New IRS Guidance Boosts Statistical Sampling
Mary Batcher, Ed Cohen, Ryan Petska, and Wendy Rotz
Federal - 12/21/2011

At the end of August, the Internal Revenue Service issued Revenue Procedure 2011-42, which provides guidance to taxpayers using statistical sampling to support a tax return position.[1] The guidance in the revenue procedure is not substantially different from guidance provided in earlier field directives.[2] The most important difference is that the revenue procedure provides a safe harbor for companies who wish to use statistical sampling as part of their tax filings.

With the issuance of the revenue procedure, companies do not have to wonder or ask their IRS Examination team whether statistical sampling will be accepted in a situation that is not explicitly covered in other guidance. As long as the statistical sample is in accordance with the revenue procedure, it will be acceptable to the IRS. This means taxpayers can claim deductions or credits that they may not have had the resources to pursue if they did not use sampling.

History of Tax Sampling Guidance

Before the release of the 2002 Field Directive on the Use of Estimates from Probability Samples, the IRS only sanctioned the use of statistical sampling in specific areas. Revolving credit, trading stamps, basis studies, and LIFO were among the few areas that were addressed in IRS guidance. Taxpayers were using statistical sampling for other purposes without the IRS’s explicit approval or after approaching their IRS exam team to seek ad hoc approval for their proposed sampling. Recognizing the value to the IRS while limiting the burden to taxpayers, the 2002 field directive was released. A benefit to the IRS is that a properly designed statistical sample narrows the scope of what the IRS needs to examine. After 2000, a series of revenue procedures or guidance were released relating to statistical sampling for Qualified Intermediaries, U.S. Withholding, Meals and Entertainment, Basis Studies, Section 199, and Utility Repairs.

Coordination with Field Guidance

The IRS has taken the additional step of revising its internal sampling guidance in the Internal Revenue Manual (IRM) so audits will be consistent with the guidance provided to taxpayers. That means companies can expect the IRS to generally follow the procedures outlined in Revenue Procedure 2011-42.

The practical consequence of this synchronization of the IRM with the Sampling Revenue Procedure is that, if the taxpayer sample is consistent with the revenue procedure, the IRS will generally use it in its audit of the taxpayer. That limits the scope of the audit for both the taxpayer and the IRS to a common set of items. Since the sample was designed and implemented before filing the return, the taxpayer is not rushing during the audit to retrieve supporting documentation since it would have already been retrieved, making the audit a smoother less stressful process for both parties.

There is, however, an exception to the synchronization of the guidance between the revenue procedure and the IRS’s internal guidance for its auditors. Specifically, the IRM guidance allows the IRS to combine multiple years into a single sample without any stated limitations on the number of years. The Meals and Entertainment Revenue Procedure allows taxpayers to combine up to three years in a single sample. That has been extended informally to other applications, but there is no explicit recognition of the appropriateness of multi-year samples in the new sampling revenue procedure. The practical consequence of combining multiple years into a single sample is that the total sample size across all the years at issue is greatly reduced, allowing taxpayers to claim three years of deductions with the same level of effort required for a single year.

Scope of Tax Sampling

Consistent with earlier statistical sampling guidance, Revenue Procedure 2011-42 provides that the propriety of the use of a statistical sample is determined by several factors including the time and cost that would be required to review the data and make a determination if sampling were not used. Although the difficulty of individually reviewing thousands of transactions is clear, the complexity and cost of the analysis of even small populations also makes sampling appropriate. For example, the complexity of the analysis of capitalized components of a building to shorten tax lives of the assets would make sampling appropriate, even if there were only a hundred or so buildings, whereas, it would not be appropriate to sample a population of 100 expenses to identify misclassified 50-percent limited expenses that could be fully deducted.

Sampling is not appropriate, under the guidance, if other evidence is readily available that can provide a more accurate answer. There are instances where books and records provide part of the information or a good starting point but in depth research is needed to make a final determination. For example, the research expenditures identified for financial statement R&D purposes may not meet the requirements for the R&D tax credit. A sample of those financial statement R&D projects is an appropriate tool to use to identify a reduced number of projects to thoroughly review, document, and quantify the qualified research activities. The qualified research costs of those projects are then statistically projected to the population they were selected from.

Statistical sampling can also be used in a confirmatory fashion, to test whether, for example, all the assets in an account designated for repairs truly are repairs that meet tax requirements.

Revenue Procedure 2011-42 provides broad acceptance for statistical sampling but does not support non-statistical or judgment sampling. Judgment sampling includes any sampling approach where the selection of items into the sample is not controlled by a random mechanism like computer generated random numbers. Sometimes people refer to haphazard sampling, where items are selected by taking a few here and there from various places in the population, as “random” but that process does not meet the requirements of statistical randomness. Similarly, the use of a skip interval to select the sample is not random unless it has a randomly chosen starting point. The use of judgment sampling for tax purposes carries serious risk that the sample and all projections from it will be disallowed in audit.

