The Battle Of Assumptions In Facebook’s U.S. Tax Court Case

Recent testimony in the U.S. Tax Court’s Facebook v. Commissioner trial highlighted a problem with the U.S. cost-sharing regulations, which establish valuation rules for a type of related-party intangible property transfer that has long been popular among U.S. tech giants.

The issue isn’t with the underlying concepts adopted by the regs, which are solidly grounded in economic theory and consistent with standard valuation practices. The problem is that the regulations’ valuation methods can be highly sensitive to seemingly modest changes in assumptions, and they often fail to offer any clear way to resolve the inevitable disagreements that will arise between the IRS and taxpayers.

The issue and its monetary significance are highly evident in the Facebook case, which concerns the U.S. parent company’s 2010 contribution of technology rights, international user base, and international trademark rights to a cost-sharing arrangement entered into with an Irish subsidiary.

The IRS’s valuation methodology generated a value of about $21 billion for those contributions in total. Facebook’s favored methodology resulted in a value of between $2.5 billion and $3.7 billion for the technology and international user base, plus a 1 percent royalty for the international trademark rights.

Although differing legal interpretations account for some of that discrepancy, much of it reflects differences in valuation assumptions. The IRS used the income method to value the transferred intangibles, which the cost-sharing regs generally favor when only one party to the transaction contributes unique and valuable intangible property. The income method, which resembles a discounted cash-flow (DCF) valuation in some respects, relies on a present-value calculation for expected future operating income, or earnings before interest and taxes. Accordingly, it requires revenue and cost forecasts, along with a discount rate to convert EBIT in future years to present value.

One of the IRS’s expert witnesses, Carl Saba of Hemming Morse LLP, defended his valuation assumptions on cross-examination during a May 12 trial session. For his expert witness report, Saba performed a DCF enterprise valuation of Facebook that corroborated the results of the IRS’s income method valuation.

Saba’s choice of discount rate was a major point of contention. Saba used a common method for calculating the appropriate discount rate: the capital asset pricing model, according to which the expected rate of return for a financial asset is equal to a theoretical risk-free return plus a return to compensate investors for the additional risk associated with the asset. Determining the compensatory return value requires calculation of the beta coefficient, which reflects the volitivity of the asset’s value relative to the volitivity of the market as a whole. Because Facebook’s beta coefficient couldn’t be directly observed before the company’s 2012 initial public offering, Saba estimated the appropriate beta for 2010 based on what he considered to be comparable tech companies.

The regulations, which offer little tangible guidance on discount rates, provide no clear-cut basis for determining whether Saba’s estimated discount rate of 17 percent was more reasonable than the higher rates favored by Facebook.

Even more problematic is the regs’ inability to clearly establish how financial projections should be determined. To estimate Facebook’s future operating margins as of 2010, Saba calculated the weighted average earnings before interest, taxes, depreciation, and amortization margin for fellow tech giants Google

GOOG
, Baidu

BIDU
, Tencent, and Gree. Facebook pressed Saba on his choice of comparables, particularly Baidu, which dominated the Chinese search engine market after Google withdrew from China in 2010.

Facebook has few, if any, comparables. But the fragility of any valuation approach that relies on such a small sample size is evident from the drastic effects of relatively modest changes, which Facebook highlighted at trial. In Saba’s EBITDA multiple valuation analysis, which he used to corroborate his capital asset pricing model valuation, simply switching from the weighted average to the median value (which the transfer pricing regulations generally favor over the mean) led to a roughly $4.5 billion drop in the aggregate valuation. Removing Baidu reduced the aggregate value by another $5 billion.

The cost-sharing regs don’t provide any authoritative pronouncement on which companies should be considered comparable, how discount rates should be determined, or how revenue and costs should be forecast. None of those variables has any one authoritative value, and each is layered on top of other subjective judgments. With no clear right answer and billions of dollars hanging in the balance, protracted and costly disputes between the IRS and taxpayers are bound to be more common than agreements.