To minimize the gap between academic and practical, and even help financial institutions and competent authorities to be helpful to the ongoing discussions on Basel III and G-SIBs.
This study intends to answer the first question is whether banks’ risk-taking is reduced after the implementation of Basel III capital requirements. It can be subdivided into the following three subquestions: 1. Test the impact of the bank's capital structure required by Basel III on bank risk (profitability). This part has not been systematically discussed. 2. This study will further explore the characteristics of banks classified as G-SIBs or D-SIBs. After becoming a G-SIBs or D-SIBs bank, is it Skin in the Game or Moral Hazard? 3. Consider the impact of the interaction for regulations and capital structure on bank risk (profitability). At current stage, exploring G-SIBs is crucial both internationally and domestically. Yet remind that their selection principles are based on the criteria set by the lack of liquidity of financial institutions that failed during the global financial turmoil. In recent years, due to the overall economy and the risk management of individual banks, the characteristics of systemic risks have changed, so it is necessary to further examine the operating conditions of global financial institutions and the corresponding risks. The second question to be answered in this study: Can we further detect possible financial failures? 1. This study preliminarily conduct a 9- year data study on 15,035 commercial banks around the world through the integration of the Bankfocus, Banksocpe and FDIC databases. The sample period spans over 2013-2021.
2. To predict insolvency banks are generally discussed among the existing literature related to financial distress model. However, it is not an accurate binary situation in practical. In this study, we utilize 7 types of bank status, such as active banks, active receivership, bankruptcy, dissolved, dissolved (merger), inactive (no precision), and in liquidation variables to perform feature engineering. 3. A variety of machine learning methods, such as extreme gradient boost (XGBoost), random forest (RF) and support vector machine (SVM) etc., are applied in order to improve the accuracy of our prediction models.