Brief Machine Learning and Tax Enforcement
Janet Holtzblatt, Alex Engler
Display Date
File
File
Download Report
(198.65 KB)

Each year, the Internal Revenue Services receives over 3 billion information returns, such as W-2s and 1099-INTs, from employers, banks, and other entities. The IRS also collects some data about taxpayers from other government agencies and the private sector. But given budget cuts and data limitations, much of the information is underused. In 2021, the Biden administration proposed that a portion of its request for a 55 percent boost (after adjusting for inflation) to the IRS budget over the next decade be used for developing machine learning. If successful, machine learning would marshal the vast trove of data currently received by the IRS to achieve more targeted and productive enforcement actions. Still, the application of machine learning to tax enforcement faces challenges, some of which are inherent to the methodology and others specific to the US tax system, including the complexity of the tax code, the scars of past budget cuts, and the uncertainty of future funding.

Research Areas Taxes and budgets
Tags Individual taxes Taxes and business
Policy Centers Urban-Brookings Tax Policy Center