The shape is familiar from every European property market where foreign capital buys buildings. Put the building inside a local company, put that company under a Dutch holding, and plan the eventual exit as a sale of shares rather than a sale of bricks. In Spain the arrangement is common enough that clients often arrive with it already drawn.
Applied to Belgian property, it usually costs more than it saves. Not because Belgium has an anti-avoidance rule aimed at it, and not because the Dutch holding fails at anything it is supposed to do, but because the variables the formula depends on are answered differently in Belgium than in the markets where the shape was learned. Most answers are worse. One is better than people expect, and not enough on its own.
The formula, and the variables it depends on
Stripped of jurisdiction, the arrangement is a bet on four propositions. First, that the situs state taxes rental income more heavily in an individual’s hands than in a company’s. Second, that the local taxable base is computed on rent actually received rather than on something notional. Third, that a sale of shares escapes a transfer charge that a sale of the building would attract. Fourth, that the eventual gain is taxed more lightly inside a corporate chain than outside one.
Where all four hold, the wrapper pays for itself. Where two hold, it is arguable and turns on the holding period. Where one holds, the structure is a cost centre with a good story attached. Belgium leaves the promoter with one, and not the one that carries most weight.
Where the taxing right sits, and where it does not
Begin with the treaty, because everything else is downstream of it. The convention in force between the Netherlands and Belgium is the one signed at Luxembourg on 5 June 2001, in the consolidated version that has applied since 1 September 2013. A newer treaty was signed at Brussels on 21 June 2023, and its text carries the note that it enters into force on a date still to be determined, so it is not the instrument to plan against.
The consolidated text is not the whole of the law, which is where descriptions of this arrangement most often go wrong. The 2001 convention is a covered agreement under the multilateral BEPS convention: Belgium listed it on depositing its instrument of ratification, and the Netherlands added it to its own list on 25 November 2021. Neither state reserved against article 7 of that convention, so a principal purpose test now sits above the treaty: a benefit is refused where it is reasonable to conclude that obtaining it was one of the principal purposes of an arrangement, unless granting it in the circumstances would accord with the object and purpose of the provisions relied on. That test governs every treaty benefit described below.
Article 6 gives the situs state the right to tax income from immovable property situated there. Income that a resident of the Netherlands derives from immovable property in Belgium may be taxed in Belgium, and the article reaches direct use, letting and any other form of exploitation. Article 13 does the same for gains: profits from the disposal of immovable property as defined in article 6, situated in the other state, may be taxed in that other state.
The building is Belgian for tax purposes no matter who owns it or through how many layers. A Dutch entity does not make Belgian rent Dutch, and the Netherlands relieves rather than competes for it: for income Belgium may tax under articles 6 and 13, article 23 requires the Netherlands to exempt by reducing its own tax. Anyone presenting the arrangement as a way of moving Belgian property income into a lighter Dutch charge is describing something the treaty does not permit.
The base Belgium taxes: a 1975 rent, indexed and increased
This is the variable that surprises people, and it runs opposite to the formula.
Belgian property carries a cadastral income, which the income tax code defines as the average normal net income of one year and fixes by reference to normal net rental values at a statutory reference date. The last general revaluation was carried out under the law of 19 July 1979, applied from assessment year 1980, and rested on net rental values at 1 January 1975. Nothing has replaced it since. Because 1975 is a long time ago, the figure is indexed annually under article 518 of the code, and the coefficient is published in the Moniteur belge: it is 2.3000 for income year 2026, against 2.2446 for income year 2025.
Cadastral income is also the base of the annual property tax, which is a regional charge rather than a federal one and is levied at rates that differ between Flanders, Wallonia and Brussels, before the municipal and provincial surcharges that multiply it. The same figure drives what an individual owner declares as property income.
The consequence is the one the formula does not expect. Where a building is let to an individual who does not use it for business, the owner is taxed not on the rent collected but on the cadastral income increased by 40 per cent, under article 7 of the code as replaced in 2021. Indexed and then increased, a 1975 net rental figure arrives at a little over three times its original amount, which on a well-let building can still sit below the rent actually received.
A company does not get this. Where the owner is a company, the building’s income is business profit and the base is what came in. Within the personal income tax the same switch happens where the building is let to a tenant who uses it for business, or to a company, though not where that company is a regional or approved social housing company passing the property on to individuals purely as housing. Outside that case the taxable amount is the rent and rental benefits actually received, subject to a floor at the same increased cadastral figure. The wrapper trades a favourable notional base for an unfavourable real one, the reverse of the trade the formula assumes.
What changes when a company owns instead of a person
The second reversal concerns the gain rather than the income.
An individual holding Belgian property within the normal management of private wealth stands differently from a company, and the Belgian treatment of a private gain in those circumstances is not the corporate treatment. A company’s gain on a building is business profit. Belgian corporate income tax is charged at 25 per cent, with a reduced rate of 20 per cent on the first tranche of 0 to 100,000 euro for companies that count as small. That reduced rate is withheld from a company which does not pay at least one of its directors a minimum remuneration, and from several other categories besides, so it should never be assumed.
The treaty leaves a gain on Belgian immovable property with Belgium whether the owner is Belgian or foreign, so interposing a Dutch entity does not move the gain. What it changes is the character of the owner, from one Belgian tax treats gently into one it treats as trading.
Then there is the second layer, which the individual owner never meets. Money inside a company has to come out. A distribution by a Belgian company meets the withholding tax on movable income at a basic rate of 30 per cent unless a relief applies, and a distribution by a Dutch holding meets Dutch dividend withholding tax at 15 per cent unless an exemption applies. Neither is fatal alone. Both are additional to a charge the individual owner would never have met.
