Key Takeaways
- $7.7 billion in additional tax liabilities were identified by the IRS under the Offshore Voluntary Disclosure Program (OVDP) during the 2013–2014 period
- $3.8 billion in additional tax liabilities were identified by the IRS under the OVDP during the 2012–2013 period
- $46 billion of the estimated 2019 tax gap was attributable to “failure to pay” (midpoint estimate)
- Taxpayers were 2.6 times more likely to voluntarily disclose with the increased likelihood of detection from IRS enforcement changes (experimentally measured in a tax compliance study published by the Journal of Economic Behavior & Organization, 2020)
- In a randomized survey experiment, 23% of participants reported they would underreport income if the probability of detection was low versus 9% when the probability was high (behavioral study in 2019, Journal of Public Economics)
- In 2022, the IRS identified 1.9 million cases with potential identity or refund fraud using information returns and analytics
- The United States received $9.9 billion in tax revenues from foreign tax authorities under information exchange agreements in 2022 (per OECD Global Forum reporting)
- In 2022, the number of CRS (Common Reporting Standard) automatic exchanges received by the U.S. was 3.5 million reports
- A 2022 U.S. Senate report found that the trade-based money-laundering ecosystem involves over $1 trillion annually globally, with a substantial portion linked to evasion of government obligations including taxes
- In a report covering 2018–2022, abusive tax shelter promoters were linked to $6.6 billion in claimed deductions and credits that were later challenged by the IRS
- 0.79% of U.S. firms reported paying any tax assessed by the IRS in 2015, but 31.1% reported making payments to “tax havens” (as captured in confidential data used by researchers), indicating mismatch between reported operations and tax outcomes
- In 2023, the IRS reported receiving 2.0 million documents from third parties (information returns and statements) that support mismatch detection for taxes
- The Internal Revenue Manual states that “correspondence exams” constituted the IRS’s largest audit activity category, with 5+ million actions annually in recent years, enhancing detection of filing and reporting errors
- In 2018–2021, the U.S. Government accountability reporting estimated that the IRS was conducting ~1.3 million audits annually (including individual and business exams), affecting noncompliance detection
- The IRS estimated that the audit rate for individual returns declined to about 0.45% in 2019, changing the expected probability of detection for tax underreporting
From offshore disclosures to audit and fraud analytics, IRS enforcement increasingly detects underreported taxes.
Related reading
01 · Category
Tax Gap Estimates3 stats
Tax Gap Estimates Interpretation
02 · Category
Compliance Signals3 stats
Compliance Signals Interpretation
03 · Category
Global Risk & Schemes3 stats
Global Risk & Schemes Interpretation
04 · Category
Abusive Schemes1 stats
Abusive Schemes Interpretation
05 · Category
Tax Compliance5 stats
Tax Compliance Interpretation
More related reading
06 · Category
Enforcement Activity4 stats
Enforcement Activity Interpretation
07 · Category
Money Laundering Links2 stats
Money Laundering Links Interpretation
08 · Category
Identity & Refund Fraud1 stats
Identity & Refund Fraud Interpretation
09 · Category
Industry Context1 stats
Industry Context Interpretation
10 · Category
Trade & Data Leakage2 stats
Trade & Data Leakage Interpretation
Cite This Report
This report is designed to be cited. We maintain stable URLs and versioned verification dates. Copy the format appropriate for your publication below.
Gabrielle Fontaine. (2026, February 13). U.S. Tax Evasion Statistics. Gitnux. https://gitnux.org/u-s-tax-evasion-statistics
Gabrielle Fontaine. "U.S. Tax Evasion Statistics." Gitnux, 13 Feb 2026, https://gitnux.org/u-s-tax-evasion-statistics.
Gabrielle Fontaine. 2026. "U.S. Tax Evasion Statistics." Gitnux. https://gitnux.org/u-s-tax-evasion-statistics.
Sources & references
25 datasets cited across this report · attribution is report-level
+12 additional datasets cited (not shown individually)