Areas Where Sampling Is Used

Sampling has been widely used by companies to establish research credit claims, to identify deductions under section 174 of the Code, to identify eligible repairs under section 162, to find deductible meals and entertainment expenses that were incorrectly limited to claiming only 50 percent of the expenditure under section 274, and to claim a domestic production deduction under section 199. Most of these situations are covered by revenue procedures that explicitly allow statistical sampling or are in areas where IRS Exam teams have been using sampling for many years.

With the release of Revenue Procedure 2011-42, sampling in less established areas can be considered without concern that the IRS will reject the use of sampling for that purpose. Typical statistical sampling applications involve issues where the books and records do not readily lend themselves to an answer and a study would be required. Some examples include estimating bad debts, asset valuation, lobbying expenses, inventory, excise tax refunds, casualty losses, compliance for pension plans, and basis studies.

The sampling and estimation methods described in Revenue Procedure 2011-42 will work quite well for most situations where a company might want to use a statistical sample for tax purposes. There are situations, however, where the methods described in the revenue procedure may not be adequate. For example, a company with decentralized records in multiple locations may need to sample locations first and then sample records within each of the sampled locations. This is a well-established and valid type of sampling called two-stage sampling, but it is not covered by the revenue procedure. The guidance does recognize that there are other acceptable statistical methods that are not covered by the revenue procedure. In such situations the sample would be referred to one of the IRS statistical sampling coordinators who have additional sampling training. Experience teaches that a valid statistical sample meeting the criteria for proper use of sampling will not be disallowed just because it is different from the methods outlined in the revenue procedure and its predecessor field directive.


The issuance of Revenue Procedure 2011-42 invites companies to use statistical sampling for tax purposes wherever it makes sense to do so. The accompanying coordination of the internal IRS sampling guidance means that taxpayers can expect the IRS to follow the same procedures and requirements described in the revenue procedure with only the exceptions noted. In fact, if the taxpayer sample is consistent with the revenue procedure, the IRS will use it to perform its audit rather than design a whole new sample, saving the taxpayer the additional effort of retrieving supporting information for a whole new set of items. The assurance of acceptance of the sample for taxpayers and the streamlining of the IRS audit makes this a win-win for both taxpayers and the IRS. For taxpayers, statistical sampling can reduce the cost of compliance, speed the time involved in determining deductions, and in many cases increase the accuracy and quality of the tax submission.

Mary Batcher is the National Director of Statistics and Sampling in Ernst & Young's Tax Department. She has successfully directed many sampling projects for tax filing purposes and has worked with the Internal Revenue Service to expand its acceptance of statistical sampling. Ms. Batcher formerly was the chief of a statistical support group at the IRS. She holds a Ph.D. degree in statistics from the University of Maryland. She is accredited by the American Statistical Association. She may be contacted at

Ed Cohen, a certified public accountant, is a Senior Manager at Ernst & Young. He has more than 20 years of statistical consulting experience in sampling and estimation. While at Ernst & Young, Mr. Cohen has worked extensively in tax sampling; before that, he served as a statistical sampling coordinator at the Internal Revenue Service where his responsibilities included tax audit statistical sampling and sampling guidance. Mr. Cohen holds a Bachelor's degree in accounting from Pace University. He may be contacted at

Ryan Petska is a Senior Manager at Ernst & Young. He has more than 10 years of statistical consulting experience in sampling and estimation. While at Ernst & Young, Mr. Petska has worked extensively in tax sampling; prior to that, he served as a statistics instructor at Florida State University. Mr. Petska holds a Master’s degree in statistics from Florida State University. He may be contacted at

Wendy Rotz is a Senior Manager at Ernst & Young. She has 19 years of experience working with tax related statistical estimation. Before joining Ernst & Young, Ms. Rotz was a mathematical statistician at the Internal Revenue Service, where she worked on a variety of statistical studies including the Taxpayer Compliance Measurement Program. She holds a Master’s degree in mathematics from Eastern Michigan University. She may be contacted at


1. Rev. Proc. 2011-42, 2011-37 I.R.B. 318 (September 12, 2011). A return position would include income, expenses, deductions, and credits as they relate to an original return or claim.

2. A description of the 2002 and 2009 sampling field directives can be found in Mary Batcher, Eric Falk, and Wendy Rotz, New IRS Guidance Makes Statistical Sampling More Attractive than Ever for Federal Tax Purposes, 62 Tax Executive 37-39 (January – February 2010).