Loss given default predictive method analysis under the perspective of Basel IV

Abstract

In response to concerns over capital calculation variability among banks, the Basel Committee revised the Basel III framework in 2017, leading to substantial changes known as Basel IV. This paper explores Loss Given Default (LGD) within the context of Basel IV, focusing on its definition, impact, mathematical measurement, and Moody's LGD model. Additionally, it compares different machine learning models relevant to LGD. Utilizing Moody's Ultimate Recovery Database, which contains detailed recovery information for over 4,600 bonds and loans, this study aims to provide readers with a foundational understanding of LGD under Basel IV and conduct a comparative analysis of machine learning techniques for LGD estimation.