Value Attribution and Cost Allocation of Platform and Social Welfare Services: An Economic Framework for the Revision of OECD Transfer Pricing Guidelines（2022） Chapter VII

Author: Liu Jianbo (Andy Liu) Author affiliation: Lianyungang City Audit Bureau, China

Abstract: Based on the author’s recommendation to the OECD tax seminar that China join the SBS mechanism and the written submission subsequently filed on the revision of Chapter VII of TPG 2022, this paper argues the necessity of China’s participation in the relevant mechanisms by reference to internationally recognised theory and practice of value economics. The implicit economic assumption underlying the simplified approach for intra-group services in Chapter VII of the OECD Transfer Pricing Guidelines is that service value varies approximately linearly with cost. In the digital economy, however, non-rival factors (algorithms, data) and cross-side network externalities in two-sided markets render the value function superlinear $(V=k\cdot n^{\beta },\beta >1)$. This paper proves that under linear cost-plus attribution the attributed value share declines monotonically to zero with scale — a directional, scale-amplifying systematic underestimation. Drawing on joint cost allocation theory, transaction cost economics, and the benefit principle of taxation, the paper develops three testable propositions — value attribution for platform-based IT services should adopt calibrated cost benchmarks; ESG inputs should be recognised at branch level under the two-dimensional framework of activity characterisation × accounting level; and the transaction-cost savings of public-interest gig and digital labour platforms constitute a measurable “benefit” under the benefit test, with social value serving as evidence of externalities — and demonstrates their interface compatibility with Amount B, UN Model Article 12B, the U.S. Services Cost Method (SCM), and the Pillar Two Side-by-Side arrangement. International legal practice — the final Apple State aid judgment and the rise and fall of India’s equalisation levy — shows that rule-based simplified benchmarks dominate ex-post discretionary adjustment. The calibrated cost benchmarks proposed here constitute a “measurement-as-cooperation” mechanism between China and its partners: under a minimum tax, shared measurement standards move the tax-competition game toward a Pareto-improving cooperative equilibrium.

Keywords: Value creation; Transfer pricing; Global Minimum Tax; Mechanical Linear Pricing; ESG; Platform-based IT services

JEL Classification: F23; H25; H26; H87; L86

1. Introduction

The revised draft of Chapter VII of the OECD Transfer Pricing Guidelines unifies the analysis of intra-group services around the chain “accurate delineation — benefit test — arm’s length pricing,” and retains the simplified approach for low value-adding services (LVAS): a cost pool, an allocation key, and a 5% markup, with explanations of specific cases. Its implicit economic assumption is that service value moves approximately linearly with cost, and the recipient’s benefit is observable through the recipient’s own operations. The assumption holds for traditional shared services (payroll, IT support, shared finance), but fails for three new configurations — platform-based IT enterprises; ESG investments coexisting at headquarters and branch level; and gig-economy and public-interest platforms. All three failures share one root: when the value function departs from the cost function, cost-based pricing rules systematically misattribute value. On the policy background: the author has already proposed to the OECD seminar, in respect of the three subject areas above, the recommendation that China join the SBS mechanism, and subsequently submitted a formal written submission on improving the value attribution of modern technology services of Chinese platform-based IT enterprises, the rational allocation of ESG public-interest investment costs, and the value attribution and concessions for purely gig-economy and freelancer platforms.

The paper argues that three policy recommendations flowing from this diagnosis — value attribution for platform IT services, localised cost recognition for ESG inputs, and cost-saving benefit measurement for gig and public-interest platforms — admit a unified theoretical foundation within the framework in which the value-creation objective is operationalised through the arm’s length principle (Devereux & Vella, 2017), and are compatible with existing multilateral rules. The contributions are fourfold: (i) a formal demonstration that linear cost-plus attribution underestimates value when value functions are superlinear, together with its methodological isomorphism with Amount B’s “calibrated simplification”; (ii) a linkage of the valuation of transaction-cost savings to the measurability of “benefit” under the benefit test, clarifying that externalities are considered only within the factual-evidence dimension; (iii) a cooperation mechanism derived from tax-competition theory, under which shared measurement benchmarks and the minimum tax are jointly Pareto-improving; and (iv) the promotion of positive incentives for public-interest gig and digital labour platforms in the social digital transformation — thereby grounding, in economic theory, the case that China’s digital exports can be taxed globally in a cooperative rather than conflictual manner.

2. Institutional Background and Theoretical Baseline: The Layered Structure of the Transfer Pricing Guidelines, the Status of “Value Creation,” and the Amount B Supplement

2.1 The four-layer structure of the TPG 2022: principles – methods – special transactions – administration

The OECD Transfer Pricing Guidelines have a clear layered structure and do not consist of principle-level rules alone. The 2022 edition (TPG 2022, released in January 2022) presents a four-layer structure:

• Core principles layer: the arm’s length principle (ALP) is the international standard agreed by OECD member states for tax purposes and embodied in Article 9(1) of the OECD Model Tax Convention (TPG 2022, paras 1.1–1.5 and Glossary). The “value creation” objective set by BEPS Actions 8–10, Aligning Transfer Pricing Outcomes with Value Creation, was proposed in 2015 and incorporated by the 2017 revisions into the application of the ALP as a normative objective, embedding the value creation of technology and other intangibles into the pricing process.

• Transfer pricing methods layer: traditional transaction methods (CUP, resale price, cost plus) and transactional profit methods (TNMM, profit split), selected under the “most appropriate method” rule (Chapter II, paras 2.1–2.12); the 2022 edition consolidates the profit-split guidance revised in 2018.

• Special-transaction guidance layer: intangibles (Chapter VI, including DEMPE analysis and the 2018 annex on hard-to-value intangibles, HTVI); intra-group services (Chapter VII); cost contribution arrangements (Chapter VIII); business restructurings (Chapter IX); financial transactions (Chapter X, added in 2020 and integrated for the first time in the 2022 edition). Intra-group services are therefore one branch of the special-transaction layer; the chain “accurate delineation – benefit test – arm’s length pricing” is itself a mapping of the Chapter I D accurate-delineation framework and Chapter III comparability analysis onto services.

• Administration and dispute resolution layer: documentation (Chapter V’s three-tier Master File, Local File and CbCR), advance pricing arrangements (Chapter IV), the mutual agreement procedure (Article 25 of the Model Convention) and dispute prevention. The acceptability of a simplified approach ultimately depends on whether the certainty gains at this layer can be verified — a dimension that must be included in any assessment of the institutional cost of the proposals below.

The test of whether a proposal fits the OECD framework is therefore not whether it repeats the language of “value creation,” but whether it can obtain an operable benchmark at the methods layer, a textual anchor at the special-transaction layer, and certainty gains at the administration layer.

2.2 “Value creation” and the ALP: normative objective versus operative standard

It is common in the literature to present “value creation” as a principle competing with the ALP. This paper does not adopt that reading. The two stand in an objective-and-standard relationship: value creation supplies the normative objective (profits should be taxed where economic activity and value creation occur), while the ALP supplies the operative comparative standard (the conditions of comparable uncontrolled transactions). TPG 2022 para 1.51 requires the functional analysis to understand “how value is created by the group as a whole” and the interdependence between the functions performed by the associated enterprises and the rest of the group, while paras 6.48 and 6.133 require the identification of all factors contributing to value creation in intangibles cases. These provisions show that value creation has been embedded into the pricing application process of the ALP rather than established as a separate rule.

Two genuine operational difficulties remain, and neither is a “dilemma” in the sense of opposition:

1. Measurement: if value creation means contribution to marginal output it cannot be observed directly; if it means equilibrium market value it must be approximated through accounting measures (Devereux & Vella, 2017). The scale of the attribution failure is documented by Tørsløv, Wier and Zucman (2023), who estimate from macro data that some 36% of multinational profits are shifted to low-tax jurisdictions. The equilibrium market value so defined raises a contested question between destination-based value and production-based value. This is precisely the proxy-benchmark problem addressed by Propositions 1 and 3.

2. Timing: the tension between ex ante rule-based benchmarks and ex post case-by-case discretion. Rule-based approaches reduce compliance and dispute costs but may mismatch the facts (Proposition 1 criticises precisely the mismatch of mechanical linear benchmarks); discretion fits the facts better but generates dispute costs and uncertainty (the Apple case being a clear illustration of the latter cost; see Section 6).

