Time Dependence in Banking Crises: A Discrete-Time Duration Approach

Abstract

This paper investigates whether the probability of a banking crisis depends on the time that has elapsed since the last banking crisis. More than two centuries of banking crises are considered, and a discrete-time duration model that is based on a cubic spline methodology is implemented. The results allow to identify when the exposure to a banking crisis is particularly high, which is a major concern for policy makers. The probability of having a new banking crisis noticeably increases over the first decade following a banking crisis. Furthermore, (as expected) the exposure to banking crises is also particularly high following a long period of time without any banking crisis because no country has ever avoided having a banking crisis in the long run. However, between the two extremes, the exposure to a banking crisis decreases, which suggests that countries benefit from a period of increasing stability.

Keywords: banking crises, discrete-time duration model

JEL Classification: G01, G21, C41