Corporate Bond Pricing via Benchmark Combination Model

This paper develops a benchmark combination model (BCM), which combines univariate-sorted basis portfolios using multiple characteristics, to perform asset evaluation. Under the non-arbitrage restriction, we estimate the common combination weights of BCM to minimize aggregate realized pricing errors. Using a long-span (45-year) sample of U.S. corporate bonds, we find BCM substantially outperforms standard factor models in pricing individual corporate bonds. BCM identifies important risk premia components, including credit rating, maturity, short-term reversal, momentum and variance. There is strong evidence of return predictability for these characteristic-sorted basis portfolios as well as high out-of-sample predictive power for individual bond returns.