摘要
以上交所债券价格隐含的利率期限结构从1996-03~2003-02的周样本数据作为分析对象,实证研究了常见的利率模型:Vasicek模型、CIR模型、仿射模型、广义高斯仿射模型,利用卡尔曼滤波法,估计了连续时间两因子利率模型。从模型给出的利率与实际观测到的利率的平均绝对误差来看,这些模型可以较好地描写利率期限结构的横截面特征,广义高斯仿射模型最好,仿射模型次之,CIR模型与Vasicek模型难分高下。但模型对各年期利率的预测误差表现出一定的序列相关性,说明这些模型不能够很好地描述利率期限结构的时间序列特征。
With weekly yield curve data in the Shanghai Stock Exchange from June 1996 to February 2003, making use of Kalman filter and maximum likelihood estimation approaches, continuous-time two-factor interest rate models are estimated. They are Vasicek model, CIR model, affine model, and Gaussian essential affine model. According to mean absolute spread of model implied yield curves and observed yield curves, the models can describe relative changes of yield curves very well. Two-factor Gaussian essential affine model is the best among them. Affine model is better than CIR model and Vasicek model. But no model fully describes the time-serial information of the yield curves, because the predicting errors of the models for yield curves show obvious serial correlations.
出处
《系统工程理论方法应用》
2004年第4期355-360,371,共7页
Systems Engineering Theory·Methodology·Applications
基金
教育部基金资助项目(01JC630008)
关键词
仿射模型
横截面
时间序列
上交所
卡尔曼滤波
affine model
cross-section
time series
the shanghai stock exchange
the Kalman filter