For nearly two decades, taxpayers have filed original returns and refund claims for the research tax credit using estimates based on statistical samples. Sampling is versatile and flexible. It may be used for the research tax credit on an originally filed return or to submit a refund claim. Some taxpayers use sampling for base period determinations; others already have these figures, so they only estimate the current year.
The regular research credit equals 20% of a taxpayer's current-year qualified research expenditures (QREs) that exceed a base amount, determined by applying the taxpayer's historical percentage of gross receipts spent on QREs (the fixed-base percentage) to the four most recent years' average gross receipts. In general, the historical base period for determining the fixed base percentage is 1984 to 1988. However, for a startup company, which is defined as a company that did not have both gross receipts and QREs in at least three of the base period years, or the first tax year in which there were both QREs and gross receipts began after Dec. 31, 1983, a different, more complicated set of rules applies. To avoid the inherent difficulties in determining the fixed-base percentage based on what may be long past years, taxpayers can elect an alternative simplified credit (ASC). The ASC is 14% of the QREs for the tax year that exceed 50% of the average QREs for the three tax years preceding the credit determination year.
Sampling may be used whether a taxpayer is using the ASC or the regular method. Sampling can be applied whether or not the credit is base-sensitive. Samples may be organized by employee, project, department, supervisor, or any other convenient means of listing, grouping, and analyzing the costs that potentially may qualify.
In statistical sampling, cases are randomly selected from a listing. The QREs are determined on the sampled items rather than the entire listing. Statisticians use these determinations to extrapolate total QREs for the entire listing. This approach saves both the taxpayer and the IRS time and money.
Sampling for the research tax credit is becoming ever more popular as the IRS pushes for more documentation and establishment of the nexus between projects and qualifying activities performed by employees on those projects. Sampling allows the taxpayer to allocate precious resources to gather robust documentation for the sampled projects. Without a sample, either the taxpayer needs to allocate more resources to gather robust documentation for each project, or the taxpayer is limited to gathering cursory documentation for each project. (When a taxpayer draws a valid statistical random sample to estimate QREs, the IRS audits the taxpayer's sample rather than drawing its own selections. Thus, a taxpayer may better prepare for its audit—indeed, control it—with the knowledge that if it is audited, the IRS will be reviewing the sampled records with QRE-supporting documentation already collected and organized.)
The research tax credit "substantially all" rules (called "sub-all" rules, here) are an important consideration when designing statistical samples. This item discusses efficient strategies for a tax department to consider when planning a statistical sample to estimate QREs—particularly so that taxpayers may claim their full allowable benefit under the sub-all rules.
Research Tax Credit Sub-All Rules
There are two 80% sub-all rules to consider with the research tax credit. When 80% or more of an employee's wages qualify as research expenditures, 100% of his or her wages can be claimed. Similarly, when 80% or more of a project's costs qualify, 100% of the project's expenditures can be claimed as QREs. When analyzing each project and each employee, it is easy enough to identify employees and projects meeting the sub-all rules. How does this work when sampling and estimation are used?
Sub-All Complexity
Applying both the sub-all rules to samples can be complex. Statistical principles the IRS applies require taxpayers to apply the sub-all rules to their sampled items before extrapolating the sample to the total population. This works well when there is only one applicable sub-all rule.
For example, consider a sample that is organized and selected by employee. Suppose there are 10 employees in the sample with qualified time between 80% and 100%. When applying the employee sub-all rule, the taxpayer adjusts the qualified time to 100% for these employees. This adjustment is made within the sample before extrapolating so that 100% of the Box 1 Form W-2 wages qualify for these 10 sampled employees. When the post sub-all qualifying wages in the sample are extrapolated up to estimate total qualifying wages, this estimation approach appropriately accounts for the sub-all application of all employees in the scope of the study—whether or not they were selected in the sample. (As a side note, the IRS emphasizes analyzing research activity by business component. Therefore, when the taxpayer opts to sample by employee, it is important for the tax analysis to consider the projects these employees worked on to get to the project/employee "nexus" stressed by the IRS when auditing research tax credit claims.)
Likewise, in a sample organized and selected by project, when 80% or more of one of the sampled projects' costs qualifies, 100% of the project costs are claimed as QREs in the sample. Extrapolating the sample findings after applying the sub-all rule then correctly accounts for the project sub-all rule when estimating total QREs for the company.
So, in situations where only one sub-all rule will apply, the sampling and extrapolation are straightforward. The math becomes trickier in the more common setting when both sub-all rules apply. For example, suppose both sub-all rules apply, and a sample is drawn by project. A sample drawn by project may contain only a portion of an individual employee's time. The employee may have worked on projects that were not included in the sample. The sample does not "see" all an individual employee's time. Therefore, within the sample, the employee time may not meet the sub-all requirement, and the employee time would not be adjusted up to 100%.
In this example, while the benefit from the project sub-all rule is captured, the additional benefit of the employee sub-all rule can be lost. Similarly, without additional care, the benefits of the project sub-all rule can go unrecognized in samples organized by employee. More thought needs to go into designing a sampling approach that will capture the additional QREs when both sub-all rules apply.
Taxpayers have attempted ad hoc post-extrapolation adjustments in an effort to quantify and claim the additional benefits from the second sub-all application. However, these attempts have met with limited success under audit. While the math may sometimes be logical, the tax position on the additional benefit calculated from these ad hoc methods can fall short of the more-likely-than-not standard. The alternatives presented in this item are statistically sound, have been widely used, and are easily accepted by the IRS under audit.
