Introduction
Another post traced the ownership of transfer pricing data from Disclosure, Inc. in Bethesda through Primark and Thomson to Ryan in Dallas. That was a history of corporations. This one is a history of a practice, and it is personal: how the comparable company entered transfer pricing examination in the United States, told by someone who was in the room. Where the account below rests on documents, the documents are cited. Where it rests on memory, it is marked as memory.
What Came Before
When I arrived at the Internal Revenue Service (IRS) from academia in the autumn of 1988, comparables were not used in the international examination division. The instrument of choice was the industry median published by Robert Morris Associates (RMA), an association of bank lending officers founded in 1914 and later renamed the Risk Management Association. Its Annual Statement Studies compiled ratios from the financial statements that commercial banks collected from their borrowers and loan applicants, arranged by industry and reported in medians and quartiles.
The logic of using it was straightforward and, on its own terms, not unreasonable. If a tested party earned a gross profit margin during the audit year less than the median firm in its industry, something required explanation. What the Annual Statement Studies could not supply was any individual company against which the tested party might be measured. The unit of analysis was the industry aggregate. No comparable was ever identified, so no comparable could ever be examined, adjusted, or rejected.
The Annual Statement Studies are still published. Robert Morris Associates and the Bank Administration Institute merged in November 2024 to form ProSight Financial Association, which continues the series.
Three Trials
Three cases changed how I thought about the problem, and they taught opposite halves of the same lesson.
The first was E.I. du Pont de Nemours & Co. v. United States, decided by the Court of Claims in 1979, the taxpayer had established a Swiss marketing subsidiary, Du Pont International S.A. (DISA), and divided the profit on foreign sales through the price charged to it. The Commissioner found the division economically unrealistic and reallocated. Under the resale price method (RPM), the prices charged were to be compared with those of comparable uncontrolled distributors, and Du Pont offered a group of distributors it had selected itself, without a showing that they were similar in products, in functions, or in geographic and marketing circumstances. The court was unpersuaded, and judgment was entered for the United States.
Note what the court did not require. It reviewed the reasonableness of the Commissioner’s result rather than the details of the examining agent’s methodology and held the amount within a zone of reasonableness. The lesson was that comparables decide these cases, and that a comparable asserted is not a comparable held in appeals or court.
The second was Westreco, Inc. v. Commissioner, decided by the Tax Court in September 1992, concerned a research subsidiary of Nestec, the Swiss research and technical assistance arm of Nestle. Westreco operated a pilot plant and laboratory at Marysville, Ohio, and from 1981 a second location at New Milford, Connecticut, and was compensated by Nestec on a cost-plus basis under contract.
The Commissioner determined deficiencies by applying a salary multiplier of three to Westreco’s total reported salaries, a method that Walter Trepashko, the IRS examining engineer, had carried over from large engineering and construction firms. At trial the IRS offered no evidence in support of it and relied instead on comparables produced by her outside economics expert witnesses, Dennis Carlton and Richard Leftwich.
Because a taxpayer contesting a section 482 allocation carries the burden of showing an abuse of discretion, the court allowed it to examine the agent on his method, an inquiry ordinarily foreclosed as looking behind the notice of deficiency. That ruling matters as much as the outcome. It made the government’s method examinable.
The experts’ method was to take the business code from Westreco’s return, code 7389 for business services other than advertising, translate it to the nearest SIC codes, 8711 for engineering services and 8731 for commercial physical and biological research, and select fifteen companies bearing those codes from the Standard & Poor’s Compustat database. The court examined what that search produced. The fifteen included an architectural and planning firm, a nuclear power plant services company, an aerial survey and mapping business, a tire and packaging research firm, and a company engaged in the importation and breeding of research primates and canines. The court said it could not see how any of these compared with food research. A footnote records that the report’s own account of its sample, twelve companies from 8711 and three from 8731, did not match its actual composition of eleven engineering firms and four research firms. The sample had been described without being read.
Judge Goffe then conducted the decisive exchange himself, asking from the bench whether the sample of fifteen was strictly statistical, based on nothing beyond the SIC code. The witness agreed that it was. Counsel had already established that, beyond reading the Forms 10-K, no inquiry had been made into the unique risks each company faced, the intangible assets each owned, the level of trade at which each operated, or the identity of their customers. Decision was entered for the taxpayer.
