COMPUTATIONAL SOCIAL SCIENCE

Department of Computational Social Science Seminar Abstract

Monday, January 28 - 12:30 p.m.
Research Hall, Room 161

DISSERTATION PROPOSAL: Mechanics and Patterns of Banking Crises – Complexity and the Panic of 1893

Wayne Zandbergen
CSS PhD Student
George Mason University

Abstract: Theoretical explanations of bank crises generally fall into two categories - Random withdrawal, as represented by Diamond & Dybvig (1983), and Asymmetric Information, as described by Stiglitz (1970) and others. Both approaches focus on modeling the cognition of those involved in specific transactions, for example a depositor and a bank. Empirical evidence provides at best mixed evidence for the more macro-level predictions of either approach (Wicker 1996). Historical, or econometric, approaches such as Friedman & Schwartz (1963) utilize macro-level patterns to draw conclusions regarding the causes of banking crises. As with the theoretical approaches, such methods fare poorly when compared with historical data. None of the approaches specifically addresses the fact that bank customers, and banks, exist in complex interactive networks.

This effort takes a naïve, or prehypothetical, approach to understanding the mechanisms and patterns of bank crises. Extensive historical research will provide the foundation for descriptions of customer and bank behavior in the 1890s. From these descriptions two models will be developed. First, a model of bank depositor behavior that extends the concepts of emotional spirals developed by Bosse, et al. (2009) will be developed. This model will be used to recreate phenomena described in empirically-based investigations of bank customer behavior. The second model will be of the National Bank system, circa 1893. This model will provide a foundation for further investigations of the mechanisms of crises as were exhibited in 1893. Both models rely upon network representations as ways to capture the amplification of localized events into more systemic changes. The primary objective is to provide an empirically-grounded description of local interactions that, when modeled over time within a complex interaction network, will provide a more accurate representation of bank crises than is available via current approaches.