Collaborative Research on the Relationship between Income Tax Auditing and Compliance

人们普遍认为不纳税是一个极其严重的问题。最近的估计表明，每年至少有900亿美元的应税收入没有申报——平均占总应税收入的10%到15%。对税收合规性下降的担忧引发了政策和学术写作的爆发，但其原因和适当的应对措施仍存在争议。早期的税收合规文献将税收合规视为一个“投资组合问题”，在发现概率固定的假设下，推导出“风险资产”（未报告收入）的最优消费规则。这些模型产生了模棱两可的结果。最近的理论创新试图从决策理论的框架转移到委托-代理问题和博弈论模型。在这些方法中，美国国税局（IRS）被允许采取战略行动，根据从纳税人那里收到的信息来调整其审计规则。这些互动模型是由Wilde、Reinganum和Graetz在之前的国家科学基金会资助下开发的，它们对IRS使用的均衡审计规则和纳税人使用的均衡报告规则的性质进行了预测。本研究的目的是检验一个博弈论模型，并为审计与合规之间的关系提供实证证据。本文分析了三个不同的数据集：(a)与1969年个人联邦所得税申报表相关的横截面数据集，这是由美国国税局在70年代收集的；(b) 1969-1985年期间与国家级变量相关的横截面时间序列数据集；(c)与1979年TCMP个人报税项目有关的数据集，该数据集最近由美国国税局收集。需要解决的主要问题是，审计费率是否以及以何种方式对纳税人的合规行为作出互动反应。IRS作为战略行为体的引入被认为是对一般执法经济理论的重大改进。税收不合规的行为框架使其成为对明确考虑到认知执法机构的相互作用和反应的理论结构进行实证检验的特别合适的案例。这项研究具有重大的科学意义，因为它不仅有望对一个重要的政策问题提供见解，而且还有力地增加了对产生遵守法律的条件的基本知识。

Tax noncompliance is widely recognized to be an extremely serious problem. Recent estimates suggest that at least $90 billion of taxable income goes unreported annually--an average of 10 to 15 percent of total taxable income. Concern about declining tax compliance has produced an outburst of policy and academic writing, but both the causes and the appropriate responses remain controversial. Early tax compliance literature treated tax compliance as a "portfolio problem," deriving rules for optimal consumption of the "risky asset" (unreported income) under the assumption that the probability of detection is fixed. These models produced ambiguous results. More recent theoretical innovations have attempted to move out of the decision-theoretic framework to a principal-agent problem and game-theoretic model. In these approaches the Internal Revenue Service (IRS) is allowed to act strategically, conditioning its audit rules on the information it receives from taxpayers. These interactive models, developed by Wilde, Reinganum and Graetz under a previous NSF grant, yield predictions about the nature of the equilibrium audit rule used by the IRS and the equilibrium reporting rule used by taxpayers. The purpose of this research is to test a game-theoretic model and to develop empirical evidence on the relationship between audits and compliance. Three distinct data sets are analyzed: (a) a cross-section data set related to 1969 individual Federal Income Tax returns, which was assembled by the IRS in the seventies; (b) a cross-section, time series data set related to state-level variables for the period 1969-1985; and (c) a data set related to 1979 TCMP individual return items, which was recently assembled by the IRS. The major question to be addressed is whether, and in what fashion, audit rates respond interactively to the compliance behavior of taxpayers. The introduction of IRS as a strategic actor is considered to be a significant improvement in the economic theory of law enforcement generally. The behavioral framework of tax non- compliance makes it an especially appropriate case for an empirical test of the theoretical construct that explicitly takes into account the interactions and responses of the cognizant law enforcement agency. The research is of substantial scientific significance in that it promises not only to offer insights on an important policy issue but also to add powerfully to fundamental knowledge of the conditions that engender conformance to law.