Risk premia and the dynamic covariance between stock and bond returns

We investigate whether intertemporal variation in stock and bond risk premia can be explained by time-varying covariances with priced risk factors. We estimate and test a conditional two-factor variant of Merton's ICAPM in which excess returns on an equity index and a long-term government bond portfolio proxy for risk factors. Conditional second moments follow the asymmetric dynamic covariance (ADC) model of Kroner and Ng (1998). We find that conditional bond variance responds symmetrically to bond return shocks but is virtually unaffected by stock return shocks, while conditional stock variance responds asymmetrically to both stock and bond return shocks. Models that impose a constant correlation restriction on the covariance matrix between stock and bond returns are strongly rejected. We conclude that the conditional two-factor model fails to adequately explain intertemporal variation in stock and bond risk premia.