Dissertations, Theses, and Capstone Projects
Date of Degree
2-2025
Document Type
Dissertation
Degree Name
Ph.D.
Program
Economics
Advisor
Merih Uctum
Committee Members
Seungho Baek
Sangeeta Pratap
Subject Categories
Econometrics | Finance | Macroeconomics
Keywords
Capital Requirements, Cross-Border Lending, Machine Learning, Natural Language Processing, Industry Spillovers
Abstract
Chapter 1: This paper examines the impact of growth in the banking industry’s capital and assets on real output growth at both aggregate and industry levels using a short-run factor-augmented vector autoregression (FAVAR) model across 19 NAICS two-digit industries from 2007Q1 to 2019Q4. Results reveal that changes in capital growth significantly affect output growth in the Manufacturing and Finance industries, while changes in risk-weighted assets impact the Construction, Manufacturing, Real Estate, and Transportation industries. Conversely, output growth in the Manufacturing and Retail Trade industries influences capital growth, while Wholesale Trade output growth affects risk-weighted assets. These relationships are robust to various differencing levels and alternative capital measures. Impulse response analysis highlights distinct dynamics, with seven industries showing meaningful interactions with banking variables. Notably, capital growth positively affects the Finance and Manufacturing industries, while reductions in risk-weighted assets negatively impact Construction and Real Estate but benefit Manufacturing and Transportation. These findings offer a nuanced understanding of the banking sector’s role in economic activity, revealing more complex interactions than previously established in the literature.
Chapter 2: Building upon previous findings, this paper further investigates the interactions between the banking industry and the broader economy using wavelet transformations. The analysis reveals that Granger causality exists between these variables at short- and medium-term time scales but not over longer-term horizons. Additionally, wavelet coherency graphs provide a clear visualization of whether these interactions are consistent over time or driven by specific events within the period.
Chapter 3: As cross-border bank lending plays an increasingly significant role in global economic growth and risk, the ability to accurately predict its future trends has become critically important. This paper addresses this need through two primary approaches: (1) employing machine learning estimation techniques and (2) incorporating newly created variables derived from the textual data of financial sector earnings calls into models alongside established literature-backed features. The findings demonstrate that the random forest technique consistently outperforms others, achieving the lowest errors across all models. Among the text-derived features, the frequency of a country’s mentions in earnings calls proves highly predictive, while sentiment measures are less impactful. These results highlight the potential for creating highly accurate forecasts of cross-border lending and underscore the value of combining textual data with theory-driven predictors for improved predictive accuracy.
Recommended Citation
Hunt, Morgan, "Banking and the Wider Economy: Essays on Capital and Cross-Border Risks" (2025). CUNY Academic Works.
https://academicworks.cuny.edu/gc_etds/6134
Included in
Econometrics Commons, Finance Commons, Macroeconomics Commons