Entity Classification: Persons, Instruments and Income Across Borders

Concept

Entity classification answers a question that sits underneath almost every cross-border tax result: when income arises inside a legal structure, who is the taxpayer, when does the tax point occur, in which country, and what kind of income is it? Company law answers none of this. A certificate of formation tells you which state's corporate statute governs the entity; it does not tell you whether Germany, the UK or the US will tax the entity itself or look straight through it to its owners. Each country answers that question with its own tax rules, from its own side of the border — which means one and the same entity is routinely a taxpayer in one country's eyes and a transparent conduit in another's, at the same time.

One example carries the whole subject. A Delaware limited partnership earns a portfolio gain. For the US, 26 U.S.C. §701 says a partnership "as such" pays no income tax — the partners are liable "in their separate or individual capacities," and under §706(a) each partner picks up its allocated share for the year, distribution or not. A UK-resident partner reaches the same practical result by a different route: HMRC's practice list treats a US limited partnership as transparent, so the partner is taxable on the share of profits as they arise (INTM180030). A German corporate investor cannot copy either answer. German law runs its own Rechtstypenvergleich — a two-sided comparison of the foreign entity's legal features against domestic entity types (BFH, I B 75/20, 18 May 2021) — and only if the LP resembles a German Kommanditgesellschaft does the investor become a co-entrepreneur taxed on the share as it arises under §15(1) no. 2 EStG. Three observers, three independent tests, one entity.

Three features of this field are routinely misjudged. First, classification is not a property of the entity — it is a property of the (entity, observer-country) pair, so a structure with investors in three countries has at least three classification answers running in parallel, plus the source country's own. Second, the entity question is only one of three: the instrument (equity or debt) and the income item (dividend, interest, capital gain, business profit) are classified separately, by each country, and each answer moves real money — a deduction here, a withholding rate there, a participation exemption somewhere else. Third, the answers can be legitimately different — and the mismatch itself is now a regulated object: since BEPS Action 2 and the EU's ATAD 2 directive, a deduction in one state without inclusion in the other is not a planning win but a trigger for corrective rules that deny the deduction or force the inclusion.

This page owns the classification logic itself. How the classification of a whole fund chain plays out is mapped in fund tax architecture; how it interacts with holding-company tests in holding structures; what happens when an opaque entity is then attributed to its controller in the CFC master guide.

Three objects of classification

Cross-border tax analysis classifies three different things, and conflating them is the most common structural error in investor memos. The table below separates the three objects; in each row the classifying observer is each country touching the income — source country, entity country and each investor's country ask the question independently.

The three layers stack: an observer first classifies the entity (deciding whose income it is), then the instrument (deciding what kind of return it produces), then the item (deciding rate, source and year). A single cross-border payment therefore passes through at least six classification decisions — three on each side of the border — before its net tax cost is knowable.

Trusts are deliberately outside this page: they are classified under separate rules (the US business-entity regulation itself excludes trusts, 26 CFR 301.7701-2(a)), and their tax treatment is mapped in trust taxation.

One entity, four national methods

There is no international classification authority. Each country has built its own machinery, and four reference designs — election, factor comparison, type comparison and codified comparison — cover most of what an investor will meet. The matrix below classifies the same object, a foreign business entity, from the standpoint of four named observers: the United States, the United Kingdom, Germany and the Netherlands.

The design lesson from the matrix: the US made classification largely elective, the UK made it judicial-then-administrative, Germany made it comparative, the Netherlands made it statutory — and none of the four defers to the entity's home-country label. The US election binds only the US; HMRC's list binds only UK analysis; the BFH's comparison binds only Germany; the Dutch method binds only the Netherlands. Each answer then feeds that observer's own follow-on machinery: an entity the US classifies as a corporation becomes the object its CFC rules test for control and attribution, and the UK CFC regime works the same way off the UK answer. Classification first, attribution second — never the reverse.

Case note: Anson v HMRC (2015)

Facts. Mr Anson, UK-resident and non-domiciled, was a member of HarbourVest Partners LLC, a Delaware limited liability company carrying on an investment business. The LLC's profits were taxed in the United States in the members' hands. The UK then taxed Mr Anson on the sums he remitted, and he claimed double taxation relief under article 23(2)(a) of the 1975 UK/US convention and section 790 ICTA 1988 (the relief provision then in force, since rewritten into TIOPA 2010). HMRC refused: on its view the LLC was an opaque company, so what reached Mr Anson was a distribution of the LLC's profits, not the income the US had taxed.

