国债收益率曲线的动态估计与预测

发布时间：2018-04-17 16:46

本文选题：国债的利率期限结构 + Diebold-Li模型　；参考：《华中科技大学》2013年硕士论文

【摘要】：收益率曲线是在某一固定时点上，不同期限下的收益率与其相应的期限之间的图形关系。它所对应的函数关系便是我们通常所熟知的利率期限结构。国债的收益率往往被视为无风险利率，是一切金融产品收益水平计算和衡量的基准。同时，国债收益率的影响现如今已经扩展到了金融市场之外。短期利率已然成为各国央行货币政策工具箱的香饽饽，它不仅是传递中央银行货币政策的主要媒介，，更是央行调控通胀，干预宏观经济的利器。因而，本文的研究落脚于利率期限结构的动态估计与预测。我们在总结现有动态利率期限结构估计方法的前提下，引入了动态因子模型作为分析框架。在这一框架下，我们将静态估计中的短，中，长期三项的待估计参数分别解释为推动收益率曲线变动的水平因子，斜率因子和曲度因子。我们的实证研究表明在动态因子模型分析框架下的这些因子无论在样本内对收益率的拟合效果，还是在样本外预测的均方根差（RMSE）上都有优异表现。在估计上，我们通过对传统模型的部分参数事先假定，大大简化了因子提取的估计方法。我们的主要结论有Dynamic Svensson模型在样本内对收益率曲线的拟合能力全面优于Diebold-Li模型，但是在样本外预测上，Diebold-Li模型在提前六个月的预测上表现要优于Dynamic Svensson模型。此外，对因子设定的动态过程对于预测结果具有较为明显影响。
[Abstract]:The yield curve is the graphic relationship between the yield and its corresponding term at a fixed time point.The corresponding function is the term structure of interest rate.Treasury yields are often seen as risk-free interest rates and a benchmark for the calculation and measurement of yield levels for all financial products.At the same time, the impact of bond yields has now extended beyond financial markets.Short-term interest rate has become a hot spot in the monetary policy toolbox of central banks. It is not only the main medium for transmitting monetary policy of central banks, but also a sharp weapon for central banks to control inflation and intervene in macro economy.Therefore, the research of this paper is based on the dynamic estimation and prediction of interest rate term structure.On the premise of summing up the existing methods of estimating the term structure of dynamic interest rate, we introduce the dynamic factor model as the analytical framework.In this framework, we interpret the three parameters to be estimated in static estimation as horizontal factor, slope factor and curvature factor, respectively.Our empirical study shows that these factors in the framework of dynamic factor model analysis have excellent performance both in the fitting effect of the return rate in the sample and in the root-mean-square difference (RMSE) predicted outside the sample.In terms of estimation, we presuppose some parameters of the traditional model, which greatly simplifies the estimation method of factor extraction.Our main conclusion is that Dynamic Svensson model is superior to Diebold-Li model in fitting yield curve in sample, but Diebold-Li model is superior to Dynamic Svensson model in predicting six months ahead of sample.In addition, the dynamic process of factor setting has a significant effect on the prediction results.
【学位授予单位】：华中科技大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：F832.51;F224

【参考文献】