Using Machine Learning to Evaluate the Existence of a Trade or Business: Olsen

Abstract

The need to determine whether a taxpayer is engaged in a “trade or business” arises frequently. Although the term “trade or business” is used throughout the IRC and the associated regulations in the Code of Federal Regulations, it is not defined explicitly. In some ways this is surprising. Crucially, establishing that there is a trade or business affects whether a taxpayer can deduct business expenses. Moreover, the phrase “trade or business” is relied on for many other taxpayer rights and obligations. Erring about whether a taxpayer is carrying on a trade or business can have wide-ranging financial consequences for deductions, credits, exemptions, disqualifications, and penalties, among other things.

Here we examine how machine learning can be used to assess the strength of the taxpayer’s position in the appeal of the Tax Court’s decision in Olsen. Based on the facts accepted by the Tax Court and the arguments advanced in the appellant’s opening brief, Blue J’s algorithm predicts with greater than 95 percent confidence that the Tenth Circuit will find that the taxpayers were not engaged in a trade or business. We appreciate that a predictive analysis with that high a degree of confidence is unusual for a matter on appeal; we therefore also examine why this may be the case.

Keywords: machine-learning, AI, tax, trade or business

JEL Classification: J00, H2, H00