Issue. Narrower than "is an LLC a company?". The question was whether the UK tax was computed by reference to the same profits or income as the US tax — identify the income by reference to which the UK liability is computed, compare it with what was taxed in the US, and decide whether they are the same ([2015] UKSC 44, para 113).

Holding. The appeal was allowed on 1 July 2015. The First-tier Tribunal had found, on Delaware law and on the LLC agreement, that the members were entitled to the profits as they arose — the agreement allocated the profit to the members as it arose and required payment to be made (para 18) — so Mr Anson's UK income was his share of the profits of the business carried on by the LLC, the same income already taxed in the US (paras 120–121). Relief was due.

Limits. Two, and they matter more than the outcome. The finding was about these documents under this governing law, reached by a fact-finding tribunal on evidence of foreign law — not a general re-classification of LLCs. And HMRC's published response was to treat the decision as "specific to the facts and findings determined by the First-Tier Tribunal on the interpretation of foreign law and relevant LLC agreement" and to continue its existing practice of treating US LLCs as companies (Brief 15/2015).

What it changes. It turns an assumed classification into an evidential one for the taxpayer willing to prove it: what the operating agreement and the governing law actually say about entitlement to profits can decide the relief question even where the authority's standing practice points the other way. It does not hand the next LLC member the answer — it hands them the burden of establishing it.

Elections: where classification can be chosen

Two of the four reference systems let a taxpayer choose — within limits — and the mechanics matter because elections have windows, lock-ins and deemed-transaction costs.

Under the US rules an eligible entity files Form 8832. The election cannot be made effective more than 75 days before filing or more than 12 months after it, and after an election that changes a classification the entity generally cannot elect again for 60 months (301.7701-3(c)). A change is not a paperwork event: electing a partnership into corporate status, or the reverse, is treated as a deemed liquidation or incorporation with real gain-recognition consequences — which is why the election is usually made once, at formation, when nothing has appreciated yet.

Germany's §1a KStG is the continental counterpart: on an irrevocable application, filed electronically no later than one month before the start of the relevant fiscal year, a commercial partnership is taxed like a corporation and its partners like non-personally-liable shareholders, with a return option (Rückoption) back to transparency. The entity remains a partnership under company law throughout — the option changes only the tax lens.

The UK has no election at all: a foreign entity is what the factor analysis says it is, and the practical instrument of certainty is HMRC's published list plus, where the stakes justify it, entity-specific analysis of the governing documents.

Elections also get missed, and the systems differ on whether that can be repaired. The US keeps a standing relief procedure: an eligible entity that failed to file on time can still obtain the classification it intended by filing Form 8832 within 3 years and 75 days of the requested effective date, provided it has reasonable cause for the failure and either has not yet filed returns for the first year or has filed all of them consistently with the classification requested (Rev. Proc. 2009-41). Outside that window the only instrument left is a change election — a deemed liquidation or incorporation, taxed as such, which also starts the 60-month clock. Germany runs no equivalent amnesty, but a taxpayer who needs certainty before acting can buy it: a binding advance ruling on a precisely described, not-yet-implemented set of facts, where a special interest exists in view of the substantial tax consequences, under §89(2) AO, for a fee charged on the value at stake (§89(3), (5)). The UK, having no election to miss, deals in evidence instead — the published list, the entity's own constitutional documents and, where the amount justifies it, the kind of foreign-law analysis a tribunal accepted in Anson.

An election made in one capital changes nothing in another. A US check-the-box election turning a Luxembourg SARL into a US-disregarded entity leaves Luxembourg taxing the SARL as a resident company exactly as before — that asymmetry is deliberate, and it is precisely the raw material of the hybrid structures discussed below.

Transparent vs opaque: who pays, when, and on what

Classification is shorthand for a bundle of operational consequences. HMRC's own definition is the cleanest statement of the dividing line: in a transparent entity the members are liable to tax "on the profits, income or gains of the entity as they arise"; in an opaque one, "only on the distributions" (INTM180010). The table compares the two regimes for one object — an investment entity with profits — as seen by any single observer country applying its own classification.

Read as a whole, the table explains why fund structuring spends so much effort on classification before anything else: transparency buys one layer of tax and preserved character at the price of dry income — tax before cash — while opacity buys deferral and simplicity at the price of a second tax layer and character conversion. The fund tax architecture page walks this trade through a full fund chain, including the blocker entities whose entire purpose is to interpose an opaque layer for investors who need one.

