Author: Cheuk Yin Jeffrey Mo
Advisor: Dr. Steve Yang
Date: August, 2024
Department: Financial Engineering
Degree: Doctor of Philosophy
Advisory Committee:
Dr. Steve Yang, Chairman
Dr. Zhenyu Cui
Dr. Zachary Feinstein
Dr. Jia Xu
Abstract: The recent financial crisis has prompted researchers to study contagion risks under different lens as the magnitude of shocks propagation cannot be assessed under the existing understanding of contagion risk. During the past decade, traditional macroeconomic models have been developed to explain the occurrence of the financial crisis. However, most of the existing studies relied on the view that the all agents act under the rational expectation and hence ignoring the individual bank performance-driven decision-making process.
In this study, we adopt agent-based modeling framework to model the complexity and dynamics of interbank markets to study interbank risk exposure and contagion risks. Agents are adaptive to the dynamic environment as they learn from past experiences and make future decisions. This dissertation aims to reconstruct the dynamics of an interbank lending market by integrating reinforcement learning control in agent-based modeling. This work incorporates various reinforcement learning algorithms, such as the temporal difference learning and policy gradient methods, to guide bank agents to derive a better lending practice in the interbank lending market. The objective of this research is to overcome the limitations of the existing rational equilibrium based macroeconomic approach to help identify potential latent fragility and shock propagation from a network system perspective.
This dissertation addressed the validity of incorporating reinforcement learning algorithms in the proposed multi-agent interbank lending framework by answering three questions:
1. How can a multi-agent system with learning agents reconstruct the dynamics of an interbank network?
2. How can adaptive learning agents endogenously form liquidity hoarding behavior?
3. How can bank agent derive an optimal lending practice to minimize counterparty risks? This research work studies three aspects of contagion risks in interbank markets:
Research Outcome 1: To investigate how the changes of risk preferences in banks’ decision-making process change the interbank market network structure.
Research Outcome 2: To analyze the impact of fire sale and liquidity hoarding behavior on shock propagation in the interbank market.
Research Outcome 3: To study the impact of information transparency in the interbank market on banks’ ability in assessing counterparty risks.
For full Dissertation, click here.