Computer Aided Tax Avoidance Policy Analysis

This thesis presents a three part methodology for analyzing the flow of taxable income in large partnership structures. The method forms the basis for prototypical software which would clarify many complicated basis adjustment issues associated with partnership taxation. Partnerships, the most common form of “flow-through” tax entities, have rapidly increased in size, complexity and economic relevance between 2005 to 2015, as well as resulting in an estimated $91 billion in underreported income. Many of these partnerships have upwards of one million direct and indirect partners, as well as 100 tiers of additional large partnerships. This surge in the number of partnerships, combined with the highly complicated nature of US partnership taxation law, requires novel techniques to evaluate the tax consequences of increasingly complex financial activity. A computational methodology is presented in this thesis for understanding and analyzing the allocation of taxable income in large partnership structures, with particular focus on characterizing abusive tax behavior. First, a formal notation is established to fully describe how taxable income is allocated in partnerships, forming the basis of a functioning partnership tax calculator. Next, a simulation is described that processes transaction sequences through partnership structures, as well as a method for assigning audit likelihood to potentially suspicious combinations of financial activity. Finally, a means by which to optimize a) transaction sequences that minimize both tax liability and audit likelihood and b) auditing procedures that characterize abusive tax behavior in a compact form is established. The proposed methodology offers taxpayers, auditors and policy-makers a computational approach to resolve uncertainty in partnership taxation, lower the cost of the auditing process through automation and provide a conceptual exploration of tax policy implications. Thesis Supervisor: Una-May O’Reilly Title: Principal Research Scientist Thesis Supervisor: Erik Hemberg Title: Postdoctoral Associate