The research credit: Using statistical sampling

Editor: Greg A. Fairbanks, J.D., LL.M.

This item discusses the Sec. 41 research credit and how statistical sampling can be used to efficiently satisfy the four-part test that governs whether research activities qualify for the credit. The focus here will be on guidance recently provided by the IRS’s Large Business & International (LB&I) division in Field Attorney Advice (FAA) 20212501F.

The FAA makes clear that taxpayers must analyze the requirements for qualified research activities under the four-part test of Sec. 41(d)(1) separately for each business component that a taxpayer uses as a factual basis for a research credit claim. Further, the FAA explains that a taxpayer’s use of statistical sampling in accordance with Rev. Proc. 2011-42 does not change the requirement that the four-part test must be satisfied separately for each business component. The FAA also states that the application of the four-part test to business components does not change when the IRS uses a statistical sample to audit a research credit claim.

The discussion below provides a basic overview of the research credit and then explains how to efficiently satisfy the IRS requirements for the credit, using statistical sampling, considering the guidance contained in the recently issued FAA.

Business components, qualified research, and the four-part test

Sec. 41 and the regulations thereunder provide a nonrefundable credit for increasing research activities. The research credit is generally computed based on the amount of qualified research expenses (QREs) exceeding a base amount that is derived from historical taxpayer data. Taxpayers compute the research credit either under the regular method as described in Sec. 41(a) or the alternative simplified credit (ASC) method in accordance with Sec. 41(c)(4). The mechanics of the research credit calculation are dependent upon the method elected by the taxpayer.

Due to the constantly evolving IRS audit environment and the contentious nature of qualified research, it is critical that taxpayers maintain sufficient records to substantiate research credit claims. Substantiation sufficient to satisfy an IRS audit can be extensive, which is why statistical sampling is an attractive approach. When performed properly, statistical sampling provides an accurate estimate of QREs while reducing the need to review extensive contemporaneous documentation. However, the analysis of sampled records still requires an evaluation of the four-part test at the business-component level.

As summarized in the FAA, Sec. 41(d)(2)(A) and Regs. Sec. 1.41-4(b)(1) generally require taxpayers to determine qualified research activities separately for each business component. Sec. 41(d)(2)(B) defines a business component as any product, process, computer software, technique, formula, or inventory that is held for sale, lease, or license, or that is used by the taxpayer in its trade or business. For example, a motorcycle engine builder may identify a new engine product as a business component.

Further, the FAA observes that qualified research activities must generally meet the four-part test, which consists of the following requirements:

• Sec. 174 test: Research expenditures must qualify as research or experimental (R&E) expenses under Sec. 174. R&E expenses are costs associated with activities intended to discover information that would eliminate uncertainty concerning the development or improvement of a product (see Regs. Sec. 1.174-2(a)).
• Technological-information test: Activities must be undertaken for the purpose of discovering technological information. As indicated in Regs. Sec. 1.41-4(a)(4), information that relies on principles of the physical or biological sciences, engineering, or computer science is generally considered technological in nature.
• Business-component test: Taxpayers are required to undertake development activities to discover information used in the development of a new or improved business component, as defined above.
• Process-of-experimentation test: “Substantially all” (i.e., 80% or more) of the activities must constitute elements of a process of experimentation conducted for a new or improved function, performance, reliability, or quality (see Secs. 41(d)(1)(C) and 41(d)(3)(A)).

As summarized in the FAA, the specific requirements with respect to the process-of-experimentation test, including the “substantially all” requirement, can only be applied at the business-component level. However, if the overall business component does not meet all four tests, certain steps are followed. In accordance with Regs. Sec. 1.41-4(b)(2), the shrink-back rule provides that the four-part test is first applied at the level of the discrete business component. If the requirements are not met at this level, then the four-part test is applied at the most significant subset of elements of the business component. For example, if the new engine product developed by the motorcycle engine builder referenced earlier does not meet the four-part test, the requirements may be applied at the next most significant subset of elements of the engine (e.g., the carburetor).

