A group with operating companies in five African countries eventually arrives at the same request. It wants one set of consolidated numbers, one treasury that can move cash where it is needed, and one legal entity that a bank, an auditor or a buyer can look at without having to understand five company registries first. The usual answer is a European holding company, and in our practice that usually means a Dutch one.
The holding is rarely the problem. What the holding starts doing is the problem. A company that owns shares and collects dividends is a shareholder. A company that runs the group’s treasury, negotiates its contracts and directs its subsidiaries is an operating business, and it can become taxable in places nobody intended. The difference between those two companies is not visible on an organization chart. It is visible in where people sit, what they decide, and what the intercompany agreements say they do.
The two jobs the holding is asked to do, and why they pull apart
Consolidated reporting and centralized treasury sound like one project. They are two, and they carry opposite risk profiles.
Reporting is passive. It requires information to flow upward, on a common calendar, in a common format. It does not require the holding to act in the subsidiary’s country, and on its own it creates almost no exposure. The obstacles are practical: five different local audit calendars, five different sets of statutory accounts, and finance teams that were never asked to close in twenty days.
Treasury is active. Sweeping cash, lending between companies, guaranteeing local facilities and hedging currency are all transactions, and each of them is priced, taxed and sometimes restricted in the country where the counterparty sits. A group that centralizes treasury before it has centralized reporting is doing the risky half first, which is the wrong order and the common one.
Consolidation is an accounting question before it is a tax question
The most frequent confusion we correct is between accounting consolidation and tax consolidation. They are unrelated, and only one of them is available.
A Dutch fiscal unity under article 15 of the Wet op de vennootschapsbelasting 1969 requires the parent to hold the full legal and economic ownership of at least 95 per cent of the shares in the nominally paid up share capital of the subsidiary, representing at least 95 per cent of the statutory voting rights and giving entitlement in all cases to at least 95 per cent of the profit and at least 95 per cent of the assets. Paragraph 4 of the same article adds that both companies must be established in the Netherlands. African operating subsidiaries are not, so they cannot join. Their profits are taxed where they arise, and no Dutch grouping election changes that.
What the holding delivers instead is a consolidated set of financial statements, which is a reporting exercise governed by accounting standards rather than by tax law. That is worth doing for its own reasons. It is what allows the group to be financed as a group, valued as a group and sold as a group. It is not a way of offsetting a loss in one country against a profit in another, and any plan that quietly assumes otherwise will produce a tax charge nobody budgeted.
Where the permanent establishment actually arises
Holding shares in a foreign company does not create a permanent establishment in that company’s country. Almost everything else the group is tempted to centralize can.
The exposure comes from people and premises. A regional finance director employed by the Dutch holding who spends most of the year in one African country, working from the local subsidiary’s offices, gives the tax administration of that country a fixed place through which the holding carries on business. A group commercial director employed by the holding who habitually negotiates and concludes contracts in a country creates the agency version of the same problem. Neither of them appears in the corporate structure, and both are usually described internally as a group function rather than as a taxable presence.
The mirror risk runs in the other direction and is easier to miss. If the real decisions of the Dutch holding are taken in Africa, by people who live there, the holding’s own residence comes under pressure. Under article 4 of the Algemene wet inzake rijksbelastingen, where a body is established is judged according to the circumstances and not by the register. A company can therefore be effectively managed somewhere other than where its board meets on paper, and once that is argued the whole structure is being defended from behind.
Treasury, and the moment the holding becomes a lender
Once the holding starts lending to subsidiaries, three questions arrive at the same time in each of the five countries. Whether the interest is deductible locally, subject to whatever interest limitation and thin capitalization rules apply there. Whether withholding tax is due on the interest paid out, and at what rate after any treaty. And whether the rate charged would survive a transfer pricing review on both sides.
Guarantees raise the same issue in a quieter form. A holding that guarantees a local facility is providing something of value, and several administrations will look for a fee. Charging nothing is a position that has to be defended, not a way of avoiding the question.
