Tutorial: Using Regression Analysis in TNMM Transfer Pricing

February 16, 2021 by Ednaldo Silva
About the Author
Ednaldo Silva
(Ph.D. Econ.) is a leading economist with over 25 years’ experience in transfer pricing and the valuation of intangibles.
He is founder and managing director of RoyaltyStat, a boutique data, software and consulting firm specializing in transfer pricing and intellectual property solutions. Dr. Silva was first Sr. Economic Advisor at the IRS Office of Chief Counsel, a drafting member of U.S. 26 IRC section 1.482 transfer pricing regulations, and introduced the “comparable profits method” (CPM in the US and TNMM in the OECD), “best method” rule and the concept of arm’s-length represented by a range of results, rather than point estimate.
Read more

More articles by this author

Find Ednaldo Silva on LinkedIn

The profit indicator under the CPM (TNMM) is calculated in four steps:

  1. Select the comparables to the tested party (taxpayer).
  2. Select the profit indicator.
  3. Calculate quartiles of the profit indicator. 
  4. Determine if asset adjustments to the profit indicator are significant.

The problem with these four steps is that the calculation of quartiles depends on special assumptions: First, the relationship between the profit indicator and the base (revenue, total cost, or assets) is linear; second, the intercept is insignificant.

In addition, there is no direct way to test the effects of assets employed on the profit indicator. As a result, the interquartile range of the profit indicator is wide (unreliable), and the assets intensity adjustments are ad hoc (arbitrary, without economic merit).

These problems can be solved by using regression analysis, which produces more defensible statistical ranges of the profit indicator that are more resistant to audit scrutiny.

However, estimating structural equations like the underlying linear equation of quartiles produces biased (omitted variable) results. Therefore, I suggest using the reduced-form equation:

(1)  R(t) = C(t) + Y(t)

(2)  Y(t) = α + β R(t) + U(t)

R(t) = α + C(t) + β R(t) + U(t)
R(t) − β R(t) = α + C(t) + + U(t)

(3)  R(t) = λ0 + λ1 C(t) + V(t)

where the intercept λ0 = α / (1 – β) and the slope coefficient λ1 = 1 / (1 – β) > 1 is the operating profit markup.

The operating profit margin can be obtained by solving for β from λ1:

(4)  λ1 = 1 / (1 – β) ↔ β = (λ1 – 1) / λ1

The necessary variables per company to estimate the reduced-form equation (3) can be found in EdgarStat®:

(a)  R(t) is revenue in year t = 1 to T audit years (plus at least two prior or post audit years).

(b)  C(t) is total cost, which is the sum of direct cost and indirect expenses (including depreciation and amortization). Thus, the indirect least squares slope coefficient β is the operating profit margin after depreciation and amortization (EBIT margin).

(c)  Y(t) is operating profit after depreciation and amortization (EBIT).

(d)  The unknown variables U(t) and V(t) are the random uncertainty.

Example from five U.S. pharmaceutical companies (annual data from 1982 to 2019):

Company

Count

λ1

t-Statistics

R2

Intercept (λ0)

Abbot Laboratories

38

1.1616

41.1

0.992

Significant (2.4)

Bristol Myers Squibb

38

1.2645

59.3

0.9669

Insignificant

Johnson & Johnson

38

1.35

87.3

0.9964

Insignificant

Eli Lilly

38

1.1589

21.5

0.9603

Significant (2.2)

Pfizer

38

1.2496

58.5

0.9868

Insignificant

All 

190

1.2976

65.5

0.9897

Insignificant

 

The OLS (ordinary least squares) regression results above can be computed in EdgarStat online (interactive applications) using the Scatterplot or the Regression function.

The Newey-West t-Statistics are corrected for autocorrelation among the regression residuals. See https://www.econometrics-with-r.org/15-4-hac-standard-errors.html.

Headquarters
EdgarStat LLC
6931 Arlington Road, Suite 580
Bethesda, Maryland, 20814-5284
USA
Customer Support
+1 (202) 558-2356
support@edgarstat.com