Comparability analysis in the TPG is therefore two-staged: at the first stage, the five economically relevant characteristics listed in Chapter I para 1.36 (contractual terms; functions, assets and risks, including how those functions relate to the wider generation of value by the MNE group; characteristics of the property or services; economic circumstances; business strategies) are used to “accurately delineate the actual transaction” in the group ALP context (para 1.37); only at the second stage does the comparability-adjustment analysis follow (mostly in Chapter III), comparing controlled and uncontrolled transactions in the group ALP context (paras 1.39, 3.1–3.4, 3.47–3.54). This distinction is critical for the present paper: the first step of Proposition 1 is not to demand a change in the pricing formula, but to argue that algorithm-intensive platform IT services are excluded under the four conditions of TPG 2022 Chapter VII Section D, para 7.45 on low value-adding intra-group services and are matched by way of analogy with Example b) in para 7.50, and that Chinese platform-based IT enterprises are not within the accurate delineation of low value-adding services at all. Only once this positioning is completed does it make sense to choose a benchmark matched to the value mechanism at the methods layer.

2.3 The 2026 draft revision of Chapter VII: the rule chain and the existing boundaries of the simplified approach

The 2026 draft revision of Chapter VII consolidates the analysis of intra-group services into the chain “accurate delineation → benefit test → pricing” (draft paras 7.5–7.70, Sections B–C), retains the simplified approach for low value-adding services (cost pool → allocation key → 5% markup → thresholds → documentation → withholding tax; draft paras 7.74–7.96 (Section E)) and adds Annex I with 21 examples. Its direction is the “internalisation of principles” — bringing service charges fully within the ALP and the general framework of Chapters I–III. The 2026 draft restructures and renumbers the paragraphs of TPG 2022 but inherits without changing these boundaries — that is, the 2022 definition of low value-adding services, the exclusion precedent, the anti-benchmark caution and the threshold are retained unchanged under new paragraph numbers. To reflect this inheritance, the three points below cite both sets of paragraph numbers, with the 2026 draft numbers in parentheses.

Three institutional points deserve emphasis:

1. The simplified approach is not a “low-value presumption” but an optional simplification with clearly defined boundaries: TPG 2022 para 7.45 (draft para 7.76) requires four conditions to be met cumulatively (supportive in nature; not part of the core business of the group; not using and not creating unique and valuable intangibles; not involving the assumption or control of significant risks by the provider).

2. The Guidelines themselves already contain a precedent excluding algorithm-intensive services: in Example b) of TPG 2022 para 7.50 (draft para 7.81, Example b)), a subsidiary of an investment-banking group uses internally developed confidential credit-risk analysis models, algorithms and software, and the service is not a low value-adding intra-group service. This gives Proposition 1 a textual anchor internal to the Guidelines — its narrow claim (that algorithm-intensive services do not qualify for the 5% simplification) holds without reliance on external empirical evidence.

3. TPG 2022 para 7.61 (draft para 7.92) itself already contains the anti-abuse warning (the LVAS markup should not be used as a benchmark for similar services outside the defined scope), and para 7.63 (draft para 7.94) authorises tax administrations to set financial-ratio thresholds and to require full functional and comparability analysis where the threshold is exceeded. What this paper advocates is the strengthening and application of these existing boundaries, not a debate on whether the 5% simplification itself is reasonable.

2.4 Amount B (2024): an institutional precedent for calibrated simplification

TPG 2022 (January 2022) did not contain Amount B at the time of its release. The final Amount B report was released by the OECD/G20 Inclusive Framework on 19 February 2024 and was subsequently incorporated as an annex to Chapter IV of the Guidelines (applicable from 2025), constituting an incremental supplement to the TPG system after the 2022 edition. Its relationship to TPG 2022 is inheritance plus supplementation, not the creation of a separate system:

• Inherited elements: (i) the ALP — Amount B is positioned as a simplified and streamlined approach to applying the ALP, not a departure from it; (ii) the comparability framework — the in-scope perimeter is defined by functional and risk profiles and by exclusions (distributors assuming significant risks, owning or using unique and valuable intangibles, or performing non-baseline distribution activities are excluded), in substance a pre-parameterisation of the Chapter I five-factor framework; (iii) the TNMM — the tested distributor’s or sales agent’s operating margin (net operating profit over sales) is the core indicator, benchmarked against global comparables, making Amount B a “pre-fabricated” form of TNMM; (iv) the function–risk–return matching logic of distribution activities.

• Added elements: (i) the standardised pricing matrix, which determines operating-margin ranges by crossing industry groupings with factor intensity, matrix cells spanning 1.50%–5.50% return on sales, with a compliance range of the matrix value ±0.5 percentage points (cells are formed by crossing industry groupings 1–3 with factor-intensity classes A–E; outcomes outside the range are adjusted to the matrix return on sales); (ii) the operating expense cross-check, which applies a cap-and-collar (floor and cap) to the equivalent return on operating expenses (net operating profit over operating expenses): a 10% collar and caps of 40%/60%/70% (default) or 45%/70%/80% (alternative caps for qualifying jurisdictions) by factor intensity, adjusting the return downwards above the cap and upwards below the collar; (iii) the data availability adjustment, which for qualifying jurisdictions with insufficient data in the global dataset applies the formula Adjusted return on sales = ROSTP + (NRAJ × OAS), where NRAJ runs from 0% (BBB+ and above) to 4.9% (B-) and about 8% (CCC- and below) by sovereign credit rating and OAS is capped at 85%, correcting the systematic underestimation caused by data availability; and (iv) tax certainty and dispute prevention, in the form of the political commitment of Inclusive Framework members to respect Amount B outcomes for covered jurisdictions, supported by MAP and APA tools.

Amount B has three methodological implications for this paper. First, it demonstrates that “replacing case-by-case comparability analysis with a calibrated parametric benchmark” is already accepted within the system — in other words, there is an institutional precedent for “replacing uncalibrated simplification with calibrated simplification.” Second, its three mechanisms — matrix calibration, cross-checking and data-availability adjustment — are precisely the operational requirements of “calibration”; the reference benchmark advocated in Recommendation 1 should be modelled on the same structure (qualified-service delineation, cost scope, calibration parameters, reverse checking and safe-harbour boundaries), rather than simply citing another country’s markup. Third, its scope exclusions and the four conditions of TPG 2022 para 7.45 share a common logic, jointly supporting the proposition that algorithm-intensive and unique-intangible-intensive activities do not qualify for low value-adding simplification. Two limits must be noted: Amount B is a simplification for distribution and does not directly cover platform IT services, so what this paper claims is methodological isomorphism, not direct applicability; and its specific parameters have been verified item by item against the 2024 report (sections 5.1–5.3 and tables 5.1–5.3) and are presented in structured form in this paragraph.

2.5 A parallel benchmark: Article 12B of the UN Model

Article 12B of the UN Model (added in 2021) provides two simplified mechanisms for automated digital services: a gross basis (a bilaterally negotiated percentage, the Commentary suggesting 3%–4%, as non-binding guidance) or a net basis (qualified profit = 30% × the ADS segment profit margin × gross revenue from the source jurisdiction). Its institutional significance is that it shows how developing jurisdictions can secure a stable tax base without relying on case-by-case comparables, sharing the “calibrated simplification” methodology with Amount B; but the two operate on different tracks (Article 12B is a withholding-type gross-basis charge, whereas Amount B is profit attribution within the ALP), and this paper therefore uses it only as a methodological parallel, without conflating the two.

2.6 Locating the three recommendations

The three recommendations can accordingly be located as follows (paragraph references use the 2026 draft, with the corresponding TPG 2022 paragraphs after the slash):

Recommendation	Layer	Principal anchors (draft / TPG 2022)	Features of the proposed landing
1. Platform IT services	Special-transaction layer (Chapter VII D boundary) → methods layer (Chapters II–III)	Draft 7.76, 7.80–7.81 / TPG 2022 7.45, 7.49, 7.50(b); draft 7.92 anti-benchmark warning, 7.94 threshold / TPG 2022 7.61, 7.63	Boundary clarification + calibrated benchmark; the 5% markup and the benefit-test standard are not touched
2. Localised ESG costs	Special-transaction layer (indirect charging, cost pool and allocation keys) + methods layer (comparability adjustments)	Draft 7.45–7.49 (indirect charging and allocation keys) / TPG 2022 7.23–7.26; draft 7.87–7.91 / TPG 2022 7.56–7.60; TPG 2022 1.36 (FAR) and 3.47–3.54 (comparability adjustments)	Cost recognition and pooling; no expansion of the markup base
3. Public-interest / gig platforms	Application of the core principle (benefit test) + special-transaction layer (allocation keys)	Draft 7.13 / TPG 2022 7.6; draft 7.68–7.70 (whether a profit element is required; tax administration discretion) / TPG 2022 7.35–7.37; draft 7.90–7.91 / TPG 2022 7.59–7.60	Extension of “benefit” along the evidence dimension; externalities do not enter the consideration

Table 1. Location of the three recommendations within the Transfer Pricing Guidelines framework

