Banking Crises and Contagion: Empirical Evidence

Recent events, such as the East Asian, Mexican, Scandinavian, and Argentinian crises, have sparked considerable interest in exploring how shocks experienced by one country can spread vis-à-vis real and nominal links to other countries' banking systems. Given the large costs associated with banking-system failures, both economists and policy-makers are interested in predicting the onset of banking crises and assessing the likelihood of contagion during crisis events. The author uses cross-country panel data to examine contagion across banking systems in developed and developing countries. Particular attention is paid to the construction of the cross-country sample: matching-method techniques are used to construct a suitable control-group sample analogue to the set of crisis countries to accurately quantify the probability of the occurrence of a banking crisis and the probability of banking-system contagion. The author finds that the sample choices of previous studies introduced bias into the estimates of the probability that a banking crisis would occur, owing to differences between the supports of the conditioning variables for the crisis and non-crisis country groups. Furthermore, the probability of a banking crisis increases when countries have macroeconomic characteristics similar to those that have recently experienced a crisis, regardless of the degree of actual economic linkages between the countries. This suggests that information contagion plays a larger role than previously suspected.