Date Thesis Awarded
Bachelors of Arts (BA)
This paper develops an agent-based network simulation model that measures systemic risk in the U.S. banking system. It is shown that ultimate losses to a bank after an initial shock to the system is greater than the direct loss they would expect to face. Using actual balance sheet data and simulating over randomly-generated interbank networks, the model captures the feedback effects that arise from a shock to the highly connected and interdependent system. In addition to capturing these "extra'' losses that arise from bank interactions over several periods, this framework measures the different channels through which the initial risk propagates, amplifies, and transforms. The model is then employed in Monte Carlo simulations for stress tests, which are analyzed from both the perspective of a bank risk manager and a regulator. It is also implemented in a Value-at-Risk framework to demonstrate its potential to inform existing VaR models employed by banks. An important feature of the banking system that is often omitted in related models is collateral underlying the majority of interbank transactions. It is included in this model, and simulations reveal that as much as 30% of total losses due to an asset shock are due to strains in the collateralized interbank debt/repo market. The regulator stress tests highlight the extra risk faced by banks heavily involved in the securities and interbank markets. A wide variety of different scenarios can be tested in this framework, and by collecting detailed information throughout the simulation, the composition of systemic risk and its evolution through the system can be analyzed in different ways.
Pickett, Christopher J., "An Agent-Based Network Simulation Model for Comprehensive Stress Testing and Understanding Systemic Risk" (2014). Undergraduate Honors Theses. Paper 84.
Creative Commons License
This work is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 4.0 License.