On the Dutch side, the item most often overlooked sits in the Wet bronbelasting 2021. Interest, royalties and, since 1 January 2024, dividends paid from the Netherlands to an affiliated entity in a low taxed jurisdiction are all subject to a conditional withholding tax. Article 3.1 of that law charges the three together, the dividend added last through article 3.4a, and that is the flow groups still leave off the map. The rate is the same for the three: article 4.1 fixes it at the highest percentage of article 22 of the Wet op de vennootschapsbelasting 1969, which for 2026 is 25.8 per cent, the upper of its two bands. Whether a jurisdiction is low taxed is not answered from its tax rate. Under article 1.2 of the Wet bronbelasting 2021 a state qualifies only once it has been designated by ministerial regulation, on one of two grounds: that it does not subject bodies to a profit tax or does so at a rate below 9 per cent, or that it appears on the EU list of non-cooperative jurisdictions for tax purposes. The instrument that operates is therefore the Dutch Regeling laagbelastende staten en niet-coöperatieve rechtsgebieden voor belastingdoeleinden, which names the states. The EU list feeds that designation and does not replace it; reading one as the other gives the wrong answer. The designation is not the whole perimeter. Article 2.1 of that law opens with five cases, and only the first turns on the recipient sitting in a designated jurisdiction. Another reaches a recipient established in a state that is not designated at all, where the recipient is entitled to the payment with the avoidance of another’s tax as one of its main purposes and the arrangement is artificial to the extent it lacks valid commercial reasons reflecting economic reality. The other three cover a permanent establishment in a designated jurisdiction and two hybrid cases. Interposing a vehicle tends to land in the artificiality case, not in the list. A group with an offshore vehicle in its history should therefore read all of article 2.1 before routing a dividend or a loan through the vehicle, not the list alone.
Currency is the constraint the tax plan usually ignores
A centralized treasury assumes that cash in the group is fungible. In several African markets it is not. Local rules on the availability of foreign currency, on the documentation required before a bank will execute a remittance, and on the order in which claims on scarce currency are met can delay a dividend or an interest payment for months.
This matters more than most tax variables, and it is a different kind of problem. A tax cost is a number that can be modeled. A repatriation delay is a timing risk that breaks covenants, distorts the group’s working capital and makes the consolidated cash position a fiction. We have seen structures that were elegant on the tax side and unworkable in practice, because the model netted a subsidiary’s surplus against another’s deficit in a currency that could not actually cross the border.
The practical response is to build the treasury around what can actually move, keep genuine local buffers in the countries with restrictions, and treat trapped cash as trapped in the model rather than as available. It is less satisfying than a single pool. It is also the only version that survives contact with a bank.
Substance, and what the Dutch company has to be
The Dutch holding earns its position by being a company. Under article 13 of the Wet op de vennootschapsbelasting 1969 the participation exemption applies where the taxpayer is a shareholder in at least 5 per cent of the nominally paid up capital of the subsidiary and the further conditions of that article are met, so the shareholding threshold is not the hard part. The hard part is everything the exemption assumes about the holder.
That means a board that meets in the Netherlands on matters that are genuinely open, with enough information in front of it to decide them. It means the holding having the means to bear the risks it reports, which for a treasury company means real equity rather than a nominal balance sheet. It means agreements that describe the group functions actually performed, and a transfer pricing file that would read the same way to a Dutch inspector and to an African one, because eventually both will read it.
What we would build for a group in five countries
Sequence the work. Put the holding in place and get the reporting consolidation running first, because it creates almost no exposure and it is what unlocks group financing. Leave treasury centralization to a second phase, once the group knows which of the five countries can actually remit and on what documentation.
Map the flows one country at a time rather than as a diagram. Dividend, interest, service fee and royalty are four different flows with four different withholding treatments in each jurisdiction, and the applicable treaty position for each of them has to be checked against the current text rather than assumed from the group’s other markets. On a dividend leaving the Netherlands the order runs the other way. First the ordinary dividend withholding tax, which article 5 of the Wet op de dividendbelasting 1965 sets at 15 per cent of the proceeds and which the withholding exemption of article 4 lifts only for a shareholder meeting its conditions. Then the conditional charge, where the recipient falls within it, which takes the total up to its own higher rate because article 5.2 of the Wet bronbelasting 2021 credits the dividend withholding tax against it. The treaty comes last, and it imposes nothing: it only limits what Dutch law has charged.
Then decide, deliberately, where each group employee is employed and from where each group function is performed. That single decision drives most of the permanent establishment exposure, and it is the one most often made by default when somebody relocates.
Where a single holding stops being the right answer
One holding above five countries is the right shape when the five assets are held for the same reasons, on the same horizon, by the same shareholders. It stops being the right shape as soon as those conditions break.
A local partner in one country, an asset the group intends to sell separately, a jurisdiction whose rules make a local intermediate company necessary for regulatory reasons: each of these argues for a sub-holding, and each of them is cheaper to build at the outset than to insert later, once the shares have appreciated and a transfer would trigger a charge. The question to ask before drawing the chart is not how few entities the group can have. It is which assets will one day leave separately, and whether the structure as drawn would let them.
That question has an uncomfortable corollary. A group that cannot answer it has not decided what it is building, and a holding structure erected on top of an undecided group tends to be redrawn within three years, usually at the worst possible moment, which is when a buyer has already been found and the timetable belongs to somebody else.
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.