Context
Korea has one of the most structured approaches to guarantee fee pricing of any major OECD jurisdiction. The International Tax Coordination Law (ICITL) and its Enforcement Decree set out two permissible methods for pricing intercompany guarantees: the NTS model, a benefit-based approach that measures the interest saving to the guaranteed subsidiary, and the Moody’s model, a risk-based approach that measures the expected loss to the guarantor. Both are legally recognised. Both can produce arm’s length results. But they produce materially different numbers, they rest on different economic premises, and choosing between them — and documenting that choice contemporaneously — is one of the most consequential decisions a Korean multinational makes when pricing a parent guarantee to an overseas subsidiary.
This FT Note examines a 2024–25 pre-assessment review in which the NTS disallowed a guarantee fee rate of 0.42 per cent applied by a KOSPI-listed chemical manufacturer to its US subsidiary, substituting the NTS model median of 1.99 per cent and adding KRW 257m to income. The FTTP element is one of seven audit issues across 2020–2022, but it is the most methodologically instructive for groups with Korean entities providing cross-border guarantees.
The facts
The taxpayer is a KOSPI-listed Korean manufacturer of acrylic and UV curing chemical products, with subsidiaries in the US, Europe, and Asia. In 2019, it established a new US subsidiary (F) to build a local manufacturing facility. F required bank financing for construction and working capital, and the taxpayer provided a corporate guarantee. The taxpayer charged F a guarantee fee and recognised the income in its transfer pricing documentation.
For 2021 and 2022 H1, the taxpayer applied a fee rate of 0.42 per cent. This rate was not derived from an independent arm’s length analysis of F’s credit position or of the guarantee’s value to F. It was borrowed from the rate that the taxpayer’s parent company (B) had applied to its own US subsidiary (E) under an existing guarantee arrangement. From 2022 H2 onwards, the taxpayer switched to 0.9 per cent — the NTS-published rate for F, which had by then become available following two years of F’s financials being on file with the NTS.
The taxpayer’s transfer pricing documentation declared that it was using the NTS model as the basis for its guarantee fee pricing. However, for the period 2021–2022 H1, the NTS-published rate for F had been available from the 2020 tax year onwards. The taxpayer did not apply it, using instead the 0.42 per cent rate derived from B’s arrangement with E. The NTS audit found that the NTS model median for F in 2021–2022 was 1.99 per cent, and that the taxpayer’s applied rate of 0.42 per cent fell below the arm’s length range by KRW 257m.
At the pre-assessment review stage, the taxpayer argued for the first time that if the Moody’s model (rather than the NTS model) were applied, its 0.42 per cent rate would fall within the arm’s length range. It submitted supporting analysis at the review stage.
The decision
NTS model upheld — Moody’s argument rejected on two grounds
The review panel upheld the NTS’s assessment on both substantive and procedural grounds.
On the substance, the panel found that the NTS-published rate for F had been available from 2020, and that the taxpayer had no documented justification for applying B’s rate for E instead. The two subsidiaries — F and E — operated in different markets, had different financial profiles, and were at different stages of development. F was a newly established manufacturing entity with no operating history; E was an established distribution company. There was no basis for treating B’s E-applicable rate as a proxy for F’s arm’s length guarantee fee. The taxpayer’s own documentation had declared use of the NTS model; having made that election, it was bound to apply the published NTS rate when available.
On the procedural ground, Article 16(6) of the ICITL provides that where a taxpayer fails to submit required documentation within the prescribed deadline and submits it only at the stage of an appeal or review proceeding, the tax authority and review body may decline to use that evidence. The panel applied this provision to the Moody’s model analysis: it had not been part of the original documentation, had not been submitted during the audit, and was produced for the first time at the review stage. The panel found it appropriate to disregard the analysis on this basis.
ajiho commentary
The two Korean models — what they measure and why it matters
The NTS model and the Moody’s model are not interchangeable approaches to the same question. They measure fundamentally different things, and groups need to understand what each model is doing before choosing between them.
The NTS model is a benefit approach: it asks how much the subsidiary saves in interest costs because the parent has guaranteed its debt. The starting point is the subsidiary’s standalone borrowing rate (what it would pay without the guarantee) versus its guaranteed borrowing rate (what it actually pays). The difference is the benefit. The NTS publishes annual rates for each subsidiary on the basis of its financial statements, which means the rate is subsidiary-specific and updated annually. For a financially weak subsidiary, the benefit is large because its standalone rate would be high. For a strong subsidiary close to investment grade, the benefit is modest.
