Transfer pricing and the cumulative advertising effects on sales

By Dr. Ednaldo Silva, founder and managing director of RoyaltyStat, Bethesda, Md.,

A legal strategy in transfer pricing is to dismember the controlled distributors or retailers of integral management or purchase functions to report reduced profit margins.

This legal fiction (no comparable third-party) is made more incongruous by booking a large fraction of the corporate group’s advertising expenses as the tax deductions of the stripped distributors or retailers.

Also, operating loss enterprises are selected as supposed comparables to the forced-invalid distributors.

Triple crown winners are rare.

Here, I test the effect of advertising on enterprise-level sales (revenue) and show that marketing intangible producing activities such as advertising expenses cannot coexist with the legal concept of limited function distributor or retailer.

Hereafter, my exegesis is focused on a group of large US retailers.

Distributed lag effects

Intangible creating activities are distributed over time; thus, the spillover effect of advertising on sales takes time from annual expenditures to fruition (future revenue recognition).

This process can be modeled by a distributed lag, such as:

      (1)     S(t) = α + β₁A(t) + β₂A(t−1) + β₃A(t−2) + … + random error

where S(t) denotes net sales and A(t) denotes advertising in period t = 1 to T years. The intercept α covers the level of net sales independent of advertising.

If the data sample is small, the degrees of freedom limitation can impair the estimation of the regression parameters. Assumptions can be tested about the time lag behavior, and here I adopt the well-accepted Koyck’s assumption that the lag effect decays exponentially. Thus, equation (1) is reduced to a regression with fewer parameters to be estimated:

    (2)     S(t) = μ + γ₁ S(t−1) + γ₂ A(t) + random error

See Kmenta (1986), pp. 529, 531 or Maddala (1977), pp. 141, 189, 360-364 for a discussion of model (2).

Palda (1964) applied Koyck’s model (2) using historical data (1907 to 1960) from a single company (Lydia Pinkham). See e.g., Palda’s equation (5.1a), p. 54. However, lacking theoretical discipline, Palda estimated several power functions, but not the linear Koyck specification of his dissertation.

The effect of advertising on sales is measured by the partial regression coefficient:

      (3)     ∆S(t) = γ₂ ∆A(t)

where the regression coefficient γ₂ is the advertising (effect) multiplier on net sales.

Empirical estimates of the effect of advertising on sales

Consider a few major US retailers and estimate their Koyck regression equation (2). The estimates were computed using the internal Koyck regression function of RoyaltyStat/Compustat (the intercept is not reported here to simplify interpretation):

     (2.1) Best Buy (GVKEY 2184): S(t) ≈ 0.655 S(t−1) + 18.5 A(t)

with the Newey-West t-statistics 6.1 and 3.6, count 29 annual (1983-2019) observations, and adjusted R² = 0.984

     (2.2) Home Depot (5680): S(t) ≈ 0.773 S(t−1) + 19.97 A(t)

with the Newey-West t-statistics 4.4 and 1.4, count 33 annual (1980-2019) observations, and adjusted R² = 0.966

     (2.3) Lowe’s (6829): S(t) ≈ 0.976 S(t−1) + 3.7 A(t)

with the Newey-West t-statistics 14.1 and 0.792, count 41 annual (1978-2019) observations, and adjusted R² = 0.995

     (2.4) Target (3813): S(t) ≈ 0.848 S(t−1) + 8.2 A(t)

with the Newey-West t-statistics 15.1 and 2.8, count 41 annual (1978-2019) observations, and adjusted R² = 0.996

     (2.5) Walmart (11259): S(t) ≈ 0.75 S(t−1) + 24 A(t)

with the Newey-West t-statistics 6.3 and 1.7, count 23 annual (1978-2019) observations, and adjusted R² = 0.975. Walmart’s advertising expenses from 1980 to 1997 were not reported to the SEC (Securities & Exchange Commission).

Substitution of routine by non-routine functions

The advertising multiplier varies among US retailers, and the effect of advertising on sales can be large.

If advertising is significant, the spillover effect of advertising on sales is high, such as the cases of Best Buy (γ₂ ≈ 18.5) and Target (γ₂ ≈ 8.2). This means that $1 million spent on advertising can generate $18.5 million or $8.2 million, respectively, of incremental sales.

Cumulative advertising expenses create a stock of marketing intangibles, and the exploitation of intangibles create the expectation of non-routine profits in addition to routine profits (which reflect the basic economic activities of the enterprise such as distribution or retail).

The legal notion that controlled entities engaged in substantial advertising expenses can be characterized as a limited function operation expected to earn routine profits is a fiction, not compatible with economic principles or reality.

The legal stripped-down (corporate reorg) amounts to a substitution of the routine management or purchase function of the controlled distributor or retailer by the non-routing advertising function, called DEMPE (development, enhancement, maintenance, protection, and exploitation of intangibles) activities by the OECD.

Contrary to optical illusions, the substitution of routine (operations management, purchase for resale, inventory control) by non-routine (advertising, intangible producing) activities in the controlled distributor or retailer’s business is expected to produce above-normal profit margins. See OECD (2017) paragraph 6.32 (“Ownership of intangibles and transactions involving the development, enhancement, maintenance, protection and exploitation of intangibles”). Inter alia, see OECD (2017), paragraph 6.59, Footnote 4 (“the use of assets includes the contribution of funding and/or capital to the development, enhancement, maintenance, protection or exploitation of intangibles”).

References

Jan Kmenta, Elements of Econometrics (2nd edition), Macmillan, 1986.

G. Maddala, Econometrics, McGraw-Hill, 1977.

OECD, Transfer Pricing Guidelines, July 2017.  Available here.

Kristian Palda, The Measurement of Cumulative Advertising Effects, Prentice-Hall, 1964.