Predicting bond betas using macro-finance variables

Abstract

We predict bond betas conditioning on a number of macro-finance variables. We explore differences across long-term government bonds, investment grade corporate bonds, and high yield corporate bonds. We conduct out-of-sample forecasting using the new approach of combining predictor variables through complete subset regressions (CSR). We consider the robustness of CSR forecasts across the 1-month, 3-month, and 12-month forecasting horizons. The CSR method performs well in predicting bond betas.