Does Machine Learning Help us Predict Banking Crises?
Johannes Beutel, Sophia List, Gregor von Schweinitz
Journal of Financial Stability,
December
2019
Abstract
This paper compares the out-of-sample predictive performance of different early warning models for systemic banking crises using a sample of advanced economies covering the past 45 years. We compare a benchmark logit approach to several machine learning approaches recently proposed in the literature. We find that while machine learning methods often attain a very high in-sample fit, they are outperformed by the logit approach in recursive out-of-sample evaluations. This result is robust to the choice of performance metric, crisis definition, preference parameter, and sample length, as well as to using different sets of variables and data transformations. Thus, our paper suggests that further enhancements to machine learning early warning models are needed before they are able to offer a substantial value-added for predicting systemic banking crises. Conventional logit models appear to use the available information already fairly efficiently, and would for instance have been able to predict the 2007/2008 financial crisis out-of-sample for many countries. In line with economic intuition, these models identify credit expansions, asset price booms and external imbalances as key predictors of systemic banking crises.
Read article
An Evaluation of Early Warning Models for Systemic Banking Crises: Does Machine Learning Improve Predictions?
Johannes Beutel, Sophia List, Gregor von Schweinitz
Abstract
This paper compares the out-of-sample predictive performance of different early warning models for systemic banking crises using a sample of advanced economies covering the past 45 years. We compare a benchmark logit approach to several machine learning approaches recently proposed in the literature. We find that while machine learning methods often attain a very high in-sample fit, they are outperformed by the logit approach in recursive out-of-sample evaluations. This result is robust to the choice of performance measure, crisis definition, preference parameter, and sample length, as well as to using different sets of variables and data transformations. Thus, our paper suggests that further enhancements to machine learning early warning models are needed before they are able to offer a substantial value-added for predicting systemic banking crises. Conventional logit models appear to use the available information already fairly effciently, and would for instance have been able to predict the 2007/2008 financial crisis out-of-sample for many countries. In line with economic intuition, these models identify credit expansions, asset price booms and external imbalances as key predictors of systemic banking crises.
Read article
Optimizing Policymakers' Loss Functions in Crisis Prediction: Before, Within or After?
Peter Sarlin, Gregor von Schweinitz
Abstract
Early-warning models most commonly optimize signaling thresholds on crisis probabilities. The expost threshold optimization is based upon a loss function accounting for preferences between forecast errors, but comes with two crucial drawbacks: unstable thresholds in recursive estimations and an in-sample overfit at the expense of out-of-sample performance. We propose two alternatives for threshold setting: (i) including preferences in the estimation itself and (ii) setting thresholds ex-ante according to preferences only. Given probabilistic model output, it is intuitive that a decision rule is independent of the data or model specification, as thresholds on probabilities represent a willingness to issue a false alarm vis-à-vis missing a crisis. We provide simulated and real-world evidence that this simplification results in stable thresholds and improves out-of-sample performance. Our solution is not restricted to binary-choice models, but directly transferable to the signaling approach and all probabilistic early-warning models.
Read article
Optimizing Policymakers' Loss Functions in Crisis Prediction: Before, Within or After?
Peter Sarlin, Gregor von Schweinitz
Abstract
Early-warning models most commonly optimize signaling thresholds on crisis probabilities. The ex-post threshold optimization is based upon a loss function accounting for preferences between forecast errors, but comes with two crucial drawbacks: unstable thresholds in recursive estimations and an in-sample overfit at the expense of out-of-sample performance. We propose two alternatives for threshold setting: (i) including preferences in the estimation itself and (ii) setting thresholds ex-ante according to preferences only. We provide simulated and real-world evidence that this simplification results in stable thresholds and improves out-of-sample performance. Our solution is not restricted to binary-choice models, but directly transferable to the signaling approach and all probabilistic early-warning models.
Read article
Predicting Financial Crises: The (Statistical) Significance of the Signals Approach
Makram El-Shagi, Tobias Knedlik, Gregor von Schweinitz
Journal of International Money and Finance,
No. 35,
2013
Abstract
The signals approach as an early-warning system has been fairly successful in detecting crises, but it has so far failed to gain popularity in the scientific community because it cannot distinguish between randomly achieved in-sample fit and true predictive power. To overcome this obstacle, we test the null hypothesis of no correlation between indicators and crisis probability in three applications of the signals approach to different crisis types. To that end, we propose bootstraps specifically tailored to the characteristics of the respective datasets. We find (1) that previous applications of the signals approach yield economically meaningful results; (2) that composite indicators aggregating information contained in individual indicators add value to the signals approach; and (3) that indicators which are found to be significant in-sample usually perform similarly well out-of-sample.
Read article
Predicting Financial Crises: The (Statistical) Significance of the Signals Approach
Makram El-Shagi, Tobias Knedlik, Gregor von Schweinitz
Abstract
The signals approach as an early warning system has been fairly successful in detecting crises, but it has so far failed to gain popularity in the scientific community because it does not distinguish between randomly achieved in-sample fit and true predictive power. To overcome this obstacle, we test the null hypothesis of no correlation between indicators and crisis probability in three applications of the signals approach to different crisis types. To that end, we propose bootstraps specifically tailored to the characteristics of the respective datasets. We find (1) that previous applications of the signals approach yield economically meaningful and statistically significant results and (2) that composite
indicators aggregating information contained in individual indicators add value to the signals approach, even where most individual indicators are not statistically significant on their own.
Read article
