In Legal Advice Issued by Field Attorneys (LAFA) 20171601F, a Large Business & International (LB&I) attorney with the IRS Chief Counsel determined that a taxpayer (Taxpayer) incorrectly estimated wages that are eligible as qualified research expenses (QREs) for purposes of computing the research credit. The LAFA did not address Taxpayer's use of statistical sampling and offered no opinion on whether: (1) Taxpayer's statistical sample was appropriate or acceptable, or (2) Taxpayer properly followed Revenue Procedure 2011-42 in its statistical sampling and extrapolation approach.

The LAFA addressed Taxpayer's method of determining the percentage of employee time spent performing qualified services under Treas. Reg. Section 1.41-2(d)(1), to determine its QREs attributable to it employees' wages. The method was found to be inaccurate and could understate or overstate QREs. According to the LAFA, "The method was based on statistical analysis that did not produce an appropriate measure of [Taxpayer's] in-house research expenses."

Background

Taxpayers will commonly use statistical sampling to estimate items for federal tax purposes. In Revenue Procedure 2011-42, the IRS has provided taxpayers with guidance on the use and evaluation of statistical samples and sampling estimates. For example, the IRS permits taxpayers to use statistical sampling, as outlined in Revenue Procedure 2011-42, to establish items on their tax return when the time or costs required to analyze large volumes of data is burdensome. When done properly, the review and extrapolation of a valid statistical sample are accepted by the IRS in determining QREs for the research credit, and EY has never had a sample rejected by the IRS.

In the case addressed in the LAFA, Taxpayer estimated its QREs attributable to wages using a two-step process. First, Taxpayer estimated the portion of its total wages paid or incurred for the performance of qualified services. Taxpayer identified which employees performed qualified services at any time during the tax year. Taxpayer then estimated the fraction of each employee's time spent performing qualified services. Taxpayer multiplied this fraction by the employee's total wages for the year. The resulting product represented Taxpayer's estimate of the total amount of a particular employee's wages for the time spent performing qualified services. Taxpayer then added the amounts calculated for each employee to compute the initial estimate of total wages incurred for qualified services.

Next, Taxpayer multiplied the estimate it calculated in step one by a fraction. The denominator of the fraction was the number of projects that Taxpayer believed to involve more than 50 hours of work by employees and that might have involved at least one employee who conducted qualified research during some part of the project. The numerator of the fraction was a subset of the projects in the denominator for which Taxpayer determined that an employee did perform qualified research during some part of the project. It appears that Taxpayer determined that 679 of its projects met the criteria for being included in the denominator and randomly selected and reviewed 29 of those projects to help determine the numerator of the fraction. Ultimately, Taxpayer determined that roughly 52% of the projects included in the denominator of this fraction involved an employee engaged in qualified research at some point.

In analyzing this sample, Taxpayer only considered whether the project involved qualified research at some point. Taxpayer did not determine what percentage of total project costs, such as wages, were attributable to the performance of qualified services. The IRS did not consider whether Taxpayer's sample selection was appropriate.

Issues

If a proper evaluation of each employee's qualified wages were conducted in the first step of Taxpayer's process, there would have been no need to further adjust the population by the sample findings determined in step two. In addition, when reviewing the sample in step two, Taxpayer did not determine what percentage of total project costs were attributable to qualified services. Instead, the sample was used to identify the portion of projects that involved at least one employee engaged in qualified research at some point. This type of ratio estimation approach is not appropriate under Revenue Procedure 2011-42, because the evaluation and extrapolation of the sample item determinations must be made on a quantifiable basis.

This ratio was then applied to a non-statistical estimate initially computed by Taxpayer in step one of the aforementioned two-step process. Therefore, not only was the sample improperly evaluated, but the results of the sample were improperly applied to an alternate position to determine total QREs.

To illustrate why this ratio method is improper, assume Project X was selected for review and it had two employees working full-time on the project. The Employee #1 had a total wage amount of $100,000, and Employee #2 had a wage amount of $500,000; therefore, the total wage amount associated with Project X was $600,000. Upon review of these two employees, Employee #1 was found to have spent all of his or her time on qualified services on Project X, and Employee #2 was found to have conducted no qualified services on Project X. Consequently, the taxpayer would have determined that Project X qualified because one employee conducted some qualified services at some point during the year, but actually only $100,000, or one-sixth of the total wages associated with this project, are indeed qualified. Applying a ratio developed from this methodology to any dollar-based population is likely to produce inaccurate results.

While the LAFA does not comment on whether Revenue Procedure 2011-42 was properly followed, it does offer that the sample might not have been representative of the population of projects containing qualified research for the years in question. The sample reportedly contained projects from years outside of the years under consideration. It is noted that "the portion of Taxpayer's projects that involved qualified research in those other years may have been different than the portion of Taxpayer's projects that involved qualified research in the years at issue." Because the results of the sample can only be extrapolated to the population from which it was sampled, the inclusion of projects outside the tax years at issue in the computation would have invalidated the position formulated for the actual years in question.

Conclusion

Statistical sampling, when done properly, is a useful tool to taxpayers that can save time and money, and provide scientifically measurable support for an estimate of an item included in a tax position. Statistical sampling is an acceptable method for determining almost any federal tax position, but proper sampling, evaluation, and extrapolation techniques must be utilized to avoid invalidating the results.

———————————————