. Unlimited carryforward is useful but can strand a loss if the taxpayer dies, emigrates, or no longer has Box 3 gains. Carryback or refunds provide faster symmetry but make revenue and administration more volatile. Restrictions are also needed to stop taxpayers from selling a loss, claiming relief, and immediately restoring essentially the same exposure.

A narrow wash-sale rule might deny the loss when the same asset is repurchased within a set period. Modern portfolios make “the same asset” an inadequate test. An investor can replace one broad market fund with a close substitute, use an option, shift the position to a related person, or pair gains and losses across accounts. Rules based on equivalent exposure are harder to explain and enforce, but without them realization rewards the selective harvesting of losses while gains remain deferred.

Then come the events in which ownership changes without an ordinary cash sale. A gift can be treated as a deemed disposal at market value, which collects tax but may leave the donor without cash. A carryover-basis rule can postpone tax and transfer the latent gain to the recipient, which preserves the claim but requires reliable records to follow the asset.

Death presents the same choice with higher stakes. A step-up in basis without tax would allow deferred gains to disappear. Immediate deemed realization can force an estate to find liquidity. Carryover basis, payment by installments, exemptions for limited cases, or combinations of these rules each decide who receives the value of deferral and whether it ever ends.

Transfers between tax boxes cannot be ignored. If an asset leaves Box 3 for Box 2 or Box 1, the law needs either a realization event or a continuity rule that preserves the latent gain and establishes the new basis. The same applies when assets enter Box 3. Otherwise the boundary can erase gain, tax it twice, or make incorporation a route around the tax.

Emigration is the most legally delicate version. A realization system needs some way to protect the Dutch tax claim on appreciation built up while a person was resident. Yet the timing and collection of an exit charge must comply with EU free-movement law. The Court of Justice’s judgments in N v Inspecteur, National Grid Indus, Verder LabTec, and Wächtler concern different taxpayers, assets, and payment arrangements. Together they show that an origin state may protect a taxing claim in some circumstances while immediate collection, deferral, security, and payment conditions remain subject to proportionality. They do not supply a ready-made answer for a future Box 3 portfolio tax. Any concrete rule would therefore require separate specialist EU-law analysis.

Finally, Parliament must decide how long deferral may last. It could accept sale as the only ordinary trigger and rely on death, gifts, migration, and box transfers as terminal events. It could impose a deemed realization after a very long holding period. It could tax certain wrappers periodically. Or it could add a retrospective interest charge intended to remove the financial benefit of waiting.

Alan Auerbach’s retrospective capital-gains proposal demonstrates the last possibility. The tax is calculated at realization, but an interest-like adjustment is used to approximate the burden that would have arisen without the timing advantage. The method avoids observing the exact path of annual gains. It replaces that task with assumptions, retrospective mathematics, and recordkeeping. It is a useful reminder that the choice is not limited to pure annual marking or uncorrected deferral.

These supporting rules affect distribution as much as administration. Carryover basis at death preserves the tax claim; a step-up can turn patient holding into permanent exemption. Generous loss refunds help volatile investors but expose the budget to market downturns. Strict wash-sale rules burden active rebalancing. An interest charge reduces the advantage enjoyed by long-horizon holders but makes the eventual bill harder to predict. The headline timing rule cannot be assessed separately from these choices.

This is realization’s honest price. The central rule is simpler, but its supporting code is not. That does not erase the advantage of a uniform tax event. It changes the standard Parliament must meet. A realization system is credible only if the Netherlands is willing to build prospective broker reporting, choose a transition basis, preserve gains through gifts and box transfers, recognize losses over time, and prevent death or emigration from becoming an unexamined escape hatch.

The awkward Dutch route from here

The permanent case for realization does not answer what the Netherlands should do with bill 36 748. The pending hybrid is already written, passed by the Tweede Kamer, and embedded in an implementation program. A full realization system for listed securities would require a different reporting architecture, especially for acquisition basis and lots. Work completed for annual values cannot simply be relabeled and used for lifetime gain records.

Timing matters because the Belastingdienst does not have spare years hidden in a drawer. Official planning gave the tax authority and financial institutions one year and nine months after Tweede Kamer adoption to prepare for the new system. The implementation letter identified 15 March 2026 as the latest lower-house deadline compatible with a 1 January 2028 start; the bill passed on 12 February. Later changes still require implementation assessment. A substantial switch in the tax base cannot be assumed to fit the same schedule merely because Parliament prefers the new destination.

Nor is the hybrid administratively light. The Belastingdienst’s execution assessment modeled about 876.9 structural full-time equivalents, including roughly 205 valuation specialists, and about €112.14 million in annual structural execution costs. A later parliamentary staffing path peaked at about 1,023 FTE before settling near 877. These are modeled requirements, not staff already recruited. They rebut the comforting fiction that annual market values and prefilled bank data make the proposed system administratively simple.

A realization pivot would exchange some of those burdens for others. It would reduce recurring valuation and classification work for appreciation, but increase basis reporting, transfer records, terminal-event administration, and oversight of loss realization. No execution test yet permits a clean numerical comparison between a mature universal realization system and the pending hybrid. The honest claim is narrower: the hybrid is expensive, and realization would need a new build rather than a cheaper version of the existing one.

