Page 4
projects is too large to conduct a full review, then statistical sampling should be used in
examining the research credit issue.
(d) Statistical sampling is particularly useful in the typical case involving
a few large projects and many small projects.
In many research credit cases, a significant portion of the expenses are incurred
in a few large projects, with the remaining expenses allocated to many small projects. If
the agent only examines the large projects, then there is no legally sustainable basis for
adjusting the small projects. In this instance, the agent would have to allow the smaller
projects, even though there is a greater likelihood of finding a basis for disallowing small
projects. This is particularly true in internal use software cases, since taxpayers have
the burden of proving that they committed substantial resources to the software
development. Norwest v. Commissioner, 110 T.C. 454, 499 (1998), appeals pending
sub nom., Wells Fargo & Co. v. Commissioner (8th Cir. Nos. 99-3878, 99-3883, 99-
4071). Also, taxpayers usually have more documentation relating to their largest
projects, and probably have more employees available to provide information about
these projects. It is therefore particularly important to consider statistical sampling of the
smaller projects. In fact, sampling the smaller projects is the only way of ensuring a
legally sustainable basis for disallowance where the agent does not review these
projects in full.
3. Should the examination be limited to only the largest projects?
No. Where there are only a few large projects these projects should be examined
in full and not be included as part of the statistical sample. The statistical sampling
approach should be used to resolve all but the large projects. Taxpayers will often
propose, however, that the audit strategy simply involve reviewing a few large projects
(i.e., the top five). This approach is generally unacceptable because it would result in
allowing the expenses paid or incurred in the smaller projects. Alternatively, a taxpayer
may propose a review of the large projects, with the results projected to the population
of smaller projects. This approach would not, however, provide a statistically sound
result since the largest projects are generally more likely to meet the requirements for
qualified research, thus skewing the results in favor of allowing a larger credit than if the
agent reviewed the smaller projects.
Where a taxpayer has projects of varying sizes, the recommended approach is to
conduct a statistical sample using stratification. For example, a taxpayer's small projects
(e.g., projects with up to $500,000 in research expenses) may be grouped into the same
stratum, and if there are too many projects in this category to review in full, a sample of
projects from this group would be reviewed, with the results projected in a statistically
valid manner. The next question examines stratification in more detail.