Statistical sampling makes it easier to claim a research tax credit than ever before

By Wendy Rotz and Rajitha Siriwardana, Washington

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

Not just large but also medium-size and even smaller companies can more easily and cost-effectively take a research tax credit (RTC) now that the IRS allows multiple tax years under one statistical sample study.

For decades, taxpayers (especially larger companies) have used estimates based on statistical samples to file original returns and refund claims for the federal RTC as well as state research and development (R&D) credits. Statistical sampling, which the IRS accepts broadly, also applies to the domestic manufacturing deduction, tangible property, repairs, and cost-segregation studies, to name a few other tax applications that rely heavily on sampling for estimating values of deductions and credits on tax returns. With statistical sampling, only a portion of a company's records require review and documentation to support qualifying expenses. That sample is used to extrapolate the credit for the entire company.

Even with a sample, the documentation the IRS requires for an R&D study can be onerous, as it requires rigorous support of a four-part test (described below) to claim that an expenditure qualifies for the credit, as well as clearly documenting the nexus (also described below) between qualified research activity (QRA) and the expenses claimed for the credit; the nitty-gritty details can be tedious. Traditionally, only larger companies have found the RTC worth their effort, even with the use of statistical sampling. Until now, many smaller companies and even some larger ones have either forgone the credit or claimed an RTC without thorough documentation — leading to more exposure in audit, as well as time and effort defending an RTC.

The IRS has long said that because the RTC is incremental in nature, it must be estimated annually — and the Service has not wavered from that position.

What has changed is how precise the estimates must be. The accuracy of the estimates is now assessed over the cumulative study rather than individually for each credit year. This greatly reduces the required sample size and, therefore, the amount of effort spent documenting and supporting the credit. That, in turn, lowers the cost of a study while achieving the same benefit. With the ability to stack as many open tax years as desired into a single study, the cost-benefit ratio of claiming the RTC is now more favorable than ever before.

Increasing R&D

The RTC is a credit for increasing R&D expenditures over a historic base amount. Taxpayers may elect one of two methods for determining the base: the regular credit or the alternative simplified credit (ASC). In both, the qualified research expenditures (QREs) for the tax year (the credit year) must exceed the base amount to obtain any credit.

For the regular method, (1) a fixed base percentage (FBP) is determined from the ratio of QREs to gross receipts during the period 1984 through 1988; (2) the smaller of the FBP or 16% is then multiplied by the four most recent years' gross receipts to determine the base amount; and, finally, (3) the excess of the credit year's QREs over the base amount (which cannot be less than 50% of the credit year's QREs) is then multiplied by 20% to calculate the credit.

The ASC is an attractive alternative for taxpayers with difficulties determining financial information and research activity from decades past. It is also attractive for startup companies, defined as either (1) companies for which the first tax year in which both QREs and gross receipts began after 1983 or (2) companies that had no more than three tax years with both QREs and gross receipts during the period 1984 through 1988. For the ASC method, the base amount for each credit year is half of the average of the prior three tax years' QREs. The amount by which the credit-year QREs exceed the base amount is multiplied by 14% to determine the RTC.

Thus, when a taxpayer first claims a credit, it must determine QREs not just in the credit years but also in a historical base period that spans three or four years, depending on the method used.

Multiple years in scope

Nowadays, when companies first claim an RTC, many chose the ASC method. For the first year an RTC is taken, this means quantifying QREs for four tax years (the current credit year and the three prior years in the base period).

Historically, the IRS has accepted a sample over the three-year combined base period, but each credit year required its own sample and precision assessment. This essentially amounted to conducting several samples, one for the combined base period and one for each credit year. This had been cost-prohibitive for many taxpayers.

Now the IRS has softened its position and will accept precision assessment (defined below) over the entire period rather than annually. Taxpayers can save even more costs when looking back to claim a refund in prior years. For example, putting three open tax years together in one study under the ASC requires six years of analysis, one for each credit year and three for the base years. The IRS now allows the precision of the estimates to be assessed over the entire six-year period, thus allowing three years of credit to be claimed with considerably less work, resulting in a substantial cost savings for all taxpayers, large and small.

Still need annual estimates

Even though the precision of the estimates is assessed cumulatively, estimates must still be extrapolated annually. Furthermore, because the IRS still considers the credit incremental in nature, the sample design and estimation must be structured to allow credit-year estimates to be made annually without influence of QREs from other tax years.

Most statisticians achieve this through stratification — each tax year is kept in a separate stratum (or "bucket"). The strata are sampled separately and estimated separately. So while the annual sample sizes cannot be made ridiculously small, they can be substantially smaller now. They only need to be sufficient to have reasonably stable estimates each year and good precision (defined below) over the cumulative period.


