Thesis title: Essays in Causal Inference Methods Designed for Financial Economics
This thesis develops robust causal inference methods to address empirical challenges in finance, moving beyond prediction to uncover underlying economic mechanisms in financial economics. While traditional empirical finance focuses on improving prediction accuracy using machine learning techniques, these methods often fall short in identifying causal relationships among financial variables. To bridge this gap, I introduce a novel causal inference approach in international finance, detailed in Chapter 1, which builds on the Generalized Synthetic Control method by instrumenting factor loadings with predictive covariates to effectively handle high-dimensional data while minimizing model misspecification and unobserved covariate bias. Chapter 2 explores the predictive power of neural networks for stock returns, investigating the interaction effects of firm-specific and macroeconomic variables and motivating novel econometric methods to impute missing counterfactuals for causal inference. Chapter 3 examines the relationship between firms' carbon emissions and stock returns, documenting a correlation but leaving causality unresolved, further motivating the need for refined causal inference tools in financial economics. Together, these chapters contribute to a more comprehensive understanding of financial outcomes, with reproducible code available to support future research.