396 – Financial Time Series Modeling
Multi-Factor Discretely Observed Vasicek Term Structure Models with Non-Gaussian Innovations and Its Applications to the Japanese Government Bond Markets
Sponsor: Business and Economic Statistics Section
Keywords: term structure models, bond yields, asymptotic expansion, Vasicek models, Kalman filter
Takayuki Shiohama
Tokyo University of Science
In this paper, we propose a multi-factor model in which the discretely observed short-term interest rates follow a non-Gaussian and dependent process. The state space formulation has the advantages of taking into account both the cross-sectional and time-series restrictions on the data and measurement errors in the observed yield curve. Clarifying the non-Gaussianity and dependency of the dynamics of short-term interest rates, we show that these features are important to capture the dynamics of the observed yield curve. Applications to the estimation of the Japanese government bond yield are illustrated.