NERA OUTLINES COMMENTS ON TRANSFER-PRICING REGS.

Find enclosed a copy of the comments I will present to the panel on proposed changes to the Section 482 on August 31 of this year.

Sincerely,

OBSERVATION 1

THE PROFIT INTERVAL TEST GUARANTEES THAT AT LEAST ONE COMPARABLE COMPANY BENCHMARK WILL BE FOUND BUT CANNOT RESOLVE CONTROVERSIES IF TWO OR MORE BENCHMARKS ARE PROPOSED.

In the past, economists attempting to use comparable company benchmarks to test transfer prices were often unable to identify appropriate benchmarks. The current proposal eliminates this problem by fiat, declaring that the best available benchmark shall be considered sufficient.

A no-fail procedure is proposed to find this "best" comparable company or companies. Basically, this procedure entails starting with the narrowest definition of acceptable benchmarks and continuing to make the definition broader until a suitable company or group of companies, for which data is available, is identified.

The proposal allows either the taxpayer or the IRS to propose a benchmark company or companies. The party that does not propose the benchmark can challenge the benchmark by offering a better one. A problem can arise in a situation where neither parties' proposed benchmark is clearly superior to the other. The regulations as currently proposed do not clearly resolve this problem.

OBSERVATION 2

THE REAL ISSUE IN SETTING THE PROFIT INTERVAL IS THE TRADE-OFF BETWEEN FAILING TO DETECT TAXPAYERS WHO ARE VIOLATING THE ARM'S- LENGTH STANDARD AND PENALIZING TAXPAYERS WHO ARE COMPLYING.

A. TYPE I AND TYPE II ERRORS

The profit interval test described in the proposed regulation is an imprecise test of the hypothesis that a particular taxpayer is paying arm's-length transfer prices. If the profit interval is very large, violators of arm's-length transfer prices will often go undetected; if it is very narrow, many complying taxpayers will be inappropriately penalized. The primary issue in setting the permissible profit interval is to determine the optimal trade-off between failing to detect taxpayers who are violating the arm's- length standard and penalizing taxpayers who are complying.

Economists define penalizing a complying taxpayer a type I error and failing to detect violators a type II error. They also have developed techniques for quantifying the probability of each type of error. In most economic research, the standard for acceptable type I errors is between one and five percent.

To apply a standard econometric hypothesis test to Example 1 from Section f(11) of the proposed changes to the regulation, one must compute the average constructed income. Then, using well-known econometric procedures, one must compute a confidence interval around the average constructed income. The average constructed income in that example is 155. The corresponding confidence interval, which allows only a 5 percent chance of penalizing a complying taxpayer, is 133 to 177. This range is considerably wider than the range of observations of constructed income, i.e., 150 to 162. Under the proposed rule, a taxpayer with actual operating income of 171 would be required to adjust the transfer payment by at least 9 to move income within the constructed income range. If a confidence interval with a 5 percent chance of a type I error were used, no adjustment would be required of such a taxpayer.

If statistical means are used to estimate constructed income, the regulations recognize the need to use statistical confidence intervals. However, the regulation does not indicate the acceptable amount of type I error. Therefore, an important policy question involves determining the acceptable amount of type I error for transfer pricing cases.

There is clearly an argument favoring the use of basic statistical methods to determine ranges of acceptable profit intervals. Adopting these methods would eliminate guessing about intervals of convergence that will prove an area of controversy for the IRS and taxpayers if left unaddressed by the proposal.

B. THE JOINT HYPOTHESIS PROBLEM

Another aspect of the trade-off between failing to detect violators and penalizing compilers concerns what statisticians call a joint hypothesis problem. If the benchmark companies are part of the same population as the tested party, i.e. closely resemble the tested party, then arm's-length transfer prices will result in the tested party falling within the profit interval. However, if the benchmark companies do not closely resemble the tested party, the transfer price may have been arm's-length and still fall outside the profit interval. In the language of statistics, the proposed profit interval would ideally test only the single hypothesis that transfer prices are arm's-length. Unfortunately, it is actually testing two hypotheses simultaneously: First, "Is the tested company from the same population as the benchmarks?" and second, "Is it paying arm's- length transfer prices?"

In my letter to the IRS dated April 6 of this year I have suggested several approaches to alleviating the joint hypothesis problem.

OBSERVATION 3

STATISTICS OF INCOME DATA IS A POSSIBLE SOURCE OF COMPARABLE COMPANIES AND SAFE HARBORS.

The IRS statistics of income data (SOI) for corporate returns is an obvious place to search for comparable company benchmarks and safe harbor benchmarks. It consists of 185 business classifications with each classification divided into categories based on asset size.

I would recommend establishing a committee or task force to study the possibility of using this data -- in a way that does not violate the confidentiality of taxpayer information -- for Section 482 purposes. An obvious starting point would be to review the average of all relevant ratios that can be computed from the SOI data for each of the 185 business classifications along with the associated standard errors. A possible safe harbor would be a profit interval of two standard errors on either side of the averages for the business classification that most closely resembles the tested parties operations.

OBSERVATION 4

LEAVE THE CHANGES IN PROPOSAL FORM FOR A FEW YEARS WHILE SUGGESTIONS SUCH AS THOSE I HAVE MADE ARE EXPLORED. THIS WILL ALLOW SUFFICIENT TIME TO UNDERSTAND THE FULL IMPACT OF THE PROPOSAL AND TO IMPROVE IT.

This concludes my remarks. I will be happy to try to answer any questions that members of the panel may have.