Maximilian Kasy

Optimal taxation and insurance using machine learning

How should one use (quasi-)experimental evidence when choosing policies such as tax rates, health insurance copay, unemployment benefit levels, and class sizes in schools? In a recent article, Maximilian Kasy suggests an approach based on maximizing posterior expected social welfare, combining insights from optimal policy theory as developed in the field of public finance, and machine learning using Gaussian process priors. The key empirical relationship the policy maker needs to learn about is the response of health care expenditures to coinsurance rates. Holding the economic model and distributive preferences constant, he obtains much smaller point estimates of the optimal coinsurance rate (18% vs. 50%) when applying his estimation method instead of the conventional “sufficient statistic” approach.

Maximillian Kasy’s current research falls into three main categories: 1) “Statistics in a social context” – rethinking issues of statistical methods as communication between interested parties, with communication constraints and conflicting interests, contributing to a process of social learning. This perspective is particularly salient in the context of current debates about p-hacking and publication bias, the replicability crisis of various fields, pre-analysis plans, etc. 2) “Adaptive experimental design for policy choice” – drawing on insights from the literature on bandits and reinforcement learning in AI, reconsidering the methodology of field experiments, especially in policy-relevant settings. 3) “Foundations of machine learning” – developing a deeper understanding of machine learning methods from the perspective of statistical decision theory.

Maximillian Kasy is an Associate Professor in the Department of Economics at Harvard University. He previously served as Assistant Professor in the UCLA Department of Economics and as Consulting Researcher with MSR Alice (Artificial Intelligence Project). He is also a Junior Associate Professor at the Institute of Advanced Studies in Vienna. He was awarded a Mag.rer.soc.oec. in Economics and an Mag.rer.nat. in Mathematics, both from the University of Vienna as well as an MA in Statistics and a PhD in Economics, both from UC Berkeley. He earned his Habilitation at WU Vienna. In 2014 he received the Young Economist Award from the Economic Association of Austria.