Intertemporal Defaulted Bond Recoveries Prediction Via Machine Learning

The recovery rate on defaulted corporate bonds has a time-varying distribution. We propose machine learning approaches for intertemporal analysis of U.S. corporate bonds' recovery rates with a large number of predictors. The most informative macroeconomic variables are selected from a broad range of macroeconomic variables using econometrics and machine learning methods. In addition to considering a large set of macroeconomic variables, bond characteristics, and industry characteristics, we study the relation between the recovery rates and news-based measures. We find that machine learning techniques significantly outperform traditional approaches not only out-of-sample as documented in the literature but also out-of-time. Examining the permutation importance of each group of explanatory variables, we find that bond characteristics, seniority dummy variables, stock market indicators, international competitiveness, and news are the most informative groups of variables for recovery rate prediction