Du Pont had shown that a handpicked set of comparables will not survive. Westreco showed that a machine-picked set will not survive either. A defensible search has to live between the two, and the only thing that gets it there is the reading of individual companies.
The third case was Nissho Iwai American Corp. v. Commissioner, T.C. Memo. 1985-578, 50 T.C.M. (CCH) 1483, Docket No. 1702-83, Korner, J., filed November 26, 1985.
The TC held against the Commissioner: Industry averages are, in most cases, of uncertain reliability in determining arm’s-length prices — citing Du Pont at 452 and Edwards, 67 T.C. 224, 236-237. The next sentence adds that the regulations impose no obligation to grope after a figure drawn from general statistics.
The facts are made for my argument. The Commissioner priced a Japanese trading company’s log sales to its parent using composite price analyses compiled by the Industrial Forestry Association and republished by the Forest Service. The court rejected it on two grounds the association did not disclose which members reported, so it was impossible to determine whether the transactions were uncontrolled sales at all, and the compiled range mixed delivery terms that could not be adjusted for. That is the anonymity defect of the industry aggregate, stated as a holding — the same defect as RMA, three years before I arrived at IRS and it was the government relying on the industry statistic.
Mueller, and a Computer Lab
The empirical apparatus came from elsewhere. Dennis Mueller published Profits in the Long Run with Cambridge University Press in 1986, an econometric study asking whether differences in profitability across firms persist, and if so why. He answered it using the thousand largest U.S. manufacturing firms across 1950 to 1972, and he answered it with Compustat.
That book showed me two things at once: that corporations-level profitability data existed in a form that would support systematic analysis, and that persistent differences in profitability across firms were an established empirical regularity rather than an anomaly requiring correction. The second point has consequences for transfer pricing that the profession has still not absorbed. A tested party earning less than the median is not, on that fact alone, mispriced.
I began working with the Compustat tapes at the City University of New York (CUNY) computer lab, using the account of my colleague at Brooklyn College, David Laibman. This is recollection, not record.
The First Subscription
Combining the three trials with Mueller’s empirical method, I started using Compustat to search for comparables in examination work. The first subscription came through litigation support. Assisting Beth Williams, a special trial attorney (STA) on the Yamaha matter, I proposed the expenditure and she approved it out of the case budget. One case, one subscription.
I did comparable work in the contemporaneous National Semiconductor matter, assisting the special trial attorney William Bonano. Both cases were docketed in the Tax Court. Yamaha settled in pretrial negotiation and produced no opinion. National Semiconductor was tried and decided in 1994, and the Commissioner acquiesced in the result.
I then argued that every IRS economist working in transfer pricing should have access to the Compustat monthly CD-ROM database. The answer was procedural: submit three competitive bids.
Three Bids
The three candidates were Compustat, Compact Disclosure, and Moody’s Industrials, all of them CD-ROM products at the time. I met with representatives of each and recommended Compustat, for three reasons.
First, coverage. The U.S. public company universe was sufficient for the searches an examination economist needed to run.
Second, normalization. Compustat did not merely reproduce what companies reported; it applied consistent definitions across companies and across years, so that an operating margin computed for one company meant a similar account as an operating margin computed for another. Anyone who has tried to build a comparable set out of as-reported figures knows what this is worth. The alternative is to spend the analysis reconciling line items instead of examining companies.
Third, technical support. This was the decisive criterion, and it was decided by the vendors themselves. Compustat sent people who could answer questions about how a field was constructed; a substantial part of their support staff held the CFA accreditation. The other two sent marketing personnel who could not. A database whose own staff cannot explain the derivation of a variable is not a database an economist can testify from.
The procurement record is not in my possession and this account is recollection.
What the Episode Was About
It would be possible to read the foregoing as a story about buying software. It is not. Three things were settled in those few years, and they have governed transfer pricing since my initial interventions.
The unit of analysis moved from the industry to the individual entity. The Annual Statement Studies could tell an examiner what an industry looked like. Compustat could tell that examiner what a comparable company with respect to major and minor economic activities looked like. Once the individual comparable became visible, it became contestable, and every subsequent development in the field — the comparable profits method, the best method rule, the arm’s length range — presupposes a contestable individual comparable.
