A controlled party’s reported operating profit is the residual of the intercompany prices under audit examination. Where transfer prices have been set to move income across borders, the reported figure is contaminated. What follows develops an internal benchmark for each party’s arm’s-length operating profit that is invariant to where profit happens to be booked, and that rests on a single parameter computed from data the tax authority already holds from corporate tax filings.
1. Notation
Let j = 1, …, N index the controlled parties. Write P(j) for the reported operating profit of party j, and let the unsubscripted P = Σ P(j) denote the combined operating profit. Similarly let W(j) be wages and salaries (payroll) of party j, with W = Σ W(j). Define the payroll share:
\omega(j)=\frac{W(j)}{W},\qquad\sum \omega(j)=1The shares are non-negative and sum to unity by construction—they form a partition of total payroll.
2. The proposed allocation
Reattribute the combined operating profit to each party in proportion to its payroll share:
\hat{P}(j)=\omega(j)\cdot P=\left(\frac{W(j)}{W}\right)\cdot PThe adjusted figure P̂(j) replaces the reported P(j) with an operating profit determined by the party’s relative compensated labor rather than by the prices it was charged. Three properties follow from the definition alone.
3. Three algebraic properties
3.1 The budget identity (exhaustiveness)
Summing the allocation over all parties:
\sum \hat{P}(j)=\sum \omega(j)\cdot P=P\sum \omega(j)=P\cdot 1=PThe allocation conserves the combined profit exactly: it is a redistribution that creates and destroys nothing. The economic consequence is that across the affected jurisdictions the aggregate tax base is unchanged. The method moves base toward economic activity; it does not expand it. This zero-sum conservation is the defining feature of a profit split as opposed to a one-sided markup.
3.2 An equalized return on payroll
Let ρ = P / W be the systemwide operating profit per dollar of payroll. Then
\begin{aligned}\hat{P}(j)&=\omega(j)\cdot P\\&=\left(\frac{W(j)}{W}\right)\cdot P\\&=\left(\frac{P}{W}\right)\cdot W(j)\\&=\rho \cdot W(j)\end{aligned}The allocation imposes a single common return on all labor compensation across all parties. It collapses the contaminated cross-section of reported returns { P(j)/W(j) } to one number, ρ. Whatever profit-shifting noise lived in the individual ratios is, by construction, removed; every party is remunerated at the group’s own rate of operating profit on payroll.
3.3 Proportionality through the origin
Property 3.2 says P̂(j) is exactly linear in W(j) through the origin: a party with twice the payroll receives twice the profit. The implicit behavioral statement is that operating profit is generated in strict proportion to compensated labor, with nothing attributed to factors that payroll does not measure. This is a deliberate specification choice; ¶4 sets out why it is the right one in the setting where the method is used.
4. The zero-intercept specification and why it is adopted
The proportional form of ¶3.2 corresponds to writing the profit–payroll relation as a regression through the origin:
P(j)=b\cdot W(j)+u(j),\qquad\text{with the intercept set to zero}The restriction is adopted because in the settings where this method is used, the data will not support estimation of a separate intercept. A tax authority typically holds about three years of data per controlled entity. A per-entity relation:
P(j,t)=a(j)+b(j)\cdot W(j,t)+u(j,t),\qquad t=1,2,3,\dots
carries two parameters against three observations—one residual degree of freedom—so neither the intercept nor its standard error can be estimated with any reliability. Pooling the reported P(j) across entities to recover a common intercept fares no better: the reported profits are the contaminated figures the method is designed to neutralize, so any intercept fitted to them re-imports the profit-shifting distortion through the back door.
Setting the intercept to zero dispenses with the problem. The proportional specification requires a single systemwide parameter, ρ = P / W, which is computed exactly from the combined totals rather than estimated from short, noisy, contaminated entity panels. The method is therefore exactly identified by the aggregates it already uses, and it remunerates every party’s labor at one common rate. The zero intercept is an assumption, openly stated: it says that, at arm’s length, operating profit is earned in proportion to compensated labor. Where long and clean entity-level panels exist the restriction could be relaxed and tested, but that is not the case for which the method is built.
5. Support in the OECD (2022) Guidelines
In OECD terms the proposal is a one-factor contribution-analysis profit split with employee compensation as the splitting factor.1 The relevant anchors in Chapter II, Section C are:
- ¶2.114 and ¶2.166—the combined profit must be split on an economically valid basis that reflects the parties’ relative contributions and approximates the division that would have obtained at arm’s length. The exhaustiveness identity of ¶3.1 (Σ P̂(j) = P) is the conservation a split presupposes.
- ¶2.169—the chosen splitting factor(s) should reflect the key contributions to value identified in the functional analysis. Where the parties’ value is created principally through people functions, employee compensation is a faithful proxy for that contribution, and a payroll key is the natural one-factor expression of it.
- ¶2.166–¶2.167—splitting factors should rest on objective, verifiable data and, where possible, be supported by comparables. Payroll taken from statutory accounts and from the Master and Local Files satisfies the verifiability criterion squarely, and—unlike reported profit—is not itself an instrument of profit shifting.
- ¶2.172—the Guidelines expressly list employee compensation (and, in the 2022 revision, headcount) among the profit-splitting factors that may be appropriate. This is direct textual support for a payroll key where people functions drive value.
- ¶2.176—where cost-based factors are drawn from profit-and-loss accounts, transactional accounts may be needed to isolate the expenses relevant to the controlled transaction. In our terms: W(j) should be the payroll attributable to the relevant functions, not indiscriminate group payroll.
- ¶2.181—”A profit splitting factor based on expenses may be appropriate where it is possible to identify a strong correlation between relative expenses incurred and relative value contributed. For example, marketing expenses may be an appropriate factor for distributors-marketers if advertising generates unique and valuable marketing intangibles, e.g. in consumer goods where the value of marketing intangibles is affected by advertising. Research and development expenses may be suitable for manufacturers if they relate to the development of unique and valuable intangibles such as patents. … Employee remuneration may be relevant in situations where functions relating to the skills and experience of staff are the primary factor in generating the relevant profits.” Emphases added.
6. Conclusion
The payroll-weighted adjusted operating profit P̂(j) = ω(j) · P is internally consistent, a budget-balanced reattribution of combined operating profit that neutralizes where profit is booked and remunerates labor at the single system-wide rate: ρ = P / W. This lone parameter is computed directly from the combined totals, so the method needs neither long entity-level panels nor the reported, contaminated entity profits to operate—an essential property given that a tax authority typically holds only about three years of data per controlled party. Under the explicit and parsimonious assumption of a zero intercept, the payroll key is identified, verifiable, and supported by the OECD profit-split factors of ¶2.172 and ¶2.181.