LAUSR

This paper investigates the role of time-varying jump tail risk component of market variance risk premium for predicting credit spreads in US and Japanese corporate bond markets. Based on a semi-nonparametric estimation procedure from option prices, we find that the implied left jump variation, which is a simple proxy of the special compensation for jump tail risks, could strongly predict lower-rated credit spreads and default spreads in Japan, even with control for traditional predictors and lagged credit spreads. The predictive pattern on forecasting horizons ranging from 1 month to 1 year differs from that of the diffusive component of variance risk premium, and thus, the variance risk premium decomposition increases the forecasting power of standard predictability regressions. Unlike in Japan, the jump tail risk component might be a weaker predictor for US credit spreads data used in this paper because it becomes insignificant after controlling volatility measures and lagged credit spreads.