Use of Sampling or Modeling Approach for Cost Segregation
One of the acceptable methods of completing a cost segregation study includes the use of statistical sampling. According to the IRS Audit Techniques Guide, one method for completing a cost segregation study is utilizing a Sampling or Modeling Approach. However, there are many areas of consideration when completing a sampling review.
First, it is important to understand the difference between a judgment sample and a statistical sample. In most cases, when preparing for a sample, taxpayers and practitioners want to pick a judgment sample. A judgment sample is a nonrandom sampling approach where records are chosen based on opinions of experts or based on a bias. Judgment sample selection is subjective, and some items have a zero chance of selection. For example, when reviewing a portfolio of homes, looking only at homes that are empty or within a 50-mile radius of a central location would be considered a judgmental sample. In a true statistical sample, every record has a nonzero chance of selection. The IRS has no guidance on judgment sampling, and this is not considered an allowable methodology under audit. The IRS will accept a true statistical sample and the rules for this sampling are covered in Revenue Procedure 2011-42.
IRS Revenue Procedure 2011-42 specifically covers the terminology and requirements related to utilizing a statistical sample. The first term that is important to understand is Relative Precision. Relative precision measures the accuracy of the sample and is calculated by dividing the point estimate by the sampling error. For example, a sample that provides a value of $5 million plus or minus $1 million would have a relative precision of 20% ($1million/$5million). This number is important as the IRS requires using the least advantageous confidence limit if the precision is 15% or worse. If the relative precision is 10% or better, the IRS allows the taxpayer to utilize the point value. Take a cost segregation estimate where the taxpayer has 10% personal property plus or minus 2%. In this case, the taxpayer would only be able to claim 8% personal property. However, if the taxpayer can refine their estimate to 10% personal property plus or minus 0.5%, they could use the point value of 10% in calculations.
It is also important to note that the relative precision needs to be calculated at a 95% confidence level. To use the above example where we have 10% personal property plus or minus 0.5%, we would need to calculate that to a confidence level of 95%. This means that we would expect the results to be 9.5% to 10.5% when calculated 95% of the time. Typically, when increasing the confidence level the Relative Precision becomes greater. The result of this is that a larger sample set is needed to provide a 95% confidence level with a relative precision of 10% or less.
Wide variations in a sample set might require the sample to be significantly larger. A real estate portfolio with homes may have some homes with carpet, and others with hardwood floors. This could cause a variation in the sample set leading to a larger variation in the precision. Whereas a portfolio of self-storage warehouses all built by the same contractor, to the same plans, may have a smaller relative precision based on a smaller sample set.
One difficulty occurs when the initial assumptions do not match the findings. A sample set may be created assuming a relatively tight distribution of the assets, however, when the study is completed, the results do not match the assumption. This could lead to a larger relative precision than initially assumed, making the deviation much larger, and potentially making the study not as valuable. To illustrate, consider a scenario where the initial assumptions assume that 15% of the property could be classified as personal property, and that a sample of 50 locations could back this up. However, when the study is complete the results come back at 15% plus or minus 5% to a 95% confidence. Since the relative precision is so high (33.33%), the least advantageous limit must be utilized. This would lead to a result of 10% personal property, leaving 1/3 of the savings on the table.
While sampling can be a valuable tool in larger portfolios, it is critical to understand the risk and pitfalls associated with this methodology. Often the perceived savings associated with sampling do not materialize due to the complexity of the requirements.