Banking Crisis Prediction using Objective Crisis Measures and AI Methods
Project Description
In the past decades, a number of financial crises originated in the banking sector. Examples include the Savings and Loan Crisis in the U.S. in the 1980s/90s, the Financial Crisis of 2007/08, or the U.S. Regional
Banking Crisis of 2023. We take the latter as a starting point for our analysis in this project. Based on the existing literature, we discuss the definition of banking crises, their importance, and predictors which have been used successfully in previous research. This provides the basis for the objective, return-based crisis definition used in this paper. Whereas crisis definitions in previous literature frequently suffer from a high degree of subjectivity, our returns-based crisis definition allows for a much more objective categorization of events and replaces a dichotomous distinction (“crisis” vs. “no crisis”) by a membership measure on a continuous spectrum. Given the multivariate, nonlinear and dynamic relations between this membership measure and potential crisis predictors, we employ various methods from artificial intelligence to model these relations and to determine the share of these predictors in the explanatory power of the resulting model.
Banking Crisis of 2023. We take the latter as a starting point for our analysis in this project. Based on the existing literature, we discuss the definition of banking crises, their importance, and predictors which have been used successfully in previous research. This provides the basis for the objective, return-based crisis definition used in this paper. Whereas crisis definitions in previous literature frequently suffer from a high degree of subjectivity, our returns-based crisis definition allows for a much more objective categorization of events and replaces a dichotomous distinction (“crisis” vs. “no crisis”) by a membership measure on a continuous spectrum. Given the multivariate, nonlinear and dynamic relations between this membership measure and potential crisis predictors, we employ various methods from artificial intelligence to model these relations and to determine the share of these predictors in the explanatory power of the resulting model.
Relevance to Liechtenstein
Given the importance of the banking sector for Liechtenstein, insights into regional banking crises are highly relevant for the country and the region.
Scientific, Economic and Societal Impact
Understanding and forecasting banking crises is of considerable practical importance for investors, financial institutions and supervisory authorities.
