Banking Panics, Information, and Rational Expectations Equilibrium

This paper shows that bank runs can be modeled as an equilibrium phenomenon. We demonstrate that some aspects of the intuitive "story" that bank runs start with fears of insolvency of banks can be rigorously modeled. If individuals observe long "lines" at the bank, they correctly infer that there is a possibility that the bank is about to fail and precipitate a bank run. However, bank runs occur even when no one has any adverse information. Extra market constraints such as suspension of convertibility can prevent bank runs and result in superior allocations. BANKING PANICS WERE A recurrent phenomenon in the United States until the 1930s. They have re-emerged as a source of public concern and much theoretical research recently. In this paper, we provide an information-theoretic rationale for bank runs. The traditional "story" is that contagion is an important aspect of bank runs. The idea seems to be that when the general public observes large withdrawals from the banking system, fears of insolvency grow resulting in even larger withdrawals of deposits. In our model, some individuals withdraw because they get information that future returns are likely to be low. Uninformed individuals observing this also have an incentive to liquidate their investments. In addition, some individuals need to withdraw deposits for other than informationally based reasons. Thus, if the random realization of such a group of individuals is unusually large, then the uninformed individuals will be misled and will precipitate a run on the bank. The technology is such that a large volume of withdrawals involves liquidation costs. Consequently, runs on the bank do impose social costs. A mechanism that may reduce these costs in our model is to suspend convertibility if withdrawals are high. However, those individuals who need to withdraw their assets for liquidity reasons are worse off ex post. Our model provides a rationalization for restrictions in demand deposits that were widespread prior to 1929. Friedman and Schwartz [7] suggest that restrictions of payments ensured