The common theoretical locus of the three recommendations is this: within the existing four-layer structure of “principles – methods – special transactions – administration,” the measurement problem of value being hard to observe directly is converted into a comparability problem — choosing the proxy benchmark closest to the value-creation mechanism and equipped with calibration and cross-checking. Recommendation 1 takes the U.S. transfer pricing cost standard (structured as Treas. Reg. §1.482-9 Services Cost Method, SCM, with its safe harbour) as its reference benchmark, with the Pillar Two Side-by-Side (SbS) arrangement as its institutional interface, providing an interface opportunity for Chinese platform-based IT service enterprises and corresponding to the “qualified-service delineation + cost scope + safe harbour” structural template; Recommendation 2 rationally takes ESG localised costs into account, recognising branch-level localised costs at the cost-pool and allocation-method level, corresponding to the “causal/beneficial matching” pooling principle; Recommendation 3 extends “benefit” to measurable transaction-cost savings, corresponding to the boundary rule of “accommodating externalities along the evidence dimension without disturbing the consideration dimension”, providing a leading policy environment for the social development of public-interest gig and digital labour platforms. This is also where the paper interfaces with Amount B, Article 12B and the U.S. services cost standard: they supply, respectively, a matrix-calibration structure, a formulaic simplification and a “qualified-service delineation plus safe harbour” template; what this paper advocates is the application of their methodological logic to digital platform services, a scenario not yet covered by dedicated rules. The shared premise is: when value cannot be observed directly, choose the proxy benchmark closest to the value-creation mechanism rather than mechanically applying linear rules designed for a different class of activities.

3. Proposition 1: Non-rivalry and Scale Superlinearity Rule Out Mechanical Linear Pricing

Economic theory sources of Proposition 1. Romer (1990) and Arrow (1962) established the non-rivalry of knowledge and algorithms — marginal replication cost approaches zero. Jones and Tonetti (2020) prove that data, as a non-rival factor, generates social value that grows with scale without depletion, and that non-rivalry generates increasing returns. Platform economics (Rochet & Tirole, 2003; Katz & Shapiro, 1985) formalises cross-side network externalities: in two-sided markets the structure of prices, not merely their level, is a strategic variable. Bresnahan and Trajtenberg (1995) classify such technologies as general purpose technologies, whose value materialises through downstream complementary innovation (Brynjolfsson & Hitt, 2000).

Formalisation. Let the value function of platform-based IT services be $V\left(n\right)=k\cdot n^{\beta }(\beta >1,n=\mathit{users}/\mathit{nodes})$ and the cost function $c\left(n\right)=a+b\cdot n.$ The value share attributed by cost-plus pricing is

$$\frac{P\left(n\right)}{V\left(n\right)}=\frac{(1+m)(a+b\cdot n)}{k{n}^{\beta }}\stackrel{n\to \infty }{\to }0(\beta >1)$$

Parameter	Definition	Economic meaning and theoretical basis
$$V\left(n\right)$$	the economic value created by the platform-based IT service at scale $n$	the object of attribution under the value-creation principle; not directly observable, hence examined through the share relation in Eq. (1) (Devereux & Vella, 2017)
$$n$$	user/node scale: combined active users of both sides of the two-sided market, or service-invocation nodes	in a two-sided market the value accruing to each side grows with the other side’s scale (Rochet & Tirole, 2003); $n$ indexes the joint scale of both sides
$$\beta$$	network-scale elasticity (the exponent of value in scale): $\beta =1$ linear, $\beta >1$ superlinear	two micro-foundations: cross-side network externalities (the classic approximation $V\propto n^2$, i.e. $\beta =2$) and increasing returns to non-rival factors (Jones & Tonetti, 2020). $\beta$ is the key structural parameter distinguishing platform and algorithm assets from traditional services — and the quantity that comparability analysis must calibrate
$$k$$	scale-invariant value-efficiency coefficient ($k>0$)	captures cross-sectional differences in technology level, data quality and matching efficiency; corresponds to general-purpose-technology attributes (Bresnahan & Trajtenberg, 1995) and organisational complementarities (Brynjolfsson & Hitt, 2000). Higher $k$ means more value created at any given scale
$$a$$	fixed cost independent of $n$: R&D, data infrastructure and other sunk investments	dominant in digital-economy cost structures; non-rivalry implies $a$ is incurred once and can be reused indefinitely without depletion (Romer, 1990)
$$b$$	constant marginal cost per additional user/node ($b>0$)	algorithms and data drive $b$ toward zero while keeping it linear — this is precisely why the cost function is approximately linear; the divergence arises because cost is linear while value is superlinear
$$m$$	the cost-plus markup rate; uniformly $m=5\%$ under the LVAS simplified approach	the markup compensates the service provider’s profit element above cost; $(1+m)$ is the cost-plus multiplier
$$P\left(n\right)$$	the related-service transfer price at scale $n$ (cost-plus basis)	the operationalisation of the “arm’s length pricing” step in the Chapter VII rule chain
$$P\left(n\right)/V\left(n\right)$$	the value share attributed per unit of cost-plus price	the core result of Eq. (1): for $\beta >1$ the share declines monotonically to zero with $n$. If value were proportional to cost ($\beta =1$), the share would be constant and scale-invariant — the scale-declining share is the mathematical expression of the systematic underestimation under mechanical linear pricing

Table 2. Parameter definitions and theoretical interpretation of Eq. (1)

The divergence mechanism in Eq. (1) decomposes into three steps. First, the cost side is linear because $b$ is constant. Second, the value side is superlinear owing to cross-side network externalities and non-rivalry. Third, cost-plus pricing transmits only cost information ($P=(1+m)c$) and carries no information about $\beta$ or $k$ — hence the larger the scale and the higher the technology and data intensity, the wider the gap between $P$ and $V$. “Calibration,” formalised, means making the markup a function of the structural parameters, $P=c\cdot (1+\tilde{m}(\beta ,k))$, so that the attributed share no longer declines monotonically with $n$ — this is the mathematical expression of “adopting cost benchmarks matched to the value mechanism”. Eq. (1) is precisely addressed to the problem of Proposition 1: it constitutes the theoretical basis of the author’s reasoning in the written submission to the OECD that “it is inappropriate to mechanically apply a low rate by way of simple formula”, and the mathematical foundation for optimising parameter selection so that the cost benchmark matches the actual value mechanism. The following proposition is accordingly advanced.

Provenance note on Eq. (1). Eq. (1) is a stylised expression constructed for this paper, a concrete elaboration of the theoretical model in the author’s submission to the OECD: it is neither drawn from the original submission (which is purely discursive and contains no equations) nor quoted in toto from any single source. Its components do, however, have verifiable counterparts in the literature. The value–scale power law $V\propto N^{\gamma }$ is known in the literature as Nivi’s law (proposed by Nivi in 2005; systematised by Scala & Delmastro, 2023), where $\gamma =1$ corresponds to Sarnoff’s law (broadcast networks), $\gamma =2$ to Metcalfe’s law (point-to-point networks), and Briscoe, Odlyzko & Tilly (2006) correct Metcalfe’s equal-weighting of connections with an $N\ln N$ form. Scala & Delmastro (2023, Scientific Reports) fit this power law to some 20 years of data on Google, Meta and other advertising-funded networks, obtaining exponents $\gamma \approx 2.2$–14 — all superlinear. Cost-plus pricing $P=(1+m)c$ is the standard textbook form. The limit result — the attributed share declining to zero as scale expands through network effects — is the present paper’s own mathematical derivation based on Eq. (1); the author has found no equivalent proposition in the transfer pricing literature, and in formal publication it should accordingly be presented as “this paper’s proposition” rather than as a result from the literature.

Proposition 1 (Non-rivalry and scale superlinearity rule out mechanical linear pricing — the Underestimation Theorem for platform-based IT service transfer pricing). For β > 1, the attributed value share under traditional linear cost-plus pricing declines monotonically to zero as scale expands through network effects. Cost-plus attribution — including the LVAS 5% markup — is not a “conservative approximation” but a directional, scale-amplifying systematic underestimation. The true markup parameter should become a function of the structural parameters, $\tilde{m}(\beta ,k)$: the $P\left(n\right)/V\left(n\right)$ ratio of platform-based IT services declines as user scale expands through network effects, which is the mathematical expression of the “systematic underestimation of mechanical linear pricing”. The LVAS simplification implicitly excludes services whose value functions are significantly superlinear; mechanically applying it violates the simplification’s own conditions of applicability.

Policy implication. The correction is not to abandon cost benchmarks but to choose cost benchmarks matched to the value mechanism — explicitly incorporating scale variables and GPT attributes into comparability analysis. This is methodologically isomorphic to Amount B: replacing uncalibrated simplification with calibrated simplification.