The standard of selection was fixed by litigation rather than by regulation. No provision of the Code or the regulations told anyone that an SIC code is insufficient. A Tax Court judge did, from the bench, in 1992, by asking a witness a question he could not answer well. The functional analysis of economic activities that every documentation report now performs is, in its origin, an answer to that Westreco question.
And the criteria for choosing a data source were, and remain, unglamorous. Coverage, consistent normalization, and competent staff who understand their own definitions. Thirty-five years on, those are still the three, and they are still the ones that separate a database an economist can defend from one he or she cannot. My repudiation of private company comparables is because the entities are not accompanied by auditable annual filings.
Sources
Westreco, Inc. v. Commissioner, T.C. Memo. 1992-561, 64 T.C.M. (CCH) 849 (September 23, 1992), Docket No. 24078-88; prior procedural opinion at T.C. Memo. 1990-501. Full text available at CourtListener:
www.courtlistener.com/opinion/4821737/westreco-inc-v-commissioner/. The comparables discussion, the list of the fifteen selected companies, the colloquy from the bench, and the footnote on the composition of the sample appear in the Opinion.
E.I. du Pont de Nemours & Co. v. United States, 221 Ct. Cl. 333, 608 F.2d 445 (1979). Summary and excerpts at tpguidelines.com/us-vs-e-i-du-pont-de-nemours-co-october-1979-us-courts-of-claims-case-no-608-f-2d-445-ct-cls-1979/
I reviewed the Du Pont trial files in the U.S. National Archives. About three broken boxes containing dusty and unindexed documents inside. Key documents were pilfered, including the reports by outside IRS testifying economists Charles Berry and Irvin Plotkin. I got a disappointing sense that regarding record keeping, the U.S. was a 3rd world country. The staff was illiterate and unhelpful, and rude to boot. They were uninterested low-paid guardians of disarrayed documents.
National Semiconductor Corp. & Consolidated Subsidiaries v. Commissioner, T.C. Memo. 1994-195, 67 T.C.M. (CCH) 2849 (May 2, 1994), Docket Nos. 4754-89 and 8031-90; acq. in result, 1995-2 C.B. 1. William Bonano appears as counsel for IRS (respondent).
Nissho Iwai American Corp. v. Commissioner, T.C. Memo. 1985-578, 50 T.C.M. (CCH) 1483 (November 26, 1985), Docket No. 1702-83 (Korner, J.). Full text available at CourtListener: www.courtlistener.com/opinion/4817422/nissho-iwai-american-corp-v-commissioner/. The rejection of the Industrial Forestry Association composite log sales analyses, and the holding on industry averages, appear in the Opinion under Issue 1.
The petitioner is Nissho Iwai American Corporation, the New York subsidiary, and the parent is Nissho Iwai Corporation of Tokyo. And there are several unrelated Nissho Iwai cases in the Court of International Trade from the same period, including a customs valuation case that transfer pricing people cite for the first sale rule. Anyone searching the name without the T.C. Memo number will land on those first, which is why the docket number and the direct link are both in the entry.
Yamaha Motor Corp., U.S.A. v. United States, 779 F. Supp. 610 (D.D.C. 1991) (Civ. A. No. 91-2125). law.justia.com/cases/federal/district-courts/FSupp/779/610/2313930/. No published Tax Court opinion.
Mueller, Dennis C. (1986). Profits in the Long Run. Cambridge University Press. www.cambridge.org/core/books/profits-in-the-long-run/861E9E41DF5BA7B7DA3A2484E0E7B508
Risk Management Association, Annual Statement Studies. On the history of the association, founded 1914 as Robert Morris Associates, and the November 2024 merger with the Bank Administration Institute to form ProSight Financial Association: moneysoft.com/rma-annual-statement-studies/
On the Compact Disclosure and Moody’s CD-ROM products of the period, and the corporate history of Disclosure, Inc., see the companion post “Where the Comparables Come From.”
Personal recollections in this post, the CUNY computer lab remote access, the Yamaha case related Compustat subscription and its settlement in pretrial negotiation, and the three-bid procurement, are not documented and are offered as recollection. Troy Haines assisted me in the Yamaha computations. The DuPont and Nisho Iwai cases were part of my training portfolio when I joined the IRS as a transfer pricing economist. The lesson drawn from the cases reviewed was the IRS needed outside experts to defend cases in litigation.