📏 The Arm's-Length Principle

When two parts of the same company trade with each other, they could pick any price they like and use it to park profit wherever tax is lowest. To stop that, tax authorities across the world apply a single rule: price the deal as if the two sides were independent strangers. The phrase itself carries the idea: you keep someone "at arm's length" when you hold them far enough away that neither side can lean on the other, so each one bargains as an independent party looking out for itself. That is the arm's-length principle, and this page shows how a transfer price is actually built from it.

Why a transfer price even exists

A multinational is a single company to its owners, but to the tax office it is many separate taxpayers: a legal entity in each country, each filing its own return and paying tax on its own profit. So the moment anything of value (goods, a service, a licence, a loan) passes from one of these entities to another, that internal transfer needs a price. The price matters because it decides how much profit is booked in each country, and therefore how much tax gets paid where.

Here is the catch. Because that internal price can be set almost anywhere, it quietly decides where the profit shows up. Price the factory's goods high and the profit stays in the factory's country; price them low and the profit moves to the sales company's country. Now suppose the factory's country taxes profit at 10% and the sales country taxes it at 30%. The group can lower its total tax bill just by setting a high internal price, so that most of the profit lands in the 10% country, even though nothing about the real business has changed. That is the temptation: shifting profit on paper into whichever country taxes it the least. To stop it, the price cannot be left to the group's own choosing. It has to match the price that two unrelated, independent companies would have agreed for the same deal. That benchmark is the arm's-length principle, written into the OECD Transfer Pricing Guidelines and into national law almost everywhere.

The five methods for finding an arm's-length price

The OECD recognises five methods. In practice you apply just one: the "most appropriate method" for the transaction. Which one fits is driven mainly by the tested party, the entity with the simpler, more routine role, and by where the best real-world comparables can be found. There is no fixed hierarchy, and no rule to run several methods and split the difference.

CUP

Comparable Uncontrolled Price

Find the price the same or similar product sells for between independent parties, and use it directly.

Best when a genuine market price exists (commodities, listed prices). The most direct, and the most demanding on comparability.
Resale Price

Resale Price Method (resale-minus)

Start from the price the reseller charges the end customer and subtract an arm's-length gross margin for the reseller. What's left is the transfer price.

Best for distributors that buy and resell without adding much value.
Cost Plus

Cost Plus Method

Take the supplier's cost and add an arm's-length markup for the function it performs. This is the markup on the transfer page.

Best for manufacturers and service providers, especially contract or toll makers.
TNMM

Transactional Net Margin Method

Compare a net profit indicator (e.g. operating margin, or return on costs) against comparable independent companies, and back out the price that yields it.

The workhorse in practice: tolerant of product differences, so it's used when CUP/resale/cost-plus can't find close enough comparables.
Profit Split

Profit Split Method

Add up the combined profit from the transaction and divide it between the entities by their relative contributions.

Best when both sides bring unique, valuable things (shared IP, highly integrated operations) and neither is a simple benchmarkable party.
In practice

How a price is chosen

Characterise each entity, pick the tested party and the method that fits it, run a benchmarking study to find comparables, and land inside the arm's-length range those comparables produce. Then document it all.

That range is the spread of comparable companies under one method, not the gap between two different methods stacked against each other.

Worked example: cost-plus vs. resale-minus

In a real analysis you normally apply just one of these methods: the one that fits the tested party, meaning the entity with the simpler, more routine functions. Cost-plus is used when the manufacturer is that routine party; resale-minus when the distributor is. Each method is benchmarked against its own set of comparable companies, so there is no rule that the two must agree. Showing both side by side here is a consistency check, not an OECD procedure: the customer price minus the maker's cost is a fixed pool of gross profit, and the two methods are simply two ways of splitting it. Enter the numbers and see where each one lands.

Markup on cost the maker earns.
Arm's-length margin the distributor keeps.

Both boxes below are the same thing: the transfer price the reseller pays the maker. They just reach it from opposite ends. Cost-plus builds it up from the maker's cost; resale-minus works it back down from the customer's price.

Cost-plus method
–
transfer price the reseller pays the maker
Up from cost: C × (1 + markup)
–
Resale-minus method
–
transfer price the reseller pays the maker
Down from price: SP × (1 − reseller margin)
–
Gap between the two methods: – A consistency check, not an OECD rule: each method is benchmarked separately, so the two need not match. A large gap flags that the assumed markup and margin aren't both realistic given the actual cost and market price.

From principle to a filed number

The workflow a transfer-pricing team actually follows:

  1. Characterise the entities. Who does what, owns what, and bears which risks? A contract manufacturer and a full-fledged principal earn very different returns.
  2. Pick the most appropriate method for the transaction and that characterisation.
  3. Run a benchmarking study: search databases for independent companies or deals that are comparable, and compute their margins.
  4. Set a price inside the arm's-length range those comparables produce (often the interquartile range).
  5. Document it in a transfer-pricing file (local file, master file, and, for large groups, country-by-country reporting), ready for any tax audit.

Get it wrong and the consequences are real: a tax authority can re-price the transaction, tax the shifted profit, and add penalties and interest, sometimes with the other country still taxing the same profit too (double taxation). That's why the markup on the intercompany transfer page is never just a number someone picked. It has to trace back to this principle.

Educational overview only, not tax advice. Real transfer-pricing analysis is jurisdiction-specific and needs a qualified adviser.