中国国债收益率曲线的拟合及预测研究  

Research on Fitting and Prediction of China′s Treasury Bond Yield Curves

作　　者：牛晓健[1] 邓昕妤 Niu Xiao-jian;Deng Xin-yu(School of Economics,Fudan University,Shanghai 200433)

机构地区：[1]复旦大学经济学院,上海200433

出　　处：《贵州商学院学报》2020年第4期9-18,共10页Journal of Guizhou University Of Commerce

基　　金：国家自然科学基金面上项目“流动性压力、信息交互与价格联动—基于中国股票和债券市场多层复杂网络的风险交叉传播机制与控制修复策略研究”(71873039,71573051);上海市“曙光计划”项目“资本流动对中国货币政策效应的实证研究”(11SG09)。

摘　　要：随着我国利率市场化的推进,国债的价格形成及其收益率曲线的完善在利率定价中发挥着越来越关键的作用,国债收益率曲线的预测对于金融机构利率风险管理至关重要。本文以wind中的中债国债即期收益率数据为样本,利用动态Nelson-Siegel模型拟合出样本内的收益率曲线,根据拟合的收益率曲线做出预测。拟合分为两步:第一步,先后使用了非线性最小二乘法和卡尔曼滤波法,估计出模型中的非线性参数衰减因子,从而使结果更稳定;第二步,使用线性最小二乘法,估计出模型中的线性参数序列为水平因子、斜率因子和曲度因子。预测分为两步:第一步,利用NARX动态神经网络方法对水平因子、斜率因子和曲度因子做预测;第二步,根据预测的下期的三个因子值和拟合过程中确定的衰减因子,得到下期的预测的收益率随期限变化的函数。并据此绘制该预测的函数图像。结果表明该方法较好地预测了国债收益率曲线的变动规律,具有一定的应用价值,可以据此制定合适的国债交易策略,也可为金融机构进行利率风险管理提供一定的决策参考。With the advancement of interest rate marketization,the formation of national debt prices and the improvement of its yield curves play an increasingly critical role in interest rate pricing.The prediction of the national debt yield curve is crucial for financial institutions′interest rate risk management.This paper takes the data of China Bond National Debt Spot Yield in wind as a sample,uses the dynamic Nelson-Siegel model to fit the yield curve in the sample,and makes the prediction based on the fitted yield curve.The fitting is divided into two steps:the first step is to use the nonlinear least squares method and the Kalman filter method to estimate the nonlinear parameter attenuation factor in the model,so as to make the result more stable.In the second step,this paper uses linear least squares multiplication method and the linear parameter sequence in the model is estimated as horizontal factor,slope factor and curvature factor.The prediction is divided into two steps:the first step is to use the NARX dynamic neural network method to predict the horizontal factor,the slope factor and the curvature factor;the second step is to predict the next three factor values and the attenuation determined during the fitting process,then get the function of the predicted return rate of the next period with the change of the period.Based on previous step,the predicted function image is drawn.The results show that the method can predict the changing law of the yield curve of government bonds well,and has certain application value.It can formulate appropriate government bond trading strategies based on this method,and it can also provide certain decision-making reference for financial institutions to manage interest rate risk.

关 键 词：国债收益率曲线 卡尔曼滤波法 NARX动态神经网络 利率预测 

分 类 号：F830[经济管理—金融学]