Wiping the Slate Clean: Individual Response to State Tax Amnesties

Abstract

Listen

Translate article

State governments in recent years have pursued a variety of novel methods to generate revenues. One approach that has become increasingly popular-and controversial-is the tax amnesty. An amnesty typically gives individuals an opportunity to pay previously unpaid taxes without being subject to the penalties and prosecution that the discovery of evasion normally brings. Since 1981, twenty-eight states have offered some form of tax forgiveness to individuals who have failed to pay all their taxes. However, despite the potential significance of amnesties, they are only now beginning to attract much attention. Most of the discussion of amnesties has focused upon the details of their administration [9; 12; 13; 15]. There have now been some theoretical analyses of the benefits and costs of amnesties [2; 3], and, as information from the state programs has become available, there have also been some examinations of the characteristics of those who participate in amnesties and the factors that cause amnesties.' However, the availability of information from those states that have held an amnesty makes examination of two different issues possible. First, do individuals respond in their amnesty compliance decisions to the incentives introduced by an amnesty? Second, how can states use this information to structure an amnesty that encourages maximum participation and generates maximum revenues? It is the purpose of this paper to examine these issues. A standard expected utility model of individual behavior under uncertainty is developed to analyze the individual's response to an amnesty. This model is used to derive a demand for taxes paid in the amnesty, a demand that depends in large part upon the structural features of the amnesty: are known delinquents allowed to participate, is interest on back taxes reduced, are post-amnesty penalties increased, and are greater funds spent on post-amnesty enforcement? This amnesty tax equation is then estimated using aggregate data from the twenty-eight states that have