4. Proposition 2: ESG Joint Costs Demand Recognition at the Value-Creation Unit

Economic theory sources of Proposition 2. Thomas (1974) established the foundational result on joint cost allocation: there is no uniquely “correct” allocation of joint costs; every allocation is a choice whose defensibility can only derive from the decision purpose and a causal or benefit anchor — testable against his three criteria of additivity, definiteness, and defensibility. Cooper and Kaplan’s (1988) activity-based costing, ABC (Activity-Based Costing), turns cost accounting into the “cost driver” logic. OECD draft para. 7.88 requires the cost pool to exclude costs incurred “solely to perform for another member” (cost-pool determination see draft paras. 7.87 and 7.89), embodying the same causal principle; Hirshleifer (1956) framed transfer pricing within decentralised (divisionalised) organisations.

ESG investments are typically joint costs: a green data-centre retrofit, a group data-governance system, or carbon-emission and other sustainability disclosure activities simultaneously serve headquarters’ compliance obligations and branches’ localised operations. Aggregation at headquarters level produces a truncated causal chain (branch-level benefits unobservable) and a measurement mismatch (recognition timing and measurement attributes differ between group and local bases). Three normative supports follow: (i) the decentralisation tradition in management accounting; (ii) the value-chain logic of sustainability disclosure — GHG Protocol Scope 3, ISSB IFRS S1/S2 (2023), and the EU CSRD (Directive (EU) 2022/2464) all require tracing emissions and impacts to value-chain nodes; (iii) Chinese practice — the Interim Provisions on Accounting Treatment of Enterprise Data Resources (MOF Cai Kuai [2023] No. 11, effective 1 January 2024) provide an accounting interface for recognising and measuring data resources, whose objects largely arise at the business front line.

The following proposition is accordingly advanced.

Proposition 2. ESG inputs of multinational enterprises have joint-cost characteristics, and their characterisation and attribution should follow a two-dimensional framework of activity characterisation × accounting level: decompose by composition (compliance / investor requirements / commercial operations / charitable input), then recognise localised costs at the value-creation unit (branch). Through this dual decomposition and confirmation by composition and accounting unit, the evidence chain of the benefit test is kept complete.

5. Proposition 3: Transaction-Cost Savings of Public-Interest Gig and Digital Labour Platforms Are Measurable Benefits Under the Benefit Test; Social Value Is Evidence of Externalities

Economic theory sources of Proposition 3. Coase (1937) interpreted the firm as a substitute for market transaction costs, making explicit that the value of the firm lies in substituting for the market and, by extension, in substituting for measurable benefits; Coase (1960) established transaction costs as the measuring rod of social-cost analysis. Gig-economy and public-interest platforms are market-aggregating infrastructure: they reduce the unit transaction cost of masses of small transactions. ILO (2021) and the EU Platform Work Directive (Directive (EU) 2024/2831) confirm digital labour platforms’ infrastructural role in labour-market matching.

Key translation. “Reducing society’s transaction costs” translates into transfer pricing language as

$$B_i={\mathit{TC}}_i^{\mathit{market}}-{\mathit{TC}}_i^{\mathit{group}}\ge 0$$

Parameter	Definition	Economic meaning and theoretical basis
$$i$$	index of the group member / service recipient	the benefit test runs member by member; “benefit” must be identified for each recipient and cannot be aggregated at group level
$${\mathit{TC}}_i^{\mathit{market}}$$	counterfactual comparable transaction cost: the full cost member $i$ would bear when sourcing equivalent infrastructure services from an independent third party	comprises two components: the market service price (comparable uncontrolled price basis) plus the transaction costs of using the market — search, contracting, monitoring and enforcement (Coase, 1937). It is precisely the latter that constitute the organisational gain of intra-group supply replacing arm’s-length sourcing
$${\mathit{TC}}_i^{\mathit{group}}$$	the cost actually borne by member $i$ through the group platform	includes the service fee or cost allocation charged by the group plus the member’s own internal coordination costs; excludes costs borne by other members, aligning with the cost-pool exclusion rules (draft para. 7.88; cost-pool determination see paras. 7.87, 7.89)
$$B_i$$	the cost-saving benefit	member $i$‘s net benefit from the group infrastructure; the direct measurement object of the benefit test, corresponding to the transaction-cost form of “economic or commercial value” in draft para. 7.13
$$\ge 0$$	participation constraint (the pass condition of the benefit test)	$B_i\ge 0$ is equivalent to “using the group platform weakly dominates arm’s-length sourcing” — the mirror image of Coase’s (1937) internalisation condition applied to intra-group services; if $B_i<0$ (external sourcing is cheaper), the benefit test fails and the draft requires a zero charge — the formalised criterion against the “zero-fee—zero-benefit” misjudgement
Measurement basis	$B$ is measured through observable proxies: market-comparable service quotations, and the full cost of platform alternatives (build-or-buy)	$B$ measures the private benefit between contracting parties and excludes third-party externalities, which belong to the evidence dimension (see below)

Table 3. Parameter definitions and theoretical interpretation of Eq. (2)

Three theoretical points about Eq. (2). First, the source of $B$ is the Coasean transaction-cost differential: the search, contracting and enforcement costs saved when group organisation replaces the market constitute the recipient’s measurable benefit — the proposition introduces no new concept, merely extends the draft’s para. 7.13 category of “benefit” to the transaction-cost dimension. Second, $B\ge 0$ formalises the benefit test as a participation constraint: the test asks whether an independent party would transact on these terms, and the constraint supplies the operational criterion. Third, the boundary with profit-based benefits: $B$ is a cost-side benefit and is not double-counted with profit-split or residual-profit attribution — the former belongs to the transaction stage, the latter to the value-creation stage; they are complementary along the activity-characterisation dimension.

Provenance note on Eq. (2). Eq. (2) is an expression constructed for this paper for Proposition 3, a formulaic supplement to the author’s original submission and to the OECD Chapter VII draft (draft para. 7.13 definition of “benefit”); both the original submission and the draft are purely discursive and contain no corresponding formula to cite. Its logical skeleton is Coase’s celebrated boundary condition (1937: 395): “a firm will tend to expand until the costs of organising an extra transaction within the firm become equal to the costs of carrying out the same transaction by means of an exchange in the open market or the costs of organising in another firm” (the firm expands up to the critical point at which the cost of organising an additional transaction within the firm equals the cost of carrying out the same transaction through exchange in the open market, or equals the cost of organising it by another entrepreneur) — $B_i={\mathit{TC}}_i^{\mathit{market}}-{\mathit{TC}}_i^{\mathit{group}}\ge 0$ is the mirror image of that equality, parameterised for the benefit test of a service recipient. A structural relative exists in the literature on firm boundaries (e.g., Bylund in the Strategic Management Review formalises Coase’s condition as a net-cost-difference inequality), but its object is the firm–market boundary rather than the recipient’s benefit, and it is therefore not listed as a formal citation.

where B is the cost-saving benefit: the comparable transaction costs a recipient would bear when sourcing equivalent infrastructure from an independent third party, minus the costs actually borne through the group platform. B is measurable (market-comparable proxies), fits the existing framework (draft para. 7.13 defines “benefit” as “economic or commercial value,” of which cost savings are a standard form — cf. centralised procurement), and respects arm’s length logic.

Discipline on externalities. The beneficiaries of positive externalities are the unspecified public, not the contracting counterparty. Pigou (1920) and the public-finance separation of the benefit and ability-to-pay principles (Musgrave, 1959) imply that externalities must not enter the consideration of related-party transactions; their correct place is as auxiliary evidence in the benefit test and as reasonableness variables in allocation-key design. What is multidimensional is the evidence dimension of the identification system; the pricing dimension of the transaction remains single-metric. The commons governance of public-interest platforms (Ostrom, 1990; Benkler, 2006) supplies institutional materials for the multidimensional evidence system; China’s “Data Twenty Articles” (2022) and the “Data Element ×” programme (2024–2026) confirm, at the policy level, that scale-value of data can be institutionally measured. Studies of public-interest gig and digital labour platforms in the United States and China show that the social value of such leading platforms is evidence of externalities.

The following proposition is accordingly advanced.

Proposition 3. The transaction-cost savings of public-interest gig and digital labour platforms are measurable “benefits”; social value is evidence of externalities.

6. International Rule-First Optimisation in Legal Practice: Rule-Based Benchmarks versus Ex-Post Discretion

The Apple State aid final judgment (ECJ C-465/20 P). On 10 September 2024 the Court of Justice (Grand Chamber) set aside the General Court’s 2020 judgment, confirmed the Commission’s 2016 decision, and ordered Ireland to recover some €13 billion from Apple (tax rulings of 1991–2014 excluding profits from IP licences from the Irish tax base). Two lessons follow. First, ex-post, case-by-case, discretionary reallocation of profit is costly and uncertain — a decade of litigation and €13 billion parked in escrow evidence the price of unpriced value-attribution rules. Second, adjudication ultimately proceeded through arm’s-length equivalence assessment, indicating that standardised, calibrated benchmarks — not one-off discretionary adjustment — are the route to reducing such disputes. Simplified, forward-looking, calibrated cost benchmarks internalise the litigation-cost externality the Apple case laid bare.