Timing deserves one more sentence, because it is the least intuitive consequence. In a transparent regime the year of taxation is fixed by allocation, not payment: a US partner includes partnership items for the partnership year ending within its own year (§706(a)); a German Mitunternehmer includes the profit share the year it arises, and even remuneration the partner charges the partnership — interest on a partner loan, fees for services — is pulled back into business income under §15(1) no. 2 EStG's Sondervergütungen rule. Cash can follow years later; the tax does not wait for it.

Who claims the treaty, and on what income

The "treaty claimant" line in the table is where classification stops being theoretical. A treaty applies to "persons who are residents of one or both of the Contracting States" (Article 1(1), 2017 OECD Model), and a transparent entity is usually not a resident of anywhere — so something has to say whose income the payment is. Article 1(2), added in 2017 on the recommendation of BEPS Action 2, says it: income derived by or through an entity or arrangement treated as wholly or partly fiscally transparent under the tax law of either state "shall be considered to be income of a resident of a Contracting State but only to the extent that the income is treated, for purposes of taxation by that State, as the income of a resident of that State."

Two consequences follow mechanically, and together they are the operative rule for a cross-border investor. Where the investor's residence state treats the entity as transparent, the investor derives the income and claims under its own treaty with the source state, at its own rate. Where the investor's state treats the same entity as opaque, the investor does not derive that income at all for this purpose, and the claim fails unless the entity itself qualifies as a treaty resident. Older treaties acquire the provision through Article 3 of the multilateral instrument where both parties adopted it; a treaty that predates the 2017 update and was not modified may contain nothing on the point, and that silence is itself a fact to plan around.

The United States legislated the mirror image. 26 U.S.C. §894(c) denies a foreign person any reduced treaty withholding rate on an item of income derived through an entity that is fiscally transparent for US purposes where three conditions hold together: the item is not treated as that person's income under the tax law of its country; the treaty contains no provision addressing income derived through a partnership; and that country does not tax a distribution of the item from the entity. The claim is then a documentary act at source: a foreign flow-through entity gives the withholding agent Form W-8IMY — "Certificate of Foreign Intermediary, Foreign Flow-Through Entity, or Certain U.S. Branches for United States Tax Withholding and Reporting" — and passes through the members' own certificates, because it is the members' residence, not the vehicle's, that the agent has to test.

Two boundaries close this off. Classification decides who derives the income; it does not decide where that person is resident, and an opaque entity still has to satisfy the corporate tax residence tests of the state whose treaty it wants to use. And the article and rate a successful claimant then gets are the subject of withholding tax.

Instruments: debt, equity and the character of the return

The second classification layer takes the instrument the investor actually holds. The stakes are symmetrical: for the payer, debt means a deductible interest expense and equity means a non-deductible distribution; for the recipient, the label selects the treaty article, the withholding rate, and the availability of participation exemptions that typically attach to dividends but not to interest.

Countries do not accept the contractual label here either. The UK statute is a worked example of re-characterisation in primary legislation: interest on "special securities" is treated as a distribution — not interest — where, among other conditions, "the consideration given by the company for the use of the principal secured depends (to any extent) on the results of the company's business" (CTA 2010 ss 1000(1)F, 1015). A profit-participating loan into a UK company is, to the UK observer, equity in substance: the return is non-deductible and dividend-like, whatever the loan agreement says. Other systems draw the line with different tools — case-law debt-equity factors in the US, arm's-length and hidden-distribution doctrines in Germany — but the structural point is constant: each country classifies the instrument for itself, on substance.

Because the tests differ, one instrument can be debt for one observer and equity for another. Whole instrument classes have been built in that gap — the Luxembourg-to-US structures using convertible preferred equity certificates (CPECs), treated in practice as debt in Luxembourg and equity for US purposes, are the best-known market example (a practice observation, not a rule stated in either statute). That gap is no longer free money, which is the next section's subject.

Hybrid mismatches: when two answers collide

A hybrid mismatch is the formal name for two observers disagreeing profitably. The EU's ATAD 2 directive defines the target outcomes in Article 2(9): deduction without inclusion (D/NI) — the payer's state allows a deduction while the payee's state sees no taxable income, because the two characterise the instrument or the entity differently — and double deduction (DD), one expense deducted twice. Both arise mechanically from everything above: a payment on an instrument that is debt for the payer and equity for the payee produces D/NI; an entity opaque at home and transparent to its owner produces payments that vanish between the systems; a reverse hybrid — transparent at home, opaque to its owners — produces income no one includes.