The Moody’s model is a risk approach: it asks how much risk the guarantor takes on by providing the guarantee. The calculation is anchored to the probability of default of the guaranteed entity and the loss given default — the amount the guarantor would have to pay if the subsidiary defaults. For a subsidiary with good credit quality, the expected loss is low and the resulting fee is low. For a financially stressed subsidiary, the expected loss is higher. The model produces a fee range rather than a fixed rate, which gives more flexibility but also requires more detailed credit analysis.
These two approaches can produce very different results in practice. A subsidiary that is financially weak (high standalone borrowing rate, low credit quality) will generate a large benefit under the NTS model but also a large risk premium under the Moody’s model — they might converge. But a subsidiary that borrows at low rates because of general market conditions (low interest rate environment) but has moderate credit quality might show a small NTS benefit while carrying meaningful default risk. In that scenario the two models diverge significantly. Bank guarantee rates, by contrast, are generally rejected in Korean and cross-jurisdictional practice — banks bear higher risk than corporate guarantors and lack group visibility.
The documentation trap — and how the taxpayer fell into it
The most instructive aspect of this case is not the rate dispute itself but the way the taxpayer’s documentation choices constrained its ability to argue at appeal. Three specific errors compounded each other.
First, the taxpayer declared in its transfer pricing documentation that it was using the NTS model. That declaration created a benchmark against which the NTS could test compliance. Having said ‘I am using the NTS model’, the taxpayer could not then argue at review stage that a different model would produce a better result. The declaration did not foreclose the Moody’s argument as a matter of law — Korean TP rules permit either model — but it fatally undermined the credibility of the late-stage submission.
Second, the taxpayer applied a rate borrowed from a different entity in a different market without any documented comparability analysis. B’s rate for E was not an NTS-published rate for F, and E and F were not comparable subsidiaries. The absence of any documentation explaining why the proxy rate was appropriate — or why the NTS-published rate for F (available from 2020) was not used — left the taxpayer with no evidential foundation.
Third, the Moody’s model analysis was produced for the first time at the review stage. Under Article 16(6) ICITL, the NTS and review bodies have a clear statutory basis to decline to use evidence submitted only in an appeal or review proceeding, and the panel exercised that discretion. The late submission problem is not unique to Korea — it mirrors the position in multiple jurisdictions and is consistent with OECD guidance on contemporaneous documentation — but the Korean provision is particularly explicit.
Which model should Korean groups use?
There is no universally correct answer, and the Korean rules do not require one model over the other. The right choice depends on the economic characteristics of the specific guarantee and the subsidiary being guaranteed.
The NTS model is the path of least resistance for most Korean multinationals. The NTS publishes rates, the calculation is relatively straightforward, and using the published rate for the specific subsidiary eliminates the primary area of audit dispute. Where the published rate is genuinely arm’s length and the subsidiary’s financial profile is stable, this is a defensible and administratively efficient approach.
The Moody’s model becomes more attractive where the NTS model produces a rate that is economically too high — for example, where the subsidiary has a very high standalone borrowing rate (generating a large benefit figure) but the guarantor’s actual credit exposure is limited because the subsidiary has good asset coverage, or where the guarantee is short-term and the probability of a claim is genuinely low. In those cases, the Moody’s model may produce a more economically accurate rate that better reflects the actual risk borne by the guarantor.
What this means for your business
Whatever model is chosen, the documentation must: (a) explicitly state which model is being used and why; (b) apply that model consistently and correctly; (c) use the NTS-published rate if the NTS model is selected, from the date it becomes available for that subsidiary; and (d) be prepared at the time of the transaction, not at audit or appeal stage.
For groups with Korean entities guaranteeing overseas subsidiaries, the practical lesson is that the choice of method is a decision to be made and documented up front — not an argument to be reached for once an assessment has landed. Borrowing a rate from a different subsidiary without a comparability analysis, and reserving a fallback methodology for the appeal stage, is precisely the pattern that failed here.
Case reference
Acrylic Resin Manufacturer v Regional Tax Office · National Tax Service Pre-Assessment Review (과세전적부심사) · tax years 2020–2022 · review filed 23 January 2025 · decision notified 26 December 2024
Decision in Korean. This FT Note covers the guarantee fee issue of a seven-issue pre-assessment review; the other issues (product pricing, seconded employee costs, trade receivables, technical service fees, and R&D tax credit) are outside its scope. Access via the NTS decision database.