India’s equalisation levy, rise and fall. India levied 6% on online advertising from 1 June 2016, extended a 2% e-commerce levy from 1 April 2020, abolished the latter on 1 August 2024 (Finance Act 2024) and the former with effect from 1 April 2025 (Finance Act 2025). The life cycle of the unilateral instrument — imposition → threat of trade retaliation (USTR Section 301) → repeal — shows that unilateral measures lack equilibrium stability, while multilateral simplified benchmarks (Amount B, UN 12B, SbS) are the stable form of source-country taxing rights. This is precisely the institutional reading behind Recommendation 1’s choice of a “multilateral interface + calibrated foreign benchmark” over unilateral assertion.

The WTO moratorium on customs duties on electronic transmissions. At MC13 (Abu Dhabi, February–March 2024) members extended the moratorium until the next Ministerial Conference or 31 March 2026, whichever is earlier; the moratorium lapsed when MC14 declined to renew it in March 2026, and from 8 May 2026 a group of members has committed, through a communication, not to impose customs duties on electronic transmissions among themselves. The tension over tariff instruments for digital trade, and the moratorium’s looming expiry, underscore the value of stable, predictable revenue-sharing rules inside the income tax system — the trade-policy backdrop to Section 7’s cooperation framework.

7. Cooperative Benchmarks in the Age of the Minimum Tax: China’s External Economic Ties and Global Tax Coordination

7.1 The theory of the minimum tax: competition failure and cooperative optimisation. The canonical result of tax-competition theory is that decentralised competition over mobile bases leads to under-taxation and under-provision of public goods (Zodrow & Mieszkowski, 1986; Wilson, 1986, 1999). Kanbur and Keen (1993) prove in a size-asymmetric model that small countries lose from harmonisation to any rate inside the non-cooperative range, but both countries gain from a minimum tax imposed within that range. In the game-theoretic setting of international taxation, the Pillar Two 15% global minimum tax has theoretically optimal cooperative features. In the light of theory and practice, the demand that “countries require a minimum tax” bears heavily on large countries, and on China as well; but given the rational cooperative equilibrium in response to tax-competition failure, such a cooperative equilibrium is more reasonable than a trade-war-style equilibrium. China also has a number of emerging industries and IT platforms that meet the exemption conditions of Side-by-Side and can apply for exemption where conditions are met; meanwhile, as an Inclusive Framework member, China should maintain the logic of active participation in international governance.

7.2 Measurement is cooperation. The deep cause of international tax disputes is the absence of a shared measurement standard: jurisdictions attributing the same value with different proxies generate double taxation and double non-taxation. The three recommendations chart a path to shared standards. Through the functional-equivalence architecture of Side-by-Side — the OECD Inclusive Framework formally adopted the Side-by-Side Package on 5 January 2026 (SbS safe harbour, UPE safe harbour, simplified-ETR safe harbour, substance-based tax-incentive safe harbour, transition CbCR extension; the United States — through its NCTI regime — being the first qualified SbS jurisdiction and the only one listed as at 1 January 2026; qualified jurisdictions are entered on an OECD Central Record and the safe harbour applies to fiscal years beginning on or after 1 January 2026) — “calibrated foreign cost standards” become a legitimate starting point for comparability analysis. The economic significance: once the measurement standard is shared, the tax-competition game turns from benchmark competition into parameter calibration — the object of negotiation shifts from “who is right” to “how to calibrate,” sharply lowering the transaction costs of disputes. This parallels the maturation of bilateral Advance Pricing Agreements: measurement consensus precedes revenue distribution.

7.3 China’s external ties: the economics of export benefits. The world-economy benefits of China’s platform and manufacturing exports run, first, through the deepening, mobilisation of consumption potential and matching efficiency: cross-border e-commerce platforms lower the fixed costs of small-parcel international trade, letting consumers and micro-enterprises — including in developing countries — access global goods at lower prices; this is the income-tax analogue of the WTO moratorium’s goal of a duty-free digital-trade environment (Section 6). Second, through revenue stability rather than erosion: Pillar Two guarantees each jurisdiction a minimum effective rate of 15%, and the baseline correction advocated here (calibrated cost standards; cost-saving benefit measurement) makes the value attribution of specific emerging and IT platform enterprises auditable, stable, and predictable — source countries obtain a determinable tax base without resorting to unilateral instruments. The Belt and Road Initiative Tax Administration Cooperation Mechanism (BRITACOM, established at Wuzhen, 18–20 April 2019, by 34 jurisdictions signing the memorandum of understanding, secretariat in Beijing) provides a ready multilateral platform for diffusing and calibrating such measurement standards.

7.4 Policy propositions. (i) Recognise “calibrated foreign cost standards” as a legitimate starting point of comparability analysis, with equivalence reasoning following the SbS precedent, providing an SbS-type interface for Chinese platform-based IT services; (ii) incorporate the cost-saving form into the definition of “benefit” via official examples, to the benefit of gig and public-interest platforms in developed and developing countries alike and facilitating a smooth social transformation; (iii) recognise the legal status of branch-level localised cost recognition in the ESG field, aligning with the substance-based carry-forward (payroll + tangible assets); (iv) use platforms such as Belt and Road tax cooperation to diffuse the measurement paradigm of data-element value accounting, forming a third path between pure formulary apportionment and pure case-by-case comparability. Stable simplified benchmarks between China and its regional partners likewise have regional-balance value, help to ease tariff disputes, and provide the micro-foundation of mutual benefit.

8. Further Testable Propositions and Conclusion

8.1 Controlling for cost and industry, platform-based IT service providers’ arm’s-length markups rise with their network-scale elasticity β; the cross-border attribution error of platform-service profits under a uniform 5% markup rises with β.

8.2 ESG inputs aggregated at headquarters yield lower branch-level benefit-test evidence completeness than localised recognition; the latter exhibits lower dispute rates in transfer-pricing audit adjustment.

8.3 For public-interest/gig platforms, benefit-test conclusions based on cost-saving benefit measures differ statistically and significantly from profit-only metrics; the former avoids “zero-fee—zero-benefit” misclassifications.

8.4 In size-asymmetric tax-competition models (Kanbur & Keen, 1993), a shared measurement standard — which compresses the technical space for profit shifting — is complementary to the minimum tax: the minimum tax compresses the range of rate competition; measurement benchmarks compress the space of base competition. Jointly they approach the cooperative equilibrium.

Conclusion. The analytical chain — superlinear value function → systematic underestimation by linear cost-plus → matched calibrated benchmarks → the two-dimensional characterisation × level framework → cost-saving benefit → the cooperation equilibrium of shared measurement — grounds all three recommendations theoretically and demonstrates their compatibility with Amount B, UN 12B, SCM, and the SbS arrangement. International legal practice shows ex-post discretion is costly and unilateral instruments unstable; rule-based simplified benchmarks and shared measurement standards are the cooperative path for international taxation in the age of the minimum tax — and the theoretical expression of mutually beneficial economic exchange between China and its partners.

References

Arrow, K. J. (1962). Economic Welfare and the Allocation of Resources for Invention. In The Rate and Direction of Inventive Activity. Princeton University Press.

Benkler, Y. (2006). The Wealth of Networks. Yale University Press.

Berg, F., Kölbel, J. F., & Rigobon, R. (2022). Aggregate Confusion: The Divergence of ESG Ratings. Review of Finance, 26(6), 1315–1344. https://doi.org/10.1093/rof/rfac033

Bresnahan, T. F., & Trajtenberg, M. (1995). General Purpose Technologies: ‘Engines of Growth’? Journal of Econometrics, 65(1), 83–108.

Briscoe, B., Odlyzko, A., & Tilly, B. (2006). Metcalfe’s Law is Wrong. IEEE Spectrum, 43(7), 34–39. DOI: 10.1109/MSPEC.2006.1653003.

Brynjolfsson, E., & Hitt, L. M. (2000). Beyond Computation: Information Technology, Organizational Transformation and Business Performance. Journal of Economic Perspectives, 14(4), 23–48.

Central Committee of the CPC and the State Council (2022). Opinions on Building Basic Data Systems to Better Leverage the Role of Data Elements (“Data Twenty Articles”).

Coase, R. H. (1937). The Nature of the Firm. Economica, 4(16), 386–405.

Coase, R. H. (1960). The Problem of Social Cost. Journal of Law and Economics, 3, 1–44.

Cooper, R., & Kaplan, R. S. (1988). Measure Costs Right: Make the Right Decisions. Harvard Business Review, 66(5), 96–103.

