Undetected Income: Identification, Estimation, and Uncertainty

• In Event: Tax Evasion and Enforcement: Measurement and Estimation

Abstract

Tax authorities rely on audits to estimate population measures of tax compliance. However, identifying the amount of taxes owed but unpaid (the “tax gap”) is complicated by the fact that audit outcomes may vary across examiners, even for returns with the same underlying compliance. When this variation in examiner outcomes reflects differences in skill, it can be leveraged to estimate the undetected portion of unreported taxes. This paper studies this class of estimation problem, proposing a new method that has three advantages over existing solutions. First, transparency: we introduce a Detection – Under-Reporting Frontier that is identified by the data, and the additional assumptions that are needed to identify an estimate along the frontier. Second, efficiency: our Bayesian estimation procedure regularizes imprecise examiner estimates and allows for the extraction of signal from more sources than existing alternatives. Third, we provide uncertainty estimates, which existing methods rarely do. We implement our method using IRS random audit data.

Authors

• Patrick Vossler

• Aviv Caspi, Stanford University

• Jonathan Hennessy

• Zaynah Javed

• John Guyton, Internal Revenue Service

• Andrew Johns, Internal Revenue Service

• Jacob Goldin, University of Chicago Law School

• Daniel Ho