Abstract: Jumping behavior is an important feature of short interest rate dynamic process. Jump-diffusion model can better describe the behavior of short rate process. Simulation experiments show that nonparametric threshold estimators perform better in finite sample than alternative model. Empirical analysis on the Shanghai Interbank Offered Rate (Shibor) finds that threshold estimators can effectively detect the jump behavior of short-term Shibor, which consistent with the micro-macro economic and financial phenomena. Comparing with diffusion model, jump - diffusion model based on threshold estimates can more accurately describe the interest rates distribution of skewness and kurtosis, and better reflect the volatility behavior of interest rate.

Key words: short rate, jump-diffusion, threshold estimation