Tax Evasion and Money Laundering: Insights from Textual and Contextual Analysis

Abstract

This study examines whether tax evasion operates as a predicate offense for money laundering by integrating an economic model of concealment with an analysis of 4,534 IRS Criminal Investigation (IRS-CI) prosecutions from 2021 through mid-2025. It addresses the gap between the Financial Action Task Force (FATF) framework, which treats serious tax crimes as predicate offenses for money laundering, and US law, which often treats underreported taxes as noncriminal. Building on the Allingham-Sandmo (1972) model and recent work by Kemsley and Morgan (2022, 2024), the study introduces a convex concealment cost function and distinguishes standalone tax evasion from sequential evasion-and-laundering behaviors.

Based on an analysis of 4,535 IRS CI prosecution news releases, cases were classified into three categories: tax evasion charges, money laundering charges, and joint prosecutions involving both offenses. The results show that 2,727 cases, approximately 60 percent, involve both tax evasion and money laundering charges. LIFT, Pointwise Mutual Information (PMI), Normalized Pointwise Mutual Information (NPMI), and Fisher's Exact Test revealed links between specific forms of tax evasion, such as corporate pass-through schemes, payroll tax violations, and false returns, and distinct laundering mechanisms, including wire transfers, shell entities, bulk cash movements, real estate acquisitions, and cryptocurrency transactions. Overall, these findings suggest that current U.S. enforcement practice commonly treats tax evasion and money laundering as separate but interconnected offenses, supporting the view that tax evasion serves as a predicate offense in a two-step process that aligns more closely with the FATF approach than with models that equate tax evasion directly with money laundering. Collectively, this study bridges the gap between economic theory and enforcement practice. By demonstrating that U.S. prosecutors treat tax evasion and money laundering as distinct, sequential offenses, the findings refine the single-stage models proposed by Kemsley and Morgan (2022, 2024) and validate the FATF two-stage framework. Furthermore, by identifying these behaviors at the subtype level, this study provides practical benchmarks to support Anti-Money-Laundering (AML) detection and strengthen cross-border enforcement.