Devereux, M. P., & Vella, J. (2017). Implications of Digitalization for International Corporate Tax Reform. In S. Gupta, M. Keen, A. Shah & G. Verdier (Eds.), Digital Revolutions in Public Finance (pp. 91–112). International Monetary Fund.

Directive (EU) 2022/2464 (Corporate Sustainability Reporting Directive, CSRD).

Directive (EU) 2024/2831 of 23 October 2024 on Improving Working Conditions in Platform Work, OJ L, 11 November 2024. https://eur-lex.europa.eu/legal-content/EN/TXT/PDF/?uri=OJ:L_202402831 (accessed 20 September 2026).

ECJ, Case C-465/20 P, Commission v Ireland and Others, Judgment of 10 September 2024 (Apple).

G7 (2025). G7 Finance Ministers and Central Bank Governors’ Statement on Global Minimum Taxes, 28 June 2025.

Hirshleifer, J. (1956). On the Economics of Transfer Pricing. Journal of Business, 29(3), 172–184.

ILO (2021). World Employment and Social Outlook: The Role of Digital Labour Platforms in Transforming the World of Work. Geneva: International Labour Office.

Jones, C. I., & Tonetti, C. (2020). Nonrivalry and the Economics of Data. American Economic Review, 110(9), 2819–2858. https://doi.org/10.1257/aer.20191330

Kanbur, R., & Keen, M. (1993). Jeux Sans Frontières: Tax Competition and Tax Coordination When Countries Differ in Size. American Economic Review, 83(4), 877–892.

Katz, M. L., & Shapiro, C. (1985). Network Externalities, Competition, and Compatibility. American Economic Review, 75(3), 424–440.

Ministry of Finance of the PRC (2023). Interim Provisions on the Accounting Treatment of Enterprise Data Resources (Cai Kuai [2023] No. 11; effective 1 January 2024).

Musgrave, R. A. (1959). The Theory of Public Finance. McGraw-Hill.

National Data Administration et al. (2023). Three-Year Action Plan for “Data Element ×” (2024–2026).

OECD (2015). Addressing the Tax Challenges of the Digital Economy, BEPS Action 1 Final Report.

OECD (2022). OECD Transfer Pricing Guidelines for Multinational Enterprises and Tax Administrations 2022. Paris: OECD Publishing. DOI: 10.1787/0e655865-en.

OECD (2024). The Inclusive Framework Report on Amount B of Pillar One (February 2024; incorporated into the OECD Transfer Pricing Guidelines from 2025).

OECD (2026). Public Consultation Document: Revisions to Chapter VII of the OECD Transfer Pricing Guidelines — Special Considerations for Intra-group Services. Discussion draft, 1 June – 22 July 2026.

OECD (2026). Side-by-Side Package: Global Minimum Tax Administrative Guidance, 5 January 2026.

Ostrom, E. (1990). Governing the Commons. Cambridge University Press.

Pigou, A. C. (1920). The Economics of Welfare. Macmillan.

Rochet, J.-C., & Tirole, J. (2003). Platform Competition in Two-Sided Markets. Journal of the European Economic Association, 1(4), 990–1029.

Romer, P. M. (1990). Endogenous Technological Change. Journal of Political Economy, 98(5), S71–S102.

Scala, A., & Delmastro, M. (2023). The explosive value of the networks. Scientific Reports, 13, 1037. DOI: 10.1038/s41598-022-26961-x.

State Taxation Administration (2009). Guo Shui Fa [2009] No. 2, Implementation Measures for Special Tax Adjustment (Trial); Announcement No. 6 (2017), Special Tax Investigation, Adjustment and Mutual Agreement Procedure.

Thomas, A. L. (1974). The Allocation Problem: Part Two. Studies in Accounting Research No. 9. American Accounting Association.

Tørsløv, T., Wier, L., & Zucman, G. (2023). The Missing Profits of Nations. Review of Economic Studies, 90(3), 1499–1534. https://doi.org/10.1093/restud/rdac049

United Nations (2021). Model Double Taxation Convention between Developed and Developing Countries, Article 12B (Automated Digital Services).

United Nations (2021). Practical Manual on Transfer Pricing.

U.S. Treas. Reg. §1.482-9 (Allocation of Income and Deductions from Sources Within the United States: Services Cost Method).

Wilson, J. D. (1986). A Theory of Interregional Tax Competition. Journal of Urban Economics, 19(3), 296–315.

Wilson, J. D. (1999). Theories of Tax Competition. National Tax Journal, 52(2), 269–304.

Zodrow, G. R., & Mieszkowski, P. (1986). Pigou, Tiebout, Property Taxation, and the Underprovision of Local Public Goods. Journal of Urban Economics, 19(3), 356–370.

APPENDIX： Analysis Report on the Landing Points and Impact Scope of Andy Liu’s Three Recommendations in the OECD Transfer Pricing Guidelines (2022)

Purpose: to analyse, recommendation by recommendation, exactly which paragraphs of Chapter VII the three recommendations in the public comment (the author’s original submission) would amend, and to assess which sections of the whole text may be affected, thereby providing a basis for precise citation in the body text of the submission to the OECD.

Source text: the English original PDF (658 pages) of the OECD Transfer Pricing Guidelines for Multinational Enterprises and Tax Administrations 2022; the paragraph numbers and page numbers were extracted and verified through this project’s pipeline (pipeline-tpg22/). The report also references the revised Chapter VII draft released for public consultation (comment period closing 22 July 2026) and marks the corresponding draft paragraphs.

Method: each recommendation is subjected to a five-step analysis — “claim → concept mapping → landing paragraphs (primary/secondary landing point) → amendment pathway → affected chapters”; all paragraph numbers were extracted from the original text and double-checked. Paragraph conventions: the body text cites TPG 2022 paragraph numbers; where the 2026 draft revision is involved, this is expressly marked by the word “draft” (the mapping is given in Section 4).

1. Recommendation 1: Value Attribution of Modern Technology Services by Platform-Based IT Enterprises

Key points of the original recommendation: platform-based sales enterprises bear substantial IT infrastructure operating costs and elevated administrative overheads; the substantive value constituted by artificial intelligence and algorithms should not be subjected to the simplified (low mark-up) approach; it is argued that the U.S. transfer pricing cost standards should be adopted by reference (for Chinese platform-based enterprises analogous to those covered under the U.S. Pillar Two “Side-by-Side (SbS)” arrangement).

1.1 Concept Mapping

Concept in the recommendation	Corresponding institution in the Guidelines	Nature
Application of the simplified method at a reduced mark-up	Section D of Chapter VII, the simplified approach for “low value-adding intra-group services” (5% mark-up, paragraph 7.61)	Direct point of conflict
IT infrastructure operating costs	The characterisation question of whether IT services are “low value-adding” (paragraph 7.49)	Characterisation issue
Substantive value constituted by AI and algorithms	Unique and valuable intangibles → Chapter VI	Exclusion condition
Elevated administrative overheads of platform-based sales	Centralised services (paragraph 7.14); group synergies (paragraphs 1.177–1.183)	Quantitative context
U.S. cost standards	U.S. Treasury Regulations §1.482-9 (the services cost method (SCM), the U.S. source of the 5% simplified rate for low value-adding services)	Country-specific practice

1.2 Landing Paragraphs (by amendment priority)

Primary landing point: Section D.1 (definition), paragraphs 7.45–7.50

• Paragraph 7.45 (p.329) — the four-characteristic definition of low value-adding intra-group services; the direct object of amendment for Recommendation 1. Paragraph 7.45 currently requires that the services “do not require the use of unique and valuable intangibles and do not lead to the creation of unique and valuable intangibles.” Amendment pathway: insert, after the third bullet point, an additional sentence of a definitional nature making explicit that “IT services operated on a continuous basis by means of data-driven intangibles such as algorithms and artificial-intelligence models (even where they bear a supportive appearance), because they use unique and valuable intangibles, do not constitute low value-adding intra-group services.” In effect, this sentence makes explicit the exclusion logic already implicit in the third characteristic of paragraph 7.45 and names the new form of undertaking.

• Paragraph 7.49 (p.330–331) — the “information technology services” item in the list of examples of low value-adding services. The current example lists, as low value-adding, “installing, maintaining and updating IT systems used in the business; information system support (which may include the information system used in connection with accounting, production, client relations, human resources and payroll, and email systems); … supporting, maintaining and supervising of IT networks (local area network, wide area network, internet)” (draft para 7.80, same wording). Amendment pathway: add a proviso within this IT services item, echoing the drafting of the case in paragraph 7.50 b): “but the data-processing and transaction-matching systems that a platform-based enterprise operates at its core using its own proprietary algorithms and artificial-intelligence models do not fall within this item.” Rationale: paragraph 7.50 already establishes the precedent structure by which “services that may seem superficially similar in nature (in the example, credit risk analysis) may or may not be low value-adding intra-group services depending on the specific context and circumstances” (case a) low value-adding / case b) not low value-adding); platform IT services are a natural extension of that precedent structure — highly persuasive and format-consistent.

