The frequency of banking crises in a dynamic setting: a discrete-time duration approach

Abstract

This paper focuses on descriptive features of banking crises. More than two centuries of banking crises are considered, and a discrete-time duration model is estimated to identify the hazard function characterizing banking crises. The model makes it possible to identify a time-dependence effect in the occurrence of banking crises. The time dependence that emerges from the hazard function is potentially generated by a wide variety of structural and cyclical factors. In this paper, the hazard function serves a descriptive purpose and provides two insights into the frequency of banking crises. First, it shows the extent to which policymakers failed in muting the exposure to a new banking crisis during the two decades following a banking crisis. Second, it provides quantitative evidence that graduation from banking crises is elusive.