The fiscal debate is equally vulnerable to bad labels. In March, the government estimated that continuing the current forfait-plus-counter-evidence regime would leave roughly €2.4 billion less annual revenue than the intended structural budget path. That is not an unconditional €2.4 billion cash cost generated by realization. It is a modeled disadvantage relative to a baseline in which the planned reform supplies more revenue. Change the baseline or the policy parameters and the comparison changes.

The much larger figure in the current political debate needs even more care. Disclosed Finance Ministry decision notes included an illustrative path that keeps the present regime through 2031 and introduces realization for property and securities from 2032. Against the pending hybrid, the annual differences shown for 2028 through 2036 sum to about €21.952 billion less revenue. The calculation assumes that a securities portfolio is sold on average once every 4.5 years, which materially affects how quickly realization revenue arrives.

According to the same notes, the realization estimate is preliminary, omits some asset categories, is sensitive to behavioral assumptions, and requires further refinement. It was not certified by CPB. It is a cumulative transition-path difference for one scenario, not an annual bill and not a permanent revenue loss from realization. In the same exercise, the modeled mature scenario produces about €30 million less annual revenue than the hybrid.

The contrast is the point. A change in timing can create a very large financing gap while one system grows in and a much smaller difference after the stock of deferred gains matures. Governments borrow, tax, and spend in actual years, so the transition gap cannot be waved away as accounting. It also cannot determine the permanent architecture by itself. Otherwise every tax system already under construction would become optimal by virtue of having a sunk cost and an earlier revenue stream.

CPB’s certification of the pending hybrid shows how unstable even the official path can be. The certified ministry estimate, measured against the pre-Kerstarrest forfaitary baseline with specified structural adjustments, produces €1.006 billion more revenue in 2028, €167 million less in 2031, €279 million more in 2036, and €1.897 billion more structurally from 2060. CPB judged the estimate reasonable and neutral, but highly uncertain. That structural uplift is not a clean measure of accrual versus realization timing: CPB says it is driven mainly by a broader tax base, especially the inclusion of rental income and changes to the netting of assets and debts.

Those numbers are not a score for a future full realization tax. They also rest on conventions that matter. Household wealth is held at its estimated 2028 level in later years so the comparison isolates the policy change. The model applies a 20 percent behavioral adjustment to taxpayers whose average burden rises. For Box 3 property, it assumes the frequency of sales falls by 50 percent under realization, while CPB says the size of that lock-in response is uncertain. The “structural 2060” figure is therefore a model comparison on a frozen wealth stock, not a prediction of the euros the Treasury will collect in 2060.

This leaves the government with an awkward but manageable obligation. If it chooses realization as the permanent destination, it must publish the basis rules, terminal events, loss system, anti-deferral method, implementation test, and a revenue-neutral mature package. It must also explain how the years before that system begins will be financed. That may justify retaining part of the existing bill as a bridge, redesigning it, or accepting delay. The evidence available now does not select one transition route without a concrete novelle, execution assessment, and fiscal package.

What it does rule out is pretending that implementation and destination are the same question. Progress on a hybrid does not prove the hybrid should govern for decades. Preference for realization does not make the current stopgap sustainable, recover the foregone transition revenue, or conjure a 2028 basis-reporting system into existence.

What Box 3 should tax

When an investment pays interest, distributes a dividend, or produces rent, Box 3 should tax that cash return when it is received. When the investment itself rises in value, realization should probably be the general permanent rule: tax the appreciation when the asset is sold or when another carefully defined event brings the gain into account.

That judgment concedes something important. Annual accrual is the cleaner measure of nominal yearly income for ordinary listed investments. It can use prices and third-party data that already exist, avoid decades of basis records, remove voluntary gain timing, and reduce realization-driven loss harvesting. The Netherlands has sound reasons for having spent years trying to build it.

The gain in annual accuracy is not free, however. A hybrid assigns different tax clocks to categories that financial markets and legal structures can partly substitute for one another. It invites disputes over wrappers, debt, derivatives, private vehicles, and the border with Box 2. It asks taxpayers to accept that economically similar appreciation is current income in one form and deferred gain in another. It requires the Belastingdienst to operate both methods and police the boundary indefinitely.

A general realization rule gives up precision where accrual is strongest in order to gain coherence across the system. It normally uses an observable transaction price and a cash event. It narrows a prominent timing difference with Box 2 and is less likely to make the tax year turn on how an investment has been packaged. Those are not decorative virtues. They reduce one source of classification-driven planning and make the organizing principle easier to explain. Political durability can never be guaranteed, but clarity gives a tax a better chance of surviving changes in coalition and fashion.

Realization earns that preference only with the unglamorous surrounding rules. Basis must follow assets. Losses must receive credible multi-year treatment. Selling a loss and instantly recreating the position cannot become a routine deduction machine. Gifts, inheritance, death, transfers between boxes, and migration must preserve or collect latent tax. Very long deferral may require an interest charge, a terminal event, or another limit. Parliament cannot promise simplicity at sale and leave every difficult sale-adjacent event for later.

The immediate transition remains a separate judgment. The pending hybrid may still be the fastest route out of the legally unstable stopgap, or it may be too compromised to justify the investment still required. That cannot be resolved by a leaked political direction, a cumulative scenario stripped of its baseline, or the fact that the existing project has already consumed years of work. It requires a published proposal and a fresh implementation and fiscal test.

The Netherlands should choose the permanent rule first and confront the transition honestly. Tax cash returns when they arrive, and tax appreciation when a sale or defined event crystallizes it. Then write the rest of the code as if that promise is meant to last.