(A) study on nonparametric modeling of Korean interest rate term structure dynamics

Abstract

Stanton (1997) and Jiang (1998) use nonparametric models applied to short-term interest rate data to estimate interest rate term structure dynamics. Their methodology allows for maximum flexibility in fitting into the data. We implement their methodology using Korean data, and compare the estimation results to those by Chan, Karolyi, Longstaff, and Sanders (1992) using the same data. In our results, the nonparametric drift function looks nonlinear and shows weaker mean reversion compared to its parametric counterpart. The diffusion functions using the two models look nonlinear and are sensitive to the level of the interest rate. The nonparametric market price of interest rate risk multiplied by -1 is distinctly different from zero, and increases as the interest rate gets higher. This fact explains that a higher risk premium reflects higher volatility. With these term structure dynamics, we obtain the bond prices and compare these with the actual prices. In our results, the nonparametric model does not show better performance in terms of the mean squared error.