ESSAYS IN COMPUTATIONAL AND CLIMATE ECONOMICS

Loading...
Thumbnail Image
Degree type
Doctor of Philosophy (PhD)
Graduate group
Economics
Discipline
Economics
Subject
Funder
Grant number
License
Copyright date
2024
Distributor
Related resources
Author
Huang, Joseph, Salera
Contributor
Abstract

Chapter 1 examines computational methods for solving continuous-time macroeconomic models, comparing grid-based methods enhanced by high-performance computing (HPC) and grid-free ap- proaches using deep learning. While HPC improves grid-based methods’ performance, I show how Amdahl’s Law constrains their scalability. Deep learning methods, particularly the Deep Galerkin Method, offer an alternative that better handles high-dimensional problems by avoiding the curse of dimensionality. In the subsequent chapters, my co-authors and I provide implementation frame- works for both approaches in solving Hamilton-Jacobi-Bellman equations and demonstrate their applications in macroeconomic and climate economics models. In chapter 2, using firm-level data, I document five empirical facts that reveal a misalignment between high ESG scores and actual environmental performance. To explain these facts, I develop a novel general equilibrium model with distorted learning. The model quantifies the economic and environmental costs of greenwashing. Key findings include: (1) firms can take up to three times as long to reduce emissions compared to scenarios with perfect information; (2) even after a decade, economies with significant greenwashing exhibit emissions that are 0.01 gigatons (0.1% of total US emissions) higher and are 9% less green than in scenarios with perfect information; and (3) greenwashing can increase social cost of carbon by up to $18 billion due to delayed emission reductions. In chapter 3, my co-authors (Lars Hansen, Michael Barnett, William Brock, and Ruimeng Hu) and I develop a neural-network-based computational algorithm that incorporates pseudo-states variables to derive accurate global solutions to high-dimensional, non-linear, dynamic models with jumps and uncertainty aversion. We use this algorithm to study optimal carbon-neutral transition policy in response to climate change, a setting where endogenous nonlinearities from model uncertainty and jump processes are first-order considerations. Our social planner places substantial value on investment in R&D and green capital, though the significant “social pay-off” of green technological innovation means only R&D investment is significantly augmented by uncertainty concerns, even with uncertainty concerns about jumps in the severity of climate change damages.

Advisor
Fernández-Villaverde, Jesús
Date of degree
2024
Date Range for Data Collection (Start Date)
Date Range for Data Collection (End Date)
Digital Object Identifier
Series name and number
Volume number
Issue number
Publisher
Publisher DOI
Journal Issue
Comments
Recommended citation