Causal Seminar: Panos Toulis, University of Chicago

The reading group will meet at 2PM before the seminar.

Network Experimental Designs for Tax Audit Policies

In this paper, we present an ongoing large field experiment in collaboration with the Tax Authority in a country of South America and the Inter-American Development Bank. The experiment randomizes tax audit notices (treatment) to firms connected through a large network determined by inter-firm VAT transactions. While the ultimate goal is to optimize tax audit policy, the short-term goal is to estimate causal effects of tax audit notices on firm behavior. Of particular interest is to understand spillovers, that is, the response of firms that are not treated but are connected to other firms that are treated. We argue that current popular approaches to experimenting on networks are limited by the reality of inter-firm networks, such as their size, high interconnectivity and heavy-tailed degree distributions. Our approach to experimentation leverages subtle sub-structures in the network and allows the application of Fisherian-style permutation tests of causal effects. These testing procedures can be computationally efficient and finite-sample valid, qualities that are important for testing in a robust way the treatment effects from complex policy interventions.

Discussant

• Tu Ni (Laboratory for Innovation Science at Harvard)