Attacking Networks of Tax Evasion: Theory and Evidence

Awardees

Michael Best
Assistant Professor
Evan Sadler
Assistant Professor

$362,577

Ability to raise sufficient tax revenues efficiently and equitably to fund public services is one of the central challenges in economic development. This problem arises partly because of inability to create the capacity to enforce taxes effectively, thus leading to increased tax evasion. Reducing tax evasion requires a detailed understanding of the drivers of tax evasion and the optimal allocation of scarce tax enforcement resources. This research combines new theory and other innovative research methodologies to provide new insights into tax evasion by firms, the strength of enforcement spillovers through production networks, and how best to target enforcement activities. Working in partnership with tax authority, the project will help to increase the capacity to reduce tax evasion and to generate and use evidence on policies’ impacts. The results of this research project will inform policies to improve tax administration in the US to make it fair, efficient, reduce the deficit, and improve the allocation of resources. The lessons from this research can then serve as evidence to inform decisions in other countries.

This project has two parts. In the first part extends the canonical model of tax evasion: firms make reports of their sales to, and purchases from, other firms in a production network, creating reporting spillovers. We derive the optimal targeting rules the government should use to minimize total tax evasion. The second part of the project leverages rich administrative tax return data on tax liabilities and production networks to perform two RCTs in collaboration with tax administrations. The first experiment will randomly deploy electronic invoicing and desk audits focused on discrepancies between trading partners' reports to learn the strength of their direct effects on targeted firms and their indirect effects on targeted firms' suppliers and clients. The project then takes these estimates and use them to calibrate the model and the second experiment tests the model’s optimal targeting rule against the status quo and random targeting. The results of this research project will inform policies to improve tax administration in the US to make it fair, efficient, reduce the deficit, and improve the allocation of resources. The lessons from this research can then serve as evidence to inform decisions in other countries.