COMPUTATIONAL SOCIAL SCIENCE

Doctoral Dissertation Defense: Wayne Zandbergen

Monday, December 2 - 10:00 a.m.
3001 University Hall
Fairfax Campus

Doctoral Dissertation Defense: Essays on the Panic of 1893 - Lessons from Helena, Montana

Wayne Zandbergen, PhD Candidate
Department of Computational Social Science
George Mason University

Abstract: The understanding of emergent financial phenomena such as bank panics, though of interest for many years, has received increased attention since the crisis of 2007-8. Previous crises are often examined, seeking to gain insight into ways in which the frequency or economic damage of such events can be reduced. The Great Depression has garnered a great deal of attention, the Panic of 1893 less so. However, the Panic of 1893 and the Great Depression share several common features: They originated in the interior of the country and spread eastwards, they were double-dip depressions, and they were the two greatest peacetime economic crises in U. S. history [Wicker, 2006]. In contrast to the several banking panics of the Great Depression, the Panic of 1893 has been called “The Perfect Panic” due to the limited institutional intervention between agents engaged in banking [Ramirez, 2009].

Two dominant theories are used to explain bank panics: random factors [Diamond and Dybvig, 1983] and asymmetric information [Greenwald and Stiglitz, 1986], [Akerlof, 1970]. Both theories assume depositors act in an individually rational manner, seeking to optimize outcomes. The difference between the two approaches is less in the cognitive models used for depositors than in the ability of depositors to understand the meaning of events external to the depositor. In this sense, the random factors approach is simply a limit point of asymmetric information, with the limit point opposite random factors representing perfect economic rationality. Empirical analysis has provided at best mixed results in terms of establishing which of the two theories best explains panics [Iyer and Puri, 2008], [Kelly and O’Grada, 2000], and [O’Grada and White, 2003], with recent evidence from the Panic of 1893 suggesting that, depending on the level of analysis, both theories hold [Carlson, 2005]. When empirical studies have focused on individual depositor behavior, they are performed using records of only one bank.

The Montana Historical Society Library in Helena, Montana, maintains daily records for four banks that operated independently during the Panic of 1893. The detailed ledgers and letters document a combined 61 years of operation. This resource provides a unique opportunity to understand features of banking and bank panics not available in any other archival record. Information garnered from this archive provides a common thread of the essays presented.

Historiographically, 1893 has been featured in narratives regarding the social and political climate of the Gilded Age, but detailed critical economic analysis is lacking. The Panic in Montana was severe, and its impact long lasting. The impact on Helena was even more damaging. Bank records show the impact of events in 1893, as well as how each of the four banks was able to respond, or not respond, to changing conditions. Helena banks were exposed to a generalized economic downturn associated with the collapse of silver prices as well as mismanagement and criminality on the part of bank managers. Daily rates of change in overall bank deposit balances and the distribution of bank deposit balances by customer are shown to be fat-tailed. The data also suggests no single common theme or event as the cause of the banking problems that occurred in Helena.

Bank panics have been observed to be localized events, spreading through means of contagion to other regions [Wicker, 1996]. It remains an open question as to whether and to what degree bank-to-bank contagion exists, and whether weaker banks are punished and healthy banks rewarded as a result of such contagion. Data on the four Helena banks provides evidence for such contagion, with depositors in Helena reacting to news of bank panics and failures in other U. S. cities. The evidence also demonstrates that those who withdrew money from one bank did not redeposit into banks deemed more solvent, which is evidence of significant information asymmetries. These findings support policies such as lender-of-last-resort.

Models of panic based upon the two dominant theories can generate panic-like behaviors, but have shown limited success at capturing more detailed empirically observed phenomena [Bryant, 1980], [Romero, 2007], [Gu, 2011]. Such models lack psychosocial characteristics, presenting a limit to their utility [Smith and Shubik, 2012]. The Bank Depositor Model (BDM) considers depositors as embedded in a social complex, with actions driven by their emotional state. Agent emotional state is a function of individual attributes and the positions of neighbors. Derived from models of group dynamics [Bosse et al., 2009], BDM is a reactive agent model that reproduces panic behaviors found in earlier models through social contagion networks, without a model of cognition or of economic rationality. Unlike economically-driven agent models based in random factors or asymmetric information reasoning, it is able to generate distributions observed in the Montana banks.

Taken together, the presented essays use the extensive data available from Helena banking records during the Panic of 1893 to demonstrate that the complexity of banking crises exists on several levels. Investigating empirical sources results in findings that are best viewed as psychosocial phenomena. Such an approach can provide unique insights into theories of financial panic, modeling such events, and policy and regulatory systems focused on limiting the occurrence and mitigating the impact of such events.