On the time-varying relation between monetary policy uncertainty and bond risk premia

Abstract

This paper examines the time-varying relationship between monetary policy uncertainty (MPU) and bond excess returns. To do so, we introduce a nonparametric time-varying coefficient predictive regression model for bond returns, and employ a kernel-based two-step method to estimate the time-varying coefficients. Next, we apply the methodologies to analyze the dynamic forecasting relationship between zero-coupon bond returns and MPU from 1985 to 2022. We find that MPU significantly and positively predicts bond returns in over 75% of the sample period, with the strongest effect observed in 2005. Thus, the expectations hypothesis is only transiently valid. After controlling for the shape of the yield curve, MPU still retains its ability to predict bond returns in 50% to 80% of the sample period. Our conclusions are robust to the so-called embedded endogeneity. Additionally, we find that bond excess returns are less responsive to MPU during periods of high economic activities and are more responsive during periods of low economic activities.