Who Panics During Panics?

Evidence from a Nineteenth Century Savings Bank

Cormac Ó Gráda, University of Dublin and Eugene N. White, Rutgers University


Contagion is greatly feared in today’s world financial system. The possibility that the collapse of one country’s equity market or banking system can spill over to other countries is a grave concern of policy makers. Contagion can occur in all types of financial markets. In this paper, we examine the most traditional form of contagion, runs on a single bank in the uninsured nineteenth century American banking system.

There are generally thought to be two types of banking contagion (Aharony and Swary, 1983; Lang and Stulz, 1992; Park, 1993, Kauffman, 1994). Industry specific contagion occurs when information about one bank adversely affects all banks including healthy ones that share few characteristics with troubled banks. Runs of this type are referred to as noninformational contagion. This type of contagion has some similarities with the mechanism that brings about a run on individual banks in models by Bryant (1980) and Diamond and Dybvig (1983). In these models, runs occur when each depositor believes that others will run on the bank and force it into liquidation. Fear of being last ignites a run, which are randomly occurring events a rational equilibrium.

The second type of contagion, bank specific contagion occurs when new information about one or more banks produces runs on other banks in the system that have some common characteristics with the former but not all banks. This type of run is considered the result of informational contagion. Contagion of this kind may be generated by asymmetric information among depositors. Jacklin (1983), Smith (1984), Gorton (1987), Chari and Jagannathan (1988), Jacklin and Battacharya (1988), Williamson (1988) and Calomiris and Schweikart (1991) have looked at the informational problems in the bank-depositor relationship. A run occurs because enough agents receive negative signals about the value of a bank’s assets. Other depositors may respond with a withdrawal of deposits because they cannot easily to observe the value of bank assets or the motives of other withdrawing depositors, producing a run.

These two types of contagion can also be characterized as the product of banking runs, by either poorly-informed or well-informed depositors (Frankel and Schukler, 1996). Bank specific runs should begin with the well-informed making use of their information while industry specific contagion should be precipitated and spread by a run of the poorly informed. Unfortunately, there is very little empirical work about what type of informational asymmetries caused banking runs and panics. Most of the contemporary empirical literature has focused on contagion in equity returns for banks not contagion among depositors.

In this paper, we provide some historical evidence on the nature of banking runs, drawing on the records of the Emigrant Savings Industrial Bank. The ESIB was subject to three serious runs in its early years. The first run occurred as the result of a local panic in 1854 when the Knickerbocker Savings Bank failed. The second run was part of the panic and nationwide crisis in 1857, and the third run took place in 1861 at the outbreak of the Civil War. The three episodes provide a natural experiment that allows us to observe how different types of information or shocks affected depositors. The Knickerbocker Savings Bank failed because of mismanagement, while other savings banks and banks were solvent and ultimately did not fail. This episode may be considered a potential example of bank specific contagion panic. Looking at the EISB, an apparently strong bank, will provide an opportunity to see if there is any evidence of noninformational contagion. Although the panic of 1857 was precipitated by the failure of the Ohio Life and Trust Company, the proximate cause of the panic was the collapse of the market for speculative western land and railroad securities, which was linked to the political uncertainty over whether Kansas and Nebraska would become slave states (Calomiris and Schweikart, 1991). This crisis affected banks differently, depending on their investment in these markets, and all banks given the general shock to the financial system. A run in 1857 should be a mix of informational and noninformational contagion. Finally, the run in 1861 was the result of a general economic shock to the system, where one would expect the informed depositors to act first. Given the stable character of the EISB deposit base, these three episodes allow us to determine how different shocks generated runs among the bank's depositors.

To examine who panicked we analyze who closed accounts during panics and who kept them open. The EISB's test books contain the names, addresses, and occupations of account holders. Usually, they also provide data on nationality, and date of arrival in New York. The account ledgers detail all transactions and interest payments for each account. Together these yield a profile of each account holder. We use a probit model to examine the determinants of account closure. Our study of 1854 and 1857, compares the 234 and 503 accounts closed during the panics with a sample of accounts that remained open. This sample is a one-in-ten sample of all accounts opened from the date of the creation of the bank in 1850. We find that account holders were less likely to close their accounts, the longer he/she was resident in the U.S., suggesting a greater familiarity with the savings' bank. Unskilled account holders and those with small opening balances had a greater probability of closing an account. The variables measuring banking activity --the greater the frequency of deposits and withdrawals, and the size of the account prior to the panic -- suggest that familiarity with banking reduced the likelihood of panic. This evidence indicates that it was the less informed depositors who panicked in 1854 and 1857.