Christine Kim (Cardozo) presents Algorithmic Tax Ownership (co-authored with Dmitry Erokhin) at Georgia today, as part of its Tax Policy Colloquium Series hosted by Assaf Harpaz:

Tax ownership is a crucial concept for determining tax liabilities, compliance, and enforcement. However, neither the courts nor the IRS have provided clear guidance on how to analyze it. Since the Supreme Court first outlined a twenty-six-factor test for determining tax ownership in Frank Lyon Co. v. United States in 1978, this multifactor test has remained largely unchanged, and there has been no further guidance from the courts or the IRS to this day. Even tests with shorter lists of factors only add to the confusion regarding compliance and enforcement, as there is no clarity on which factors are most dispositive. 

This paper provides a better understanding of tax ownership analysis by addressing this question with machine learning algorithms. After analyzing 2,000 IRS rulings and court cases related to sale and repurchase transactions, this paper’s algorithmic model found that a single factor of identifying who bears the economic benefits and burdens of the underlying asset can predict a tax ownership outcome with 94.6% accuracy. Additionally, this paper demonstrates how the proposed model and its findings could be applied to other types of transactions, providing various stakeholders with an advanced tool to assess and refine their decisions without overhauling existing substantive case law. 

This paper’s analysis of multi-factors in tax ownership may resonate with broader audiences who are studying multi-factor tests in another area of law. Also, the comparison of the proposed algorithm using classic machine learning with regression, natural language processing, and large language model would inspire legal scholars exploring cutting-edge methodologies. 

For more information on the Tax Policy Colloquium, please contact Assaf Harpaz.