The response, designed in BEPS Action 2 (2015) and made mandatory across the EU by ATAD 2, is a pair of linking rules: the primary rule denies the deduction in the payer's member state; if the payer's state does not act, the defensive rule forces the payee's state to include the payment in income (Article 9). Reverse hybrids get their own rule — Article 9a treats the entity as resident and taxes it directly where non-resident owners holding 50% or more see it as opaque — applicable since 1 January 2022, the rest of the regime since 1 January 2020. The framework reaches hybrid instruments, hybrid entities, hybrid permanent establishments and imported mismatches routed through third countries.

Classification is not reporting transparency

One terminological trap needs cutting before the scenarios. "Transparency" in this page means liability: whose income it is and when. The same word is used in a second, unrelated sense — tax transparency as automatic exchange of information under CRS and FATCA, which decides who gets reported to whom, not who owes tax. The two systems classify independently: an entity that is opaque for income tax can still be a passive vehicle whose controlling persons are reported under CRS; a tax-transparent partnership is still an "entity" with its own CRS status and documentation duties. Nothing in a CRS self-certification answers a single question on this page, and no check-the-box election changes what a bank must report. When a term sheet says "the vehicle is transparent," the first follow-up question is always: transparent for which purpose, and in whose eyes?

Scenario 1: one partnership, two investor countries

The facts, fictional throughout: Ashford Ventures LP, a Delaware limited partnership, holds portfolio investments and in year 1 realises a $1,000,000 long-term capital gain. Its partners include a UK-resident individual (10%) and a German GmbH (10%). No cash is distributed in year 1. The table classifies one object — Ashford Ventures LP — from three named observers' standpoints.

The result is aligned — three observers, one timing — but only because the entity happens to classify the same way everywhere. Change one fact and the alignment breaks: make the vehicle a Delaware LLC with the same economics, and the US answer stays transparent (default for a multi-member LLC absent an election), while the UK observer flips to opaque under HMRC's list and continuing practice after Anson (Brief 15/2015). Now the UK member is taxed only on distributions, in a later year, as dividend-type income — and because the US taxed the member personally in year 1 on income the UK does not see until year 3, the double-tax relief mechanics stop matching, which is the precise dispute Anson litigated. Same economics, one different suffix on the entity, materially different UK result: this is why LLCs held by UK residents are a standing classification problem rather than a solved one — a topic continued in US LLCs for non-residents.

Scenario 2: one payment, dividend or interest

Fictional facts: a Cyprus company lends €10,000,000 to its UK subsidiary under a loan whose interest is 2% fixed plus 6% of the borrower's annual profits. To the lender's home system the instrument is a loan producing interest. The UK observer disagrees: the profit-linked element makes the securities "special securities" — the consideration for the principal "depends (to any extent) on the results of the company's business" (CTA 2010 s1015, condition C) — so the interest is treated as a distribution: non-deductible for the UK payer and dividend-like for treaty purposes. One payment, two characters.

Run the same instrument the other way — into a payer state that allows the deduction while the recipient state exempts the receipt as a dividend under a participation exemption — and the result is a textbook D/NI hybrid: deducted once, included nowhere. Within the EU, ATAD 2 Article 9 now closes exactly this: the payer state denies the deduction (primary rule), failing which the recipient state must include the payment (defensive rule). The practical consequence for structuring: a profit-participating shareholder loan needs the instrument classified by both observers before signing, because the after-tax cost of the same coupon can differ by the full corporate rate depending on which characterisation pair you land in — deductible/taxable, non-deductible/exempt, or (post-ATAD 2) forcibly re-matched.

Scenario 3: cash arrives after the tax

Fictional numbers, continuing scenario 1: Ashford Ventures LP allocates the UK partner $100,000 of gain in year 1 but distributes the cash only in year 3, after an exit. For every observer that classifies the LP as transparent, the tax year is year 1 — the allocation year (§706(a) for the US; the arising basis for the UK per INTM180010) — and the partner funds the tax from other resources: this is dry income, the standing cash-flow cost of transparency. The year-3 distribution is then, in outline, a return of already-taxed profit rather than a second taxable event (US mechanics track this through the partner's basis in its interest). For an observer that classifies the vehicle as opaque, the sequence inverts: nothing is taxable to the member in year 1, and the year-3 cash is the taxable event, as a dividend. Neither answer is wrong; they are different classifications doing exactly what they say. Fund documents manage the transparent-side problem contractually — tax-distribution clauses that advance enough cash to cover the partners' dry-income liability are standard in institutional partnership agreements, as described in fund tax architecture.