Monetary Policy, Banking Crises, and the Friedman Rule

Abstract

Banking crises are frequent events. Gerard Caprio and Daniela Klingebiel (1997) catalog over 80 banking crises during the last 25 years. Interestingly, some banking crises are associated with no output losses whatsoever, while others involve massive recessions. Finally, it is known that the probability of a banking crisis rises as the rate of inflation rises (see Asli Demirguc-Kunt and Enrica Detragiache, 1997; John Boyd et al., 2001a, b). While received wisdom exists about the conduct of monetary policy while a crisis is underway, there is no formal treatment of how the conduct of monetary policy during “normal times” affects the potential for banking crises to occur. To fill that gap, I consider economies where spatial separation and limited communication create a transactions role for money, and random shocks to agents’ liquidity preferences create a role for banks. In addition, banks confront randomness in withdrawal demand. When withdrawal demand is sufficiently high, banks exhaust their cash reserves. There are good reasons to associate this with a banking panic. The output lost during a banking panic depends on how great withdrawal demand is. Some banking crises generate no output losses, while others generate large reductions in resource availability. In addition, the conduct of monetary policy during “normal times” affects the probability of a banking crisis. The higher the nominal interest rate (the inflation rate), the higher is the probability of a panic. Driving the nominal interest rate to zero (following the Friedman rule) eliminates bank panics. This constitutes a new rationale for the Friedman rule. Nonetheless, conventional methods of implementing the Friedman rule never produce an optimal resource allocation. In particular, low nominal interest rates induce banks to hold large cash reserves, thereby forgoing socially more productive investments. In effect, the Friedman rule induces banks to become narrow banks voluntarily. The banking system is very safe, but it undertakes a suboptimal level of investment. Less conventional methods for implementing the Friedman rule, such as allowing unrestricted access to the discount window at a zero nominal interest rate, lead either to the nonexistence of equilibrium, or to massive indeterminacies. None of the equilibria will be consistent with full optimality.