The relationship between credit expansion and banking crises is complex and cannot be fully explained by total credit alone.A systematic analysis of the relationship between the amount and structure of total credit an...The relationship between credit expansion and banking crises is complex and cannot be fully explained by total credit alone.A systematic analysis of the relationship between the amount and structure of total credit and banking crises is important for an objective prediction of the influence of potential financial risks.This paper,drawing on data from 15 selected countries,delves into the power of credit indicators in the early warning of banking crises from the perspectives of industrial structure,sector structure,and term structure of credit.Various machine learning methods were used,including Logistic Regression,Random Forest,Decision Tree,Support Vector Machine(SVM),Bagging,and Boosting models.The empirical findings indicate that credit expansion plays a crucial role in triggering banking crises.However,total credit is better suited for the early warning of short-term banking crises,whereas credit structure is more useful for the early warning of long-term banking crises.Moreover,in an early warning system,identifying key early warning indicators is more meaningful than merely increasing the number of indicators.Machine learning can somewhat enhance the early warning power,but it may not always be robust.Therefore,more attention should be paid to potential systemic banking crises resulting from an imbalance in credit structure while regulating the total credit threshold.展开更多
基金funded by the Chongqing Social Sciences Planning Project (2023NDQN22)the Social Sciences and Philosophy Project of the Chongqing Municipal Education Commission (23SKGH097)the Youth Program of Science and Technology Research of Chongqing Municipal Education Commission (KJQN202300545)。
文摘The relationship between credit expansion and banking crises is complex and cannot be fully explained by total credit alone.A systematic analysis of the relationship between the amount and structure of total credit and banking crises is important for an objective prediction of the influence of potential financial risks.This paper,drawing on data from 15 selected countries,delves into the power of credit indicators in the early warning of banking crises from the perspectives of industrial structure,sector structure,and term structure of credit.Various machine learning methods were used,including Logistic Regression,Random Forest,Decision Tree,Support Vector Machine(SVM),Bagging,and Boosting models.The empirical findings indicate that credit expansion plays a crucial role in triggering banking crises.However,total credit is better suited for the early warning of short-term banking crises,whereas credit structure is more useful for the early warning of long-term banking crises.Moreover,in an early warning system,identifying key early warning indicators is more meaningful than merely increasing the number of indicators.Machine learning can somewhat enhance the early warning power,but it may not always be robust.Therefore,more attention should be paid to potential systemic banking crises resulting from an imbalance in credit structure while regulating the total credit threshold.