The exit, and the clause the treaty does not contain
Here the formula gets its one good answer, and it is worth stating accurately, because it is usually stated wrongly in both directions.
Article 13 of the 2001 treaty allocates gains on immovable property to the situs state and leaves gains on everything else to the state in which the alienator is resident. It contains no clause allowing the situs state to tax gains on shares deriving their value from immovable property. Many modern treaties carry such a clause. This one does not, and the paragraph that does deal with shareholdings is an emigration provision: it reaches an individual who was resident in the first state, and only for as long as an assessment raised on the deemed disposal at departure remains outstanding.
The multilateral convention does not supply the missing clause either, and this is worth checking rather than assuming, because the consolidated treaty text is silent about it. Article 9 of that convention would add a land-rich rule with a 365 day look-back, but it reaches a treaty only where both states have notified a choice to apply it, and neither the Netherlands nor Belgium has done so; Belgium reserved in addition against the paragraph that would attach the look-back to a clause already in place. Article 13 is untouched on this point.
So a sale of the shares in a Belgian property company by a Dutch holding is not delivered to Belgium by the treaty in the way a sale of the building would be. On the Dutch side the deelnemingsvrijstelling can exempt the gain. It asks for 5 per cent of the nominal paid-up capital and sets no minimum holding period, and the tests that withhold it from passive holdings are easier for property to pass than is generally assumed: immovable property, and rights relating directly or indirectly to it, are excluded from the free investments that can make a participation low-taxed, provided the property is not held through a body designated as an investment institution or an exempt investment institution.
One route out is closed off. Under the protocol to the treaty, what a shareholder receives on a buy-back of its shares or on a liquidation is not a gain from alienation at all: it falls under article 10 and is treated as a dividend. The choice of exit route is therefore not neutral.
This is a genuine advantage, and smaller than it looks. A share sale transfers the company with its latent Belgian gain intact and with no step-up in the building’s tax base, so a rational buyer prices that latent liability into the offer. The seller does not escape the tax. The seller receives a price reduced by a discounted version of it, and takes on the work of persuading a buyer to accept a wrapped asset, which Belgian domestic buyers frequently decline.
What the Dutch side contributes, and what it does not
It is worth separating what the Dutch holding does from what it is often said to do.
It consolidates. A Dutch holding over property companies in several countries gives one place where dividends land, where the participation exemption relieves them on that same 5 per cent test and with no holding period, and where reinvestment is decided. Dutch corporate income tax is charged as a bracket rather than a flat rate: 19 per cent on the first 200,000 euro of the taxable amount, and 38,000 euro plus 25.8 per cent above that. It reaches what the holding itself earns, which in a well-built structure is very little, because the operating returns are relieved.
It does not shelter Belgian rent, it does not shelter the Belgian gain on the building, and it does not turn a Belgian property company into a Dutch asset for Belgian purposes. Substance is required throughout: a board that meets and decides in the Netherlands, minutes recording deliberation rather than ratification, accounts filed on time, and transfer pricing documentation matched to the functions performed.
The cost of entry
The last variable is acquisition cost, and it often settles the question before the tax analysis is reached.
Belgian registration duty on the acquisition of immovable property is a regional charge, set separately by Flanders, Brussels and Wallonia, and it is substantial. The Flemish sale duty is 12 per cent; Wallonia and Brussels each set their general rate under article 44 of the registration duties code as it stands in their region. The relief for an owner-occupier is not a single pattern and is never automatic: Flanders drops the rate to 2 per cent for a buyer’s own and only home who owns nothing else and actually moves in, Wallonia reduces it to 3 per cent on comparable conditions, and Brussels leaves the rate alone and cuts the taxable base instead. Nor are the reduced rates confined to owner-occupiers: the Flemish codex alone also sets 6 per cent for professional property traders, 10 per cent on agricultural land and 1.5 per cent for approved social housing bodies, and the other two regions have their own lower rates for professional traders and social housing.
A purchase of shares does not attract the duty, which is the argument for buying the company rather than the building, and for a buyer it is a real one. But it is an argument about the second transaction, not the first. The wrapper only avoids duty on exit if it existed on entry, and if the property went into a company at acquisition the duty was paid then. Structures are frequently sold on the strength of a saving available only to a future buyer, funded by costs the present client pays now, across a holding period nobody has committed to.
Where a Dutch entity does earn its place
None of this makes the Dutch entity wrong. It means the reason for it has to be something other than the Belgian building.
It earns its place where there is a portfolio rather than an asset, particularly where the assets sit in more than one country. It earns its place where several investors need a single vehicle with a governing law they can agree on and a share register they can rely on. It earns its place where a family needs ownership to survive a succession without reopening title to a building in every jurisdiction. Those are structuring reasons, and they hold whether or not the Belgian tax arithmetic cooperates.
In one mandate we were asked to put a single Belgian commercial building under a Dutch holding for a client who had used the same shape in another market. The modelling took an afternoon. Between the loss of the notional base, the corporate characterization of a gain that would otherwise have sat in private hands, and the second layer needed to get cash out, the wrapper cost more across any realistic holding period than it could recover on exit. The client bought that building personally and put the next three, in three countries, under the holding.
The formula is not wrong. It is local. It was learned in markets where the local company is taxed more gently than the local individual, and Belgium is not one of those markets. Where the tenant is an individual who does not use the building for business, Belgium taxes the owner on a notional figure anchored to 1975, indexed and then increased by 40 per cent; where the owner is a company, or the tenant uses the building for business, it taxes what actually came in. Any structure that begins by moving a Belgian building from the first category into the second has to earn back that difference before it delivers anything at all, and a single building rarely does.
This article is informational and does not constitute tax or legal advice. The treatment of any structure depends on its facts and on the law of each jurisdiction involved. Each engagement is subject to scope and applicable regulation.