Yield Curve Factors, Yield Volatility, and the Predictability of Bond Excess Returns
34 Pages Posted: 8 Mar 2008
Date Written: January 2008
We study the predictability of U.S. government bond excess returns using yield curve factors as well as yield volatility components. The yield curve factors are the level, slope and curvature factors extracted from a dynamic Nelson and Siegel (1987) framework. The yield volatility factors correspond to stochastic volatility components associated with the Nelson-Siegel factors, and capture time-varying risk inherent to the yield curve. The model is estimated using Markov chain Monte Carlo techniques based on U.S. government yields.
We find that the slope and curvature yield factors explain up to 36% of the variation in future yearly bond excess returns revealing the same predictability as the return-forecasting factor proposed by Cochrane and Piazzesi (2005). Moreover, it turns out that the volatility factors parsimoniously and effectively capture yield curve risk. Including them in forecasting regressions for bond return premia increases the forecasting R-Squared to up to 50%. It is shown that the predictive power of the volatility factors is neither subsumed by Nelson-Siegel type yield curve factors nor the Cochrane-Piazzesi return-forecasting factor. Moreover, we illustrate that the extracted yield curve factors and volatility factors are closely connected to underlying macroeconomic fundamentals.
Keywords: term structure, factor volatilities, bond return premia
JEL Classification: C11, G12, G14
Suggested Citation: Suggested Citation