Modeling Expatriate Tax Evasion Using Unsupervised Machine Learning

Abstract

The level of tax noncompliance amounts to a very high sum globally. For the United States alone, there is a projected tax shortfall of $600 billion US dollars per year, which is $7 trillion over ten years, not adjusted for inflation [1]. Under the current system, it is estimated that 15% of individual taxes are uncollected [2]. This research aims to model the phenomenon of tax avoidance and evasion through the administration and collection of 67 survey responses and 22 expert interviews, collected by the researchers. While tax evasion is criminally illegal and tax avoidance is more about exploiting existing loopholes in the code, both result in less revenue being collected. Understanding the current state of tax evasion and avoidance for expatriates is important, given the global shortfall in tax revenue.