Dissertation Defense: “Essays on Macroeconomics and Machine Learning”, Yuanzhe Liu
Speaker
Yuanzhe Liu, University of California, Santa Barbara
Title
“Essays on Macroeconomics and Machine Learning”
Abstract
This dissertation comprises three chapters.
In the first chapter, I investigate the surge in durable goods consumption during 2020–2021 through the lens of a household production model with working-from-home. Historically, recessions have been associated with declines or slowdowns in durable goods consumption. However, during and after the COVID-19 pandemic, such consumption rose sharply. I develop a household production model incorporating working from home and estimate it using a Bayesian approach. Applying a Kalman smoother, I decompose the increase in durable goods consumption into its contributing channels. The results show that working from home accounts for up to one-third of the increase, while substitution from nondurable to durable goods explains another one-third.
The second chapter introduces an innovative approach that leverages Deep Reinforcement Learning (DRL) to solve and estimate heterogeneous agent models. Unlike conventional solution methods, the DRL-based approach delivers global solutions while preserving the full nonlinearity of the model. It also scales efficiently, making it applicable to models with hundreds or even thousands of state variables. I further explore combining this solution method with amortized likelihood-free Bayesian inference, opening new avenues for advanced probabilistic modeling and estimation.
The third chapter, coauthored with Dr. Fanyu Liu, examines how expanding access to health insurance for low-income households affects real estate investment decisions. Specifically, I estimate the impact of Affordable Care Act (ACA) marketplace subsidies on mortgage applications and originations, using comprehensive mortgage application records. I first develop a theoretical model in which simulations indicate that reducing medical expenses and uncertainty through insurance increases the demand for housing. Empirically, I employ a difference-in-differences strategy that exploits county-level variation in health insurance coverage rates at the time of the ACA reform. The results show that the subsidies increased both the volume of residential mortgage applications submitted by low-income households and the number of mortgages originated by financial institutions. Mechanism analysis reveals that the subsidies raised insurance coverage rates among low-income households, reduced large medical expenditures, and lowered mortgage delinquency rates. These findings highlight the broader economic effects of health insurance beyond alleviating healthcare costs for low-income groups.
JEL Codes: C63, E21, C11
Event Details
Join us to hear Yuanzhe’s dissertation defense. He will be presenting his dissertation titled, “Essays on Macroeconomics and Machine Learning”. To access a copy of the dissertation, you must have an active UCSB NetID and password.
