VARIANCE RISK PREMIUM AND BOND RETURN PREDICTABILITY

Abstract

This paper proposes a new way of estimating the risk-neutral volatility of fixed income securities using derivatives data, which can further be used to construct the corresponding yield curve variance risk premium (VRP). After controlling for the shape of yield curves, the VRP strongly predicts one-year holding period excess returns for two-year to ten-year zero coupon bonds. The marginal R^2 of VRP is as high as 40.6%. One standard deviation increase in the VRP is associated with 5.77% increase in the (annualized) bond excess return. This result is robust to controlling for other bond return predictors, including the Cochrane-Piassezi "tent-shaped" factor. Out-of-sample analysis suggests that this predictability is not only statistically signifi cant but can also be translated into economic gains. Additional test suggests that this predictability varies with macroeconomic conditions.