Understanding the determinants of bond excess returns using explainable AI

Abstract

Abstract Recent empirical evidence indicates that bond excess returns can be predicted using machine learning models. However, although the predictive power of machine learning models is intriguing, they typically lack transparency. This paper introduces the state-of-the-art explainable artificial intelligence technique SHapley Additive exPlanations (SHAP) to open the black box of these models. Our analysis identifies the key determinants that drive the predictions of bond excess returns produced by machine learning models and recognizes how these determinants relate to bond excess returns. This approach facilitates an economic interpretation of the predictions of bond excess returns made by machine learning models and contributes to a thorough understanding of the determinants of bond excess returns, which is critical for the decisions of market participants and the evaluation of economic theories.