I, like most who have played around in this field, think that the arm's length standard was always the ultimate guide and that any specific rules as to methodology in the 482 regulations were simply ways of administering the arm's length standard in specific contexts. After all, the Regulations specifically state that "the standard to be applied in every case" is the arm's length standard. § 1.482-1(a)(b)(1). Yet, the Xilinx Court held that, although the arm's length standard may be the general benchmark for valid 482 adjustments, the IRS can by regulation change that benchmark even if it achieves a demonstrably non-arm's length result.
I do not doubt that the IRS has a great deal of authority to promulgate regulations that, under Chevron (not mentioned by the majority opinion but clearly in the background), will be the law so long as reasonable even when not commanded by the plain meaning of the statute being interpreted. So, you might ask, does the statute as interpreted not require the arm's length standard? The statute itself does no mention the arm's length standard. The statute merely says that the IRS may make the adjustment when it is "necessary in order to prevent evasion of taxes or clearly to reflect the income of any of such organizations, trades, or businesses." Evasion of taxes is a term of art in the tax area, generally meaning the willful intent to violate a known legal duty applicable in order to convict under Section 7201. But, evasion as a separate concept has no discernible meaning in § 482, so the courts have focused on the "clearly reflect income" standard in the text. What does clearly reflect income mean? It too is not self defining in this context, but the overwhelming body of law, including the IRS's own regulations, says it means the income that would be reflected in a transaction between parties dealing at arm's length -- the arm's length standard.
You might ask about the legislative history. It is sparse and unhelpful given that it was enacted in 1928 with early revenue statute predicates, a time when Congress was less wordy. I cover the sparse legislative history in my earlier article John A. Townsend, Reconciling Section 482 and the Nonrecognition Provisions, 50 Tax Lawyer 701, 702-705 (1997). Until Xilinx, the arm's length standard has become the standard for resolving transfer pricing issues where it really matters -- in cross-border contexts where our treaties uniformly adopt some variation of language that is interpreted to mean that the arm's length standard applies. (The Court discusses the treaty context, and I have discussed in more detail in John A. Townsend, Tax Treaty Interpretation, 55 Tax Law. 219 (2001).)
Readers will forgive me if I digress for a moment on the arm's length standard. Back in the old (real old) days while I was with DOJ Tax Appellate I briefed and argued Liberty Loan Corp. v. United States, 498 F.2d 225 (8th Cir. 1973). In that case, the taxpayer was a parent corporation with numerous consumer finance subsidiaries. The parent-taxpayer borrowed at 5.55% and re-lent to its subsidiaries much, if not most, of the cash the subsidiaries needed to lend in their consumer finance businesses. Each of the loans to the subsidiaries was reflected in separate lending documents. The taxpayer lent its profitable subsidiaries at a 5.75% rate and its unprofitable subsidiaries at a 0% rate. These two rate structures produced overall interest income to the parent-taxpayer at its cost of borrowing rate -- 5.5%, so the parent-taxpayer had neither income or loss at its level. The parties stipulated that, had the subsidiaries borrowed from a third party unrelated lender, they would each have paid a rate exceeding 5.75%. Exercising its authority under § 482 to adjust the 0% interest rate loans to 5% (a safe harbor rate) without adjusting the 5.75% interest rate loans. The 5.75% interest rate loans were not adjusted because they were within the "safe harbor" provided in the regulations. The parent-taxpayer objected, paid the tax and sued for refund. the parent-taxpayer won at the trial level. The district court treated the subsidiaries as a group borrower who, as a group but not as individual corporations, could borrow at the 5.5% rate, thus in effect making the transaction between the parent-taxpayer and the "group" at arm's length. The Government appealed to the Eight Circuit. The Government argued, that the no-interest loans could be adjusted to the safe harbor rate because that was an adjustment toward the actual arm's length rate (exceeding 5.75%) but could not adjust the 5.75% rate downward to mitigate the effect of the first adjustment because a downward adjustment of the 5.75% rate would impermissibly move the interest rate away from the arm's length rate stipulated to exceed 5.75%. The Government won the appeal. The Court of Appeals in Liberty Loan held:
Thus, it would have been contrary to the regulations for the Commissioner to balance out the transactions at 5% (or 5.55%) when the stipulated arm's length rate was in excess of 5.75%. This limitation on the offset provision logically follows when one remembers that the purpose of a § 482 adjustment is to more clearly reflect income. The upward adjustments to the no-interest loans do just that. Downward adjustments to the 5.75% loans, however, would have the effect of moving those loans even further away from the actual arm's length rate. In adjusting the no-interest loans upwards to 5%, the Commissioner therefore made the only adjustment he could.In Xilinx, the court did exactly what the Government argued was impermissible in Liberty Loan -- i.e. it forced the taxpayer under the guise of the regulations to move the stipulated arm's length transfer price away from the arm's length standard. It is true that the court did that under what it perceived as authority of the IRS to require § 482 adjustments by regulation under Chevron-type notions (although the court does not refer to Chevron).
I think the dissent got this right.