| Abstract: | This article develops and empirically implements an arbitrage-free,dynamic term structure model with "priced" factor and regime-shiftrisks. The risk factors are assumed to follow a discrete-timeGaussian process, and regime shifts are governed by a discrete-timeMarkov process with state-dependent transition probabilities.This model gives closed-form solutions for zero-coupon bondprices, an analytic representation of the likelihood functionfor bond yields, and a natural decomposition of expected excessreturns to components corresponding to regime-shift and factorrisks. Using monthly data on U.S. Treasury zero-coupon bondyields, we show a critical role of priced, state-dependent regime-shiftrisks in capturing the time variations in expected excess returns,and document notable differences in the behaviors of the factorrisk component of the expected returns across high and low volatilityregimes. Additionally, the state dependence of the regime-switchingprobabilities is shown to capture an interesting asymmetry inthe cyclical behavior of interest rates. The shapes of the termstructure of volatility of bond yield changes are also verydifferent across regimes, with the well-known hump being largelya low-volatility regime phenomenon. |
