Empirical facts characterizing banking crises: an analysis via binary time series

TL;DR

This paper analyzes binary time series data of banking crises across 66 countries from 1800 to 2014, revealing patterns of interconnected crises, clustering by economic and geographical factors, and changes post-financial liberalization.

Contribution

It provides empirical evidence on the interconnectedness and clustering of banking crises over two centuries using novel heatmap and clustering analyses.

Findings

Banking crises tend to occur simultaneously across countries.

Developed European countries show similar crisis paths.

Post-2008 crises are linked by financial ties rather than geography.

Abstract

Various works have already showed that common shocks and cross-country financial linkages caused the banking systems of several countries to be highly interconnected with the result that during bad times, banking crises may arise simultaneously in different countries. Our aim is to provide further evidence on the topic using a dataset made by dichotomous banking crises time series for 66 countries from 1800 to 2014. Via the use of heatmap matrices we show that several countries exhibit pairwise correlation, which means that banking crises tend to occur in the same year. Clustering analysis suggests that developed countries (for the most European ones) are highly similar in terms of the path of events. An analysis of the events that followed the Great Depression and the Great Recession shows that after the crisis of 2008, banking crises tend to characterize countries tied by financial links whereas before 2008 contagion seems to affect countries in the same geographical area. Clustering analysis shows also that after financial liberalization, crises affected countries with similar economic structures and growth. Further researches should enlighten the origin of these linkages investigating how the process of contagion eventually happens.