Substantiation requirements

Regs. Sec. 1.41-4(d) provides that taxpayers claiming a research credit are required to keep records “in sufficiently usable form and detail to substantiate that expenditures claimed are eligible for the credit.” Sec. 6001 and the regulations thereunder provide taxpayers specific guidance with respect to record retention, including the requirement to keep permanent books of account or records. As highlighted in the FAA, the substantiation requirements must also be satisfied at the business-component level. Taxpayers can look to case law for examples of how the courts interpret the substantiation requirement with respect to business components and qualified research activities.

There are a variety of cases that examine the substantiation requirement and demonstrate that the four-part test is applied at the business-component level. With respect to substantiation, the FAA cites several notable cases including Union Carbide Corp.,T.C. Memo. 2009-50, aff’d, 697 F.3d 104 (2d Cir. 2012), and Suder, T.C. Memo. 2014-201. Based on the recent audit environment, IRS examinations appear to be placing more focus on taxpayers’ business components as well as certain four-part test requirements that have historically received less scrutiny, such as the technological-information test.

Two recent cases that reflect this trend and emphasize that the burden of proof is on the taxpayer to satisfy and substantiate the four-part test at the business-component level are Siemer Milling Co.,T.C. Memo. 2019-37, and Max,T.C. Memo. 2021-37. In Siemer Milling, the court analyzed and evaluated the four-part test for each of the taxpayer’s seven business components independently. The court disallowed the taxpayer’s research credits because none of the business components met all four parts of the four-part test requirement.

Similarly, in Max,the Tax Court concluded that the taxpayer could not claim a research credit because the taxpayer’s activities did not meet three of the four-part test requirements, including the process-of-experimentation test. In particular, the court agreed with the IRS that the taxpayer did not fundamentally rely on principles of the hard sciences and did not employ a process similar to the scientific method during its pre-production process.

The FAA concludes that, while statistical methods can be used for estimation purposes, the general requirement that the four-part test be applied separately to each business component and substantiated at such level is not altered by using statistical methods. Therefore, careful consideration of statistical approaches that capture and allow proper substantiation of the four-part test by each business component is key to successfully defending a research credit claim during audit.

Benefits of statistical sampling

The IRS has a history of accepting taxpayer estimates based on statistical analysis and has provided guidance on sampling through field directives, revenue procedures, and audit technique procedures. Guidance that best supports the use of statistical analysis in research credit computations is included in Rev. Proc. 2011-42, which provides general guidance for taxpayers that use statistical sampling and statistical estimates for tax filing purposes. Provided that taxpayers follow the IRS guidelines, statistical sampling can be a cost-effective, efficient approach to calculating and substantiating the research credit. In addition, the IRS often encourages the use of statistical sampling for the research credit due to the efficiencies provided during examination.

Statistical sampling often minimizes the level of involvement required from the taxpayer to collect the quantitative data required to compute QREs and calculate the research credit. Additionally, statistical sampling may significantly reduce the level of qualitative analysis required to meet the substantiation requirements of Regs. Sec. 1.41-4(d). Taxpayers with a considerable number of business components may find it burdensome to properly document the four-part test requirements for each business component. By implementing a statistical sampling approach, the analysis is limited to the taxpayer’s sample of records.

Further, the use of statistical sampling can make the IRS examination process more efficient for the taxpayer and the IRS. Provided the taxpayer uses valid statistical methods to determine its sample selections, the IRS will generally limit its purview to the taxpayer’s sampling selections during exam. As a result, the process of issuing and responding to information documentation requests (IDRs) during an IRS exam is often more streamlined because taxpayers generally find it less arduous to sufficiently document and substantiate their research credit claims as part of the initial filing rather than in response to an IDR.

Sampling process overview

In a typical research credit study involving sampling, the taxpayer’s list of business components under development during the tax year is reviewed to identify only those business components with qualified activities and QREs. The revised population is next divided into several strata (e.g., buckets) of similar types and sizes, and random samples are selected from each stratum. Tax research credit professionals must subsequently analyze the samples from each stratum in accordance with Sec. 41 and the regulations thereunder to determine the QREs of each sample. After QREs have been determined and substantiated for each sample, statisticians use the analysis results to estimate total QREs for the entire population list.

