We analytically decompose corporate bond yields into eight risk and yield curve factors and find that five factors among them are important determinants of corporate bond yield spreads and that there exists a non-linear relation between bond yields and betas. Our model explains 99.4% of the crosssectional variations of corporate bond yield spreads compared to 72.9% of Fama-French two factor models in terms of R-squared. We claim that each risk contribution among interest rate, credit and illiquidity may be wrong if the illiquidity factor is simply added to the existing Fama-French two factor models when each risk premium is estimated. Specifically, the illiquidity premium is underestimated by 34% and the credit premium is overestimated by 39% when an inaccurate model is used. We also show that the relationship between credit and illiquidity varies depending on the economic situation. Specifically, the liquidity premium showed a positive correlation with the credit premium in global financial crisis, whereas it showed a negative one in European national debt crisis. We find that liquidity black holes arise at the beginning of a financial crisis and financial markets quickly become unstable. Our method disentangles a flight-to-quality from a flight-to-liquidity and identifies the risk contributions of credit and illiquidity and among the yield curve factors (level, steepness, and concavity). Our model also explains a flight-from-maturity phenomenon.
Keywords: Analytic factor decomposition method, Credit premium, Illiquidity premium, Yield curve factors, Liquidity black holes
JEL Classification: G12, G14
