Timing the corporate bond market

Abstract

This paper develops a transparent, data-driven strategy for timing the corporate bond market using real-time macro-financial predictors. We implement a forecast combination approach—conceptually aligned with ensemble learning in machine learning—to dynamically adjust exposure to the bond market based on expected excess returns. The resulting managed portfolio significantly outperforms traditional static strategies and leading factor models, both in- and out-of-sample, delivering higher Sharpe ratios, investor utility, and robustness to transaction costs and leverage constraints. Our strategy proves particularly effective during periods of macroeconomic turbulence, capturing countercyclical return patterns and enhancing portfolio performance when economic conditions shift rapidly. By sidestepping black-box algorithms, our work contributes to the literature on AI-assisted investing by demonstrating how interpretable, real-time prediction models can improve portfolio allocation in fixed-income markets.