Modernizing Tax Administration with Predictive Algorithms

This research project plans to leverage artificial intelligence techniques to improve fairness in IRS audit selections while also making certain forms of business tax evasion easier to identify.  

The annual tax gap—the difference between taxes owed by taxpayers and taxes received by the federal government—was projected by the Internal Revenue Service (IRS) to be around $696 billion in 2022. This gap is logistically challenging to address in part because the IRS lacks the technology needed to accurately estimate the risk of taxpayer noncompliance and allocate enforcement resources accordingly.  

Without this technology, the agency instead relies on methods of audit selection which have high rates of inaccuracy, overburdening compliant taxpayers with delayed refunds and exacerbated financial strain because of unneeded audits. Moreover, Goldin and coauthors provided evidence in prior work that legacy audit systems disproportionately selected Black taxpayers for audit even when audits directed elsewhere would have detected more underreported taxes. 

Any effort to modernize these processes will certainly be costly. To make the most of available resources, it is critical that the IRS has the tools it needs to allocate resources efficiently and focus audits where there is serious concern for evasion.  

This timely research project, led by Jacob Goldin, takes a novel approach to supporting the efficiency and fairness of IRS tax enforcement. Using artificial intelligence (AI) techniques, including data-driven machine learning and predictive algorithms, this project aims to significantly improve the accuracy of IRS audit selections.  

Building on Goldin’s prior collaborations with Stanford RegLab and the IRS, this project seeks to strengthen the identification of certain forms of tax evasion among partnerships and other complex “pass-through” business structures. The project will also work to promote fairness in the current IRS audit selection procedures.

Goldin’s work promises to generate crucial insights for optimizing audit resources, enhancing tax compliance, and increasing fairness in the United States tax system.