Precision is a statistical measure describing how much an estimated value might vary. It is influenced by sample sizes, random selection, estimation method, and intrinsic variability within the population and sample data itself. It is that plus or minus amount that is reported with an estimate. For example, suppose total QREs are estimated to be $1,000,000 plus or minus $80,000. Relative precision is a measure of how much an estimate may vary compared to its size. In this example, the relative precision is $80,000 ÷ $1,000,000 = 8%.

Relative precision is the measure of accuracy the IRS uses to assess taxpayer estimates. According to Rev. Proc. 2011-42, when the relative precision is below 10% (smaller is better), an estimate may be used without an adjustment due to estimate accuracy. However, when the relative precision exceeds 15% (larger is worse), the estimate is unfavorably adjusted by the relative precision percentage. Between 10% and 15%, there is a sliding scale adjustment.

Historically, the IRS has allowed the precision to be assessed over cumulative tax years in other tax applications. Two examples are that the IRS allows unlimited years to be combined in cost-segregation studies and three consecutive years for the meals and entertainment deduction. With the exception of the base years in an ASC, until now the IRS had not allowed multiple years under one precision assessment in R&D. The shift substantially changes sample requirements.

For example, a company with 100 projects annually claiming three years of credit using the ASC method would have six years' worth, or 6 × 100 = 600 projects, to consider. Previously, the 300 projects in the base years may have required a sample of about 40 projects, and each credit year would require separate samples of about 30 projects to achieve reasonable precision annually. This is a total of 40 + (3 × 30) = 130 projects for which to sample, review, conduct the four-part test described below, establish nexus (also described below), and gather supporting documentation. Now this hypothetical company may only need to provide support for a sample of about 10 a year, for a total of 60 projects, which is much more feasible. Many companies may find the effort could be about half of what it may have been in the past.

It is wise to get a statistician who understands IRS expectations to design and extrapolate the sample. Not all statisticians are the same. Just as your cardiologist and dermatologist have different skill sets, a sampling statistician familiar with IRS requirements has a different skill set from those who may be within a taxpayer's company. This specific skill set will help taxpayers comply with the IRS field directives and revenue procedures and design an efficient study that allows taxpayers more time to assemble the thorough documentation to support their QRE findings.


With fewer cases to document, taxpayers can devote more resources to better documentation and support on the sampled records. Sampling does not exempt the taxpayer from complying with the four-part test, each part of which is applied to business components separately. Activities must pass all four tests to be qualified research activities.

The four-part test includes the following:

  • Permitted purpose: The research purpose should be to produce a new or improved product or process that, consequentially, increases the performance, functionality, reliability, or quality of the product or process.
  • Technological in nature: The research activities should be performed on principles of physical sciences, biological sciences, computer science, or engineering.
  • Elimination of uncertainty: Taxpayers must demonstrate an attempt to eliminate uncertainty regarding the capability or method in the development or improvement of the product or process.
  • Process of experimentation: Taxpayers must establish a process of experimentation, such as evaluating alternatives, simulations, confirmation of hypotheses, trial and error, prototypes, or modeling. Failed attempts should be documented.

What is more, once an activity passes all four tests, the taxpayer must still demonstrate the nexus between the activity and the expenditure. For example, they must connect which employees worked on the project, for how long, and performing which activities. However, this is much less work with smaller samples to consider and will disrupt normal business operations less with employee interviews and documentation requests. This lessened workload may allow some taxpayers to more thoroughly pull together and organize their supporting documentation, putting their company in a better audit position.

A more favorable outlook

Larger companies have been taking advantage of sampling to estimate the RTC for a long time. Now medium and smaller companies may find their cost/benefit ratio for an RTC has a more favorable outlook — especially for multiple credit years.

There are many ways to organize a sample; the most common are by project or by employee. Once these are listed, they can be sampled.

There has never been a better time to conduct an R&D study.


Greg Fairbanks is a tax managing director with Grant Thornton LLP in Washington.

For additional information about these items, contact Mr. Fairbanks at 202-521-1503 or

Unless otherwise noted, contributors are members of or associated with Grant Thornton LLP.

Tax Insider Articles


Business meal deductions after the TCJA

This article discusses the history of the deduction of business meal expenses and the new rules under the TCJA and the regulations and provides a framework for documenting and substantiating the deduction.


Quirks spurred by COVID-19 tax relief

This article discusses some procedural and administrative quirks that have emerged with the new tax legislative, regulatory, and procedural guidance related to COVID-19.