Content area
Full Text
ABSTRACT
This study investigates whether the nonrandom element of human behavior could facilitate the detection of tax evasion. Unplanned Evasion (UPE) is defined to be blatant manipulation by the taxpayer of line items at filing time. Planned Evasion (PE) is the result of planned actions to conceal an audit trail. UPE requires that the taxpayer invent a number(s) for the line item(s).
Benford's Law (Benford 1938) is used as an expected distribution for the digits in tabulated data. The assumption is that the digits of data that are truthfully reported, or are subject to PE, should conform to the expected digital frequencies. A Distortion Factor model that quantifies the extent of UPE is developed. Tax returns on the U.S. Internal Revenue Service Individual Tax Model Files are analyzed. The analysis, based on digital frequencies, indicates that Low Income taxpayers practice UPE to a greater extent than High Income taxpayers.
The U.S. Internal Revenue Service (IRS) estimated that the individual gross tax gap (the difference between income tax due and income tax voluntarily paid) for 1992 would amount to $91-94 billion (IRS 1990). Jackson and Milliron (1986, 15) describe the state of compliance understanding as "primitive" and Roth et al. (1989, 247) believe that our base of compliance knowledge is "small compared with the importance of the phenomenon." More recently Alm and McKee (1992, 107) claim that despite a decade of empirical analysis "our understanding of compliance remains surprisingly limited." Long and Swingen (1991, 664) stress a need to focus on specific types of noncompliance. The objective of this study is to introduce and identify a type of evasion described as Unplanned Evasion (or filing-time evasion), to increase our understanding of the evasion phenomenon.
Building on the work of Slemrod (1985), recent studies by Christian and Gupta (1993) and Dusenbury (1993) take a new approach to compliance research by drawing inferences from unaudited taxpayer data. Similarly, the basis of this study is that the digital frequencies of numbers reported in tax returns will deviate from the expected frequencies due to taxpayer evasion. The expected frequencies are those that would be observed in the event of truthful reporting, and are assumed to be governed by Benford's Law (Benford 1938). This idea is not new in...