广义仿射模型在上交所债券市场的实证分析

以上交所债券价格隐含的利率期限结构从1997年1月至2002年3月的月度样本数据作为分析对象，利用卡尔曼滤波法，实证研究了连续时间两因子广义高斯仿射模型．结果表明模型下的利率期限结构与实际观测到的利率期限结构形状基本相同，说明模型能够反映利率期限结构的横截面特征．模型对1，2，3，5年期利率的预测误差没有表现出序列相关性，而对4年期利率的预测误差表现出序列相关性，说明两因子广义高斯仿射模型基本上可以反映利率期限结构的时间序列特征。

Empirical study on the term-structure of interest rates in the SSE with essential affine model

With monthly interest rate data in the Shanghai stock exchange from January 1997 to March 2002, making use of Kalman filter and maximum likelihood estimation approaches, continuous-time two-factor Gaussian essential affine model is estimated. The result indicates that two-factor Gaussian essential affine model fits the shapes of the term structures very well, the predicting errors of the model for 1-year, 2-year, 3-year 5-year rates show no obvious serial correlations, but predicting errors of 4-year rate show some serial correlation. It can be concluded that two-factor Gaussian essential affine model reflects the cross-section and time-series information quite well.