Determinants of aggregate income-tax-evasion behavior: The case of the US

Banca Nazionale del Lavoro Quarterly Review, Sep 1998 by Cebula, Richard J

In turn, the expected gross benefits from not reporting income or from underreporting income are hypothesized to be an increasing function of the income tax rate (Cagan 1958, Bawley 1982, Tanzi 1982 and 1983, Clotfelter 1983, Slemrod 1985, Pyle 1989, Feige 1994). Presumably, federal income taxation rate measures could effectively take at least three forms, the personal income tax rate (P7), the social security tax rate (S7), and the corporation income tax rate (C7).

The specification in (8) constitutes the basic model for the empirical estimates provided in Sections 4 and 5 below.

3. The data

In this analysis, three income tax rate measures are investigated: (1) the average effective federal personal income tax rate (AEPI?*, (2) the average effective social security tax rate (AESS*, and (3) the average effective federal corporation income tax rate (ACS). In addition to AEPIT, AESST and ACIT, the variable AUDIT, which is the percentage of filed federal income tax returns in the US that has actually been subjected to an IRS audit (investigation) in each year, is included as a measure of the expected likelihood of being subjected to an IRS audit. The variable PEN, which is the total pecuniary penalty (inclusive of both penalties per se plus interest) assessed by the IRS per dollar of reported AGI in each year, is included to reflect the penalty (above and beyond added tax liabilities per se) from not reporting income if said activity is detected. As observed above, the variables AUDIT and PEN are adopted in this study, based on arguments found in Pestieau, Possen and Slutsky (1994), Alm, Jackson and McKee (1992), and Errard and Feinstein (1994), as identifiable and quantifiable measures of risks associated with underreporting income. The variable DIS is represented by the `dissatisfaction index'. This index is constructed as an equally weighted average of three normalized indices reflecting answers to the University of Michigan's Institute for Social Research (ISR) surveys concerning whether government officials can be trusted (to honor obligations to the public), whether they are dishonest, and whether government wastes tax dollars. Values for this index of dissatisfaction lie within a range of (-1.5), which corresponds to least dissatisfied, to ( 1.5), which corresponds to most dissatisfied, so that the algebraic value of this index is higher as the public is more dissatisfied with government.

Naturally, there are other possible measures of the federal personal income tax rate that could have been adopted rather than AEPIT (the average effective federal personal income tax rate), including perhaps the maximum marginal federal personal income tax rate, MMPIT. Nevertheless, in effect paralleling Feige (1994, p. 135),2 we focus on the variable AEPIT. We adopt a view that, given the complexity of the Internal Revenue Code and the variety of marginal tax brackets in the Internal Revenue Code, a variable such as AEPIT may be a reasonably useful (albeit only proximal) measure for tax filers generally of tax benefits from underreporting income. By contrast, the MMPIT variable can be viewed as too narrow (and hence irrelevant) for most of the income spectrum and thus potentially as neglecting a very large portion of taxpayers. Essentially paralleling Feige (1994), we define the variable AEPIT as the ratio of total federal personal income tax collections to aggregate reported AGI, expressed as a percentage; we adopt variable AEPIT as the personal federal income tax rate measure/proxy. Similarly, the focus on (choice of) variables AESST and ACIT, as defined, as the measures of the social security tax rate and the federal corporation income tax rate, respectively, also parallels the approach in Feige (1994). The data for variables AEPIT, AESST and ACIT were obtained from the IRS (1971-1996) and the Council of Economic Advisors (1997).