• Paragraph 7.50 (p.331–332) — should the OECD Secretariat prefer to add an example rather than amend the definition, it may be proposed that Example c) be added: platform-based Enterprise X operates a trading platform on the basis of its own proprietary recommendation algorithms, and its internal IT department provides systems support for the platform’s operations — because that support activity uses unique and valuable algorithmic intangibles and bears significant risk, it does not constitute low value-adding intra-group services. This is the least intrusive and most acceptable pathway.

Secondary landing point: Section D.2 (simplified approach mechanism), paragraphs 7.52–7.63

• Paragraph 7.61 (p.336) — the 5% uniform mark-up rate provision. The final sentence of this paragraph already carries a caution: “the low value-adding intra-group services mark-up should not, without further justification and analysis, be used as benchmark for the determination of the arm’s length price for services not within the definition of low value-adding intra-group services, nor for similar services not within the elective, simplified scheme.” Amendment pathway: append a half-sentence after that caution, naming “in particular for platform-type services whose value creation depends mainly on algorithms and data assets.” This converts the value proposition of Recommendation 1 (“substantive value should not be subjected to a mechanically reduced mark-up”) into a reinforcement of the anti-abuse provision, without disturbing the 5% itself.

• Paragraph 7.63 (p.336–337) — the threshold provision (permitting tax administrations to set financial-ratio thresholds and, where exceeded, to require a full functional analysis). Amendment pathway: add to the enumeration of threshold examples an “intangible-intensity indicator relating to platform activities,” providing tax administrations with an operational handle for requiring platform-based enterprises to step outside the simplified approach.

1.3 Affected Chapters (full-text impact scope)

Chapter	Paragraphs	Mechanism of impact	Intensity
Chapter VI Intangibles	paragraphs 6.1–6.32 (identification of intangibles, in particular 6.6, 6.30–6.32); paragraphs 6.49–6.57 (DEMPE functional contributions)	If algorithms/AI models are identified as “unique and valuable intangibles,” the functional-positioning discussion under the DEMPE framework of Chapter VI must respond to this; the standard for determining whether data-driven intangibles are “unique and valuable” must be consistent between the two chapters	★★★
Chapter II Transfer pricing methods	paragraph 2.2 (selection of the most appropriate method); paragraph 2.99 (the meaning of full cost under the cost-based TNMM)	If platform IT services are removed from the simplified approach, they fall back to the ordinary method selection in paragraph 7.31 (CUP / cost plus / cost-based TNMM); the full-cost basis in paragraph 2.99 determines how a platform’s high IT costs enter the mark-up base	★★
Chapter I The arm’s length principle	paragraphs 1.33–1.40 (accurate delineation of the actual transaction as a precondition); paragraphs 1.177–1.183 (D.8 MNE group synergies)	Part of a platform’s IT costs is a structural advantage attributable to group synergies (the three-part “deliberate concerted group action” framework in paragraphs 1.179–1.181); the cost-recognition logic asserted in Recommendation 1 must be compatible with the rule for sharing synergy benefits (paragraph 1.182, “shared in proportion to their contribution”)	★★
Chapter V Documentation	paragraphs 5.16–5.17 (three-tier structure) and the local file requirements (Annex II to Chapter V)	If platform IT services are removed from the simplified approach, the taxpayer must substantiate under the ordinary path, and the local file should add disclosure items for the holding of algorithmic intangibles and the aggregation of costs	★
Chapter X Financial transactions	(indirectly) Section D.1(1) precise delineation from paragraph 10.6	Relevant only indirectly in platform-related financing scenarios; not developed in this report	★

2. Recommendation 2: The Computation of ESG Public-Interest Investment

Key points of the original recommendation: within the conventional ESG computation environment based on the headquarters (legal entity), the accounting recognition and measurement of branch-level localised costs should be rationally taken into account.

2.1 Concept Mapping

Concept in the recommendation	Corresponding institution in the Guidelines	Nature
Headquarters-based ESG computation	The entity-level basis of transfer pricing documentation (the three-tier documentation of Chapter V: master file / local file / country-by-country report)	Basis/scope issue
Branch-level localised costs	Cost aggregation and allocation in functional analysis (cost pools and allocation keys in paragraphs 7.56–7.60; full cost in paragraph 2.99)	Accounting issue
ESG (environmental, social and governance) investment	The Guidelines contain no dedicated ESG provision; the closest are Section D.1 of Chapter I (accurate delineation) plus Chapter III comparability adjustments	Institutional gap

2.2 Landing Paragraphs

Primary landing point: Section B.2.2.2 of Chapter VII (indirect-charge methods) and Section D.2.2 (cost pools)

• Paragraphs 7.23–7.26 (p.323) — indirect-charge methods and selection of allocation keys (draft paras 7.45–7.49, substantively carried over). ESG localised costs (e.g., environmental facilities, community investment and localised labour costs at the branch’s location) are frequently recovered in group centralised accounting through indirect allocation. Amendment pathway: after the enumeration of allocation-key examples in paragraph 7.25, add: “for location-specific ESG compliance costs borne by a branch, they may be collected separately on the benefit-recipient principle and not be merged into the group’s uniform allocation key.” The key here is to liberate local ESG costs from “uniform group-level formula allocation,” corresponding directly to the recommendation’s call to “rationally take into account branch-level localised costs.”

• Paragraph 7.56 (p.334) — determination of cost pools (draft para 7.87; the “solely on behalf of one other group member” removal at draft para 7.88): “the cost pool should exclude costs that are attributable to an in-house activity that benefits solely the company performing the activity.” Amendment pathway: make explicit that, where localised ESG investment benefits the operations at the branch’s location, it should remain in the local entity’s costs rather than being aggregated upward into the headquarters’ overall ESG base — this is in the same direction as Recommendation 2’s assertion that computation “must not be based on the headquarters alone.”

Secondary landing point: Chapter III comparability analysis

• The five economically relevant characteristics are listed in paragraph 1.36 (paragraphs 1.33–1.40 set out the accurate-delineation and two-stage framework); comparability adjustments are in paragraphs 3.47–3.54 (including working capital adjustments), while paragraphs 3.55–3.63 deal with the arm’s length range and extreme results. Amendment pathway: list “location-specific ESG compliance cost differences” as a possible item of comparability adjustment (a regulatory-cost difference under the “economic circumstances” factor). This is new language, but is anchored within the existing five-factor framework and is consistent with OECD amendment practice.

2.3 Affected Chapters

Chapter	Paragraphs	Mechanism of impact	Intensity
Chapter III Comparability analysis	the five comparability factors in paragraph 1.36 (see paragraphs 1.33–1.40); comparability adjustments in paragraphs 3.47–3.54	Differences in ESG regulatory costs are brought into comparability adjustments	★★
Chapter V Documentation	master file (Annex I to Chapter V: intangibles and the overall description of the group’s business); country-by-country report (Annex III)	The call for ESG data to be disclosed by entity/jurisdiction; however, the country-by-country report template is a BEPS Action 13 international consensus and should not be proposed for amendment — only the local file could carry an ESG cost note	★★
Chapter I	paragraphs 1.33–1.40 (accurate delineation); paragraphs 1.44–1.55 (functional analysis)	The ESG functions actually assumed by a branch must enter into the functional analysis	★
Chapter X Financial transactions	from paragraph 10.189 (Section E, captive insurance)	Indirectly relevant if ESG includes green insurance / captive arrangements	★

3. Recommendation 3: Value Attribution of Purely Gig-Economy and Freelancer Platforms

Key points of the original recommendation: purely public-interest, non-profit digital platforms have emerged which do not directly generate profit but, through the provision of infrastructure, reduce aggregate societal transaction costs and generate substantial positive externalities; assessing them solely by financial-profit indicators would understate their economic contribution; exploration should be made of introducing quantified social-benefit indicators (user scale, activity levels, employment effects) so as to build a multi-dimensional value-and-cost attribution system and avoid unduly suppressing emerging public-interest forms of the digital economy.