Strategies to Apply Both Sub-All Rules
Teamwork and good communication between the statistician, tax practitioner, and taxpayer are critical. Sample design is both art and science. Understanding the company structure, research documentation, and concentrations of research activities is important to design a sample that will result in the greatest benefit to the taxpayer. Talk early and often during the planning stages.
Slice and Dice
Strategically slicing and dicing the company into parts (or "strata") for sampling is a common solution to the dual sub-all problem. Here is how it works:
The most common example is manufacturers with qualified activities both in a core research and development (R&D) engineering group and in production departments. Consider such a manufacturer where the employees meeting the sub-all rule are all engineers sitting in the core R&D department. Further suppose this manufacturer pulls a handful of employees off production now and then to work on R&D projects with qualified activities. Suppose all of the production employees spend most of their time doing production. So, while some production employees will have qualifying time, no one in production will satisfy the employee sub-all rule. By contrast, in production, suppose—on a project-by-project basis—most of the employee time spent on these R&D projects is qualified, as are the project supply costs.
Splitting this company by the core engineering versus production will then allow a sample structure with the R&D department, sampled by employee, and the production department, sampled by project. In this approach, both sub-all rules are applied where they would be satisfied, and no benefit is lost.
Note that this is not exactly two entirely separate studies. Think of it more as two "buckets" under one "umbrella" of a single study. Although each bucket is sampled and separately evaluated, the estimated QREs from both pieces are summed together, and the IRS assesses the accuracy of this combined estimate. The sample sizes and, therefore, the cost of the study would be much higher if the accuracy were assessed on each piece separately. Combined under one umbrella, the level of effort is less.
Use knowledge of the company to slice and dice its expenditures into appropriate buckets to recognize the sub-all rules where they apply. As long as the expenditures go into only one bucket (no double counting), the approach is statistically sound and can even be applied using the safe-harbor methods outlined in Rev. Proc. 2011-42.
Sample at a Higher Level
Sometimes slicing and dicing will not resolve the whole problem, as there can be areas in a company where both sub-all rules will potentially apply. The problem is sometimes that one cannot get to the full benefit from a combination of employee and project listings—no matter how the organization is sliced and diced.
The sampling unit of employee or project is what is causing the problem. A different sampling unit can resolve it. Strategically choose the sampling level so that 100% qualifying employees and 100% qualifying projects do not get split across sampling units, thereby losing the ability for a random sample to "see" all that qualifies.
As an illustration, consider companies where research activity may be spread across dozens or even hundreds of departments or even subunits. Within each, there may be some employees and projects that qualify, and among those, some of each may satisfy the employee and/or project sub-all rules.
Sampling by department rather than at either the employee or project level may be a good solution here. Essentially, the projects and employees are grouped under their department for sampling purposes. That is, list the departmental expenditures—for wages, supplies, and contractor costs. Randomly sample departments, and from the departmental expenditures, identify the QREs in the sampled departments. Apply both the employee and project sub-all rules to qualifying employees and projects within the department. Then extrapolate the post-sub-all QREs found in the sampled departments to estimate total QREs for all the departments in the scope of the study.
While the sampling unit may be a department, this approach should be distinguished from the cursory "departmental" R&D studies that were frequently conducted in the past. In those studies, tax professionals interviewed all department heads about the activity of their staff but did not get to the project/employee nexus sought by the IRS. In this example, for the best audit success, when evaluating the sampled departments, it is important to note that the project/employee nexus requires performing assessment and documentation in these sampled departments—just as any other research tax credit analysis would.
In concept, this same approach works when sampling by subunit or even supervisor (if projects are all contained under one supervisor). So other, more convenient sampling units can be considered.
Group Sampling Units
Sometimes listing expenditures strictly by the company's defined departments may not quite be the perfect solution, as two departments may commonly team up on projects, and one of the two may have more qualified activity. The less qualifying department may not meet a project sub-all rule on its own. However, there is enough activity in the more qualified department that if projects were not split between the two departments, the higher-qualifying department would have enough activity to raise its combined efforts up to satisfy the project sub-all rule, thus picking up additional benefit in the less-qualifying department.
Be flexible. Combine those pairs of departments as though they are only one department when listing them for sampling purposes. If sampled, the expenditures of the two departments are analyzed as though they were only one department when quantifying their QREs and applying the sub-all rules. Again, as long as there is no double counting on the list of expenditures, sampling units can be grouped, sliced, diced, paired, or listed in any convenient manner for sampling.
Conclusion
Slice, dice, split/pair, departmentalize, etc. These are all tools for constructing a sample best suited for the company's organization. These tools can increase the benefit the taxpayer will recognize from the sample. Strive to keep the sampling approach as simple as possible. Complication in a sample design can drive up sample sizes and costs. However, be strategic. Statistical modeling allows several types of samples to be used within a single large population.
There are good reasons to use sampling to estimate QREs. It reduces costs and audit risks without reducing benefits. Achieving the full benefit can require some thought, planning, company knowledge, and good communication among tax professionals, statisticians, and leaders in the company. Work together for the best success.
EditorNotes
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 greg.fairbanks@us.gt.com.
Unless otherwise noted, contributors are members of or associated with Grant Thornton LLP.