During the sampling process, statisticians will calculate the accuracy of the estimates, as measured by confidence and precision. Broadly, confidence represents the chance that the selected sample will produce an estimate close to the true value (i.e., total QREs of the entire population), and precision indicates the proximity of the estimate to the true value. The IRS specifies confidence and precision requirements in Rev. Proc. 2011-42, and taxpayers should collaborate with an experienced statistician knowledgeable of IRS statistical expectations when sampling.

Constructing the sample and substantiation requirements

The first step in the statistical sampling process is to construct the population listing or sample frame (i.e., listing of every item in the scope of the study). Building the sample frame requires expenses to be aligned with a sampling unit (e.g., a row of data in the population sampling frame data set listing). Common sampling units in research credit studies include business components, projects, employees, cost centers, locations, departments, supervisors, or a combination of these.

Guidance provided in the FAA amplifies the critical nature of determining sampling units, as application of the four-part test is required to be at the business-component level. When developing a statistical sampling methodology, taxpayers should first attempt to use business components as the sampling unit because this most directly aligns with the requirements of Sec. 41(d)(2)(A) and Regs. Sec. 1.41-4(b)(1) and the substantiation requirements of Regs. Sec. 1.41-4(d). Often, projects directly align with business components and are easily listed, and projects are therefore a natural sampling unit chosen by many taxpayers.

For various reasons, including ease of data collection, taxpayers often establish sampling units as something other than business components. For example, sampling by employee may be a more practical choice for taxpayers that do not capture employee time via time-tracking software. However, taxpayers must use caution when using non-business-component sampling units. To properly substantiate the research credit under this sampling approach, taxpayers must generally demonstrate nexus between QREs for each randomly sampled employee and qualified business components that meet the four-part test requirement. Taxpayers must track and document the complete list of business components on which each randomly sampled employee worked to appropriately establish nexus. Taxpayers may choose to use a two-stage sampling methodology whereby business components are randomly selected for the randomly selected employees to minimize demanding documentation requirements. Alternatively, taxpayers with few business components and/or few employees may find it practical to survey all employees who do not use time tracking at the business-component level.

Sampling by cost center or department can be highly efficient if performed properly and adequate data is available. For example, statisticians may use historical cost center or department information to generate preliminary qualifying percentages. These preliminary qualifying percentages can often be leveraged to create smaller sample sizes. However, taxpayers are still required to analyze the four-part test and substantiate such analysis at the business-component level under this sampling approach.

Again, taxpayers must use caution to implement this kind of study properly. Taxpayers must demonstrate nexus between employee time and qualified business components for each randomly sampled cost center. A two-stage sampling approach can be applied to reduce the burden of demonstrating nexus for each cost center. Under this approach, a random sample of business components is analyzed for each randomly sampled cost center or department. To maintain statistical validity, the assessed business components should not be determined based on qualifying activities and should instead be generated based on valid, statistical random samples. In general, other non—business-component sampling (e.g., location, supervisor, or a combination) follows the same methodology.

Choosing the right sampling method

The above-described statistical sampling methodologies are some common approaches to statistical sampling for research credit studies. However, there are many other acceptable methods for statistical sampling, including variations in sampling units. Taxpayer-specific facts including types of research performed, data availability, and organization structure should be considered to develop the most appropriate statistical sampling methods.

Statistical sampling is an IRS-accepted approach that offers several analysis- and documentation-related benefits that taxpayers can implement to compute the research credit. Further, statistical sampling may prove to be a more efficient methodology for computing the research credit for many taxpayers. However, taxpayers must be aware that the use of statistical sampling does not circumvent the requirement that the four-part test be analyzed and substantiated at the business-component level.

EditorNotes

Greg A. Fairbanks, J.D., LL.M., is a tax managing director with Grant Thornton LLP in Washington, D.C.

For additional information about these items, contact Mr. Fairbanks at 202-521-1503 or greg.fairbanks@us.gt.com.

Contributors are members of or associated with Grant Thornton LLP.