3.1 Concept Mapping

Concept in the recommendation	Corresponding institution in the Guidelines	Nature
Non-profit platforms generate no profit	paragraphs 7.35–7.36 (whether the service charge must include a profit element)	Direct point of engagement
Reduction of aggregate societal transaction costs	the “economic or commercial value” standard in the benefits test in paragraph 7.6	A claim to extend the standard
Positive externalities / social benefit	group synergies in paragraphs 1.177–1.183; incidental benefits in paragraphs 7.12–7.13	Partial correspondence
Quantification of user scale and activity levels	allocation keys for intra-group services (the staff/user/volume-type keys in paragraph 7.59)	Methodological tool
Multi-dimensional value attribution	Section D.1 of Chapter I (accurate delineation) plus the overarching value creation principle (the Preface and paragraphs 1.1–1.5)	Principle level

3.2 Landing Paragraphs

Primary landing point 1: Section B.2.3.2 (the profit element), paragraphs 7.35–7.37

• Paragraph 7.36 (p.326) — already acknowledges that “it may be the case that the market value of intra-group services is not greater than the costs incurred by the service provider” and that “an MNE group may still determine to provide the service intra-group rather than using a third party for a variety of reasons, perhaps because of other intra-group benefits (for which arm’s length compensation may be appropriate)” (draft paras 7.68–7.69, substantively unchanged). Amendment pathway: append at the end of paragraph 7.36: “in evaluating the relationship between cost and value, for platform infrastructure that is provided for public-interest purposes and does not pursue commercial return, the benefit assessment may have regard to non-financial dimensions such as user scale, activity levels and employment effects, but such dimensions do not alter the application of the benefits test itself.” That is, first secure the institutional boundary that “the benefits test still applies” (non-negotiable for the OECD), and then bring social-benefit indicators into the benefit assessment as an evidentiary dimension. This is consistent with the proposition in the theoretical-support paper that “social value is evidence of externalities, and externalities are limited to the evidentiary dimension”.

• Paragraph 7.37 (p.327) — the tax administration’s practical discretion (a cost-benefit analysis may exceptionally accept a cost-reimbursement model). Amendment pathway: taking paragraph 7.37 as the landing point, assert that for service charges of non-profit public-interest platforms a tax administration may, within its discretion, accept a cost-reimbursement (no mark-up) arrangement — this is fully compatible with the existing flexibility in paragraph 7.37, requires no new rule, and merely makes explicit that public-interest platforms fall within the reach of that flexibility.

Primary landing point 2: Section B.1.1 (the benefits test), paragraphs 7.6–7.8

• Paragraph 7.6 (p.316–317) — the standard sentence of the benefits test. It is not recommended to amend the “economic or commercial value” standard itself (it would undermine the foundation of Chapter VII, and the OECD would not accept “social value” entering the standard). Amendment pathway: within the framework of paragraph 7.7 (“it is not possible in the abstract to set forth categorically the activities that do or do not constitute the rendering of intra-group services. However, some guidance may be given to elucidate how the analysis would be applied for some common types of services undertaken in MNE groups”), recommend adding a public-interest platform case to the examples in Section C: a non-profit professional information platform provides employment-matching infrastructure to group members free of charge, its costs being shared among the group members that use it — under the “whether an independent enterprise would have been willing to pay” test in paragraph 7.6 (from the functional-substitution perspective: the using members, were they independent, would have had to bear the job-search infrastructure costs themselves), an intra-group service may be found to exist; the degree of benefit may have regard to the number of active users and the volume of successful matches. That is, do not amend the standard, amend the examples, so that the benefits test naturally accommodates public-interest platforms.

Secondary landing point: Section D.2.3 (allocation keys), paragraphs 7.59–7.60

• Paragraph 7.59 already lists keys such as “the share of total users” (IT services). Amendment pathway: add a sentence to the examples along the lines of “for platform services driven by user value, indicators such as active user numbers and transaction-matching volumes, which reasonably reflect the degree of benefit, may be used as an allocation key or as a component of an allocation key,” putting the quantitative claim of Recommendation 3 into the technical layer of allocation, with zero institutional conflict.

3.3 Affected Chapters

Chapter	Paragraphs	Mechanism of impact	Intensity
Chapter I	paragraphs 1.177–1.183 (D.8 MNE group synergies)	If the “positive externalities” of a public-interest platform are characterised as a structural advantage arising from deliberate concerted group action, they must be handled under the three-part framework of paragraph 1.181 (nature / amount / allocation) and must remain consistent with the evidentiary-dimension claim in Chapter VII	★★★
Chapter I	paragraphs 1.1–1.5 (the overarching value creation principle); the Preface	If “multi-dimensional value attribution” is written in, the Preface’s value-creation statement would acquire a “non-financial dimension” footnote — the OECD has consistently been conservative at the principle level, so this should be presented as an interpretive paragraph rather than an amendment to the principle	★
Chapter II	the TNMM in paragraphs 2.58–2.113 and the profit split method from paragraph 2.114	If a public-interest platform uses profit splitting, its “profit” is near zero and the split method fails; in practice it falls back to cost plus / cost reimbursement, so the impact scope is limited	★
Chapter VI Intangibles	paragraphs 6.34–6.47 (legal ownership) and 6.48–6.58 (contributions by functions, assets and risks)	The attribution of data/reputational intangibles arising from a platform’s user network effects is adjacent to Recommendation 3 but not claimed by it; recorded as an associated observation item	★

4. Summary of Amendment Pathways for the Three Recommendations (directly usable in the OECD submission)

Recommendation	Highest-priority landing point (TPG 2022)	Corresponding 2026 draft paragraphs	Minimal-intrusion amendment	Alternative landing point	Key affected chapters
1: Platform IT value attribution	paragraph 7.45 (definition) + paragraph 7.49 (IT example item)	draft 7.76, 7.80–7.81	add platform-algorithm Example c) at paragraph 7.50	the anti-benchmark caution in paragraph 7.61; the threshold in paragraph 7.63	Chapter VI (DEMPE), Chapter II (2.99), Chapter I (1.177–1.183)
2: Localised ESG costs	paragraph 7.25 (allocation key) + paragraph 7.56 (cost-pool exclusion)	draft 7.45–7.49 (indirect charging and keys), 7.87–7.91 (cost pools and allocation)	add separate collection of location-specific ESG costs at paragraph 7.25	add a regulatory-cost-difference adjustment under the five factors of Chapter III	Chapter III (comparability adjustment), Chapter V (documentation)
3: Value of public-interest platforms	paragraph 7.36 (value ≥ cost case) + paragraph 7.37 (cost-reimbursement discretion)	draft 7.13, 7.68–7.70, 7.90–7.91	add a non-profit platform example in Section C (the example extension authorised by paragraph 7.7)	add a user-scale-type allocation key at paragraph 7.59	Chapter I D.8 (the three-part synergy framework), the Preface (value creation)

5. Consistency Notes (joint drafting constraints for the three recommendations)

1. Do not touch the 5% mark-up rate (paragraph 7.61) or the standard sentence of the benefits test (paragraph 7.6) themselves — both are products of the international consensus that followed the BEPS Actions 8–10 of 2017, and any proposal to amend them would be read as reopening a closed issue; all three recommendations can be achieved through “definitional clarification + example extension + anti-abuse reinforcement,” without frontal engagement.

2. The benchmark against U.S. §1.482-9 (Recommendation 1) should be placed in the “rationale/commentary” part; the body-text landing points should still refer to OECD paragraph numbers; the OECD Guidelines do not accept direct incorporation of a country’s domestic-law provision, but its institutional effect may be invoked as comparative reasoning.

3. The SBS (Side-by-Side arrangement) reference (Recommendation 1): this falls within the Pillar Two legislative template and is not part of the TPG body text; in the submission it should appear as a footnote or in a rationale paragraph, to avoid being read as a request to amend the TPG body text.

4. The “quantification of social benefit” for public-interest platforms (Recommendation 3): always express it as “an evidentiary dimension of the benefit assessment under the benefits test,” keeping it fully consistent with the proposition in the theoretical-support paper that “social value is evidence of externalities, and externalities are limited to the evidentiary dimension,” so as not to be read as a request to amend the arm’s length principle.

5. Localised ESG costs (Recommendation 2): align with the “activity characterisation × accounting level” two-dimensional framework of the theoretical-support paper — composition decomposition (compliance / investor requirements / commercial operations / charitable input) is completed along the activity-characterisation dimension, and localised cost recognition along the accounting-level dimension; the two landing points (allocation keys at 7.25; cost pools at 7.56) correspond respectively to the two boundaries of “no expansion of the mark-up base” and “cost recognition and pooling”.

Appendix: Verification Checklist for the Paragraph References Cited in this Report

Paragraph number	PDF page	Method of verification
7.6–7.65 (the whole of Chapter VII)	p.315–338	segment-by-segment extraction from blocks.jsonl + 65/65 numbering completeness check (no references to 7.66+ in the whole text)
1.177–1.184	p.88–90	checked against the extracted original text (group synergies, D.8)
2.2 / 2.12 / 2.99	p.95 / 98 / 125	checked against the extracted original text (method selection / multiple methods / full cost)
5.24–5.26 (country-by-country report)	p.235	checked against the extracted original text
7.49 / 7.50 truncated-text recovery	p.330–332	restored word-for-word against the original paginated text (has been incorporated into the translation)

End of report