基于动态Nelson Siegel模型的银行间国债市场收益率曲线研究

发布时间：2018-04-12 16:18

本文选题：银行间国债市场 + 动态Nelson-Siegel期限结构　；参考：《西南财经大学》2013年硕士论文

【摘要】：在利率市场化的大环境下,金融机构对于利率风险的控制和对资产配置的需求使得关于利率期限结构的研究变得越来越有意义。利率期限结构不仅能为投资者提供组合管理的建议,还能为监管机构提供宏观调控的思路。在实际应用中,对于利率期限结构的拟合效果和预测能力的研究更是利率类模型的核心。现代利率期限结构研究主要包含静态模型、无套利模型、均衡模型、宏观金融模型、混合模型。静态模型对某个时点上的利率期限结构进行估计,主要的相关研究包括McCulloch (1971)、Vasicek等(1982)的样条法和Fama等(1987)的息票剥离法；无套利模型和均衡模型从利率的随机过程出发,主要的模型包括Vasicek模型,CIR模型和HJM模型；宏观金融模型将宏观变量与利率期限结构相结合,主要的研究包括Wu(2003)实证了宏观经济冲击收益率曲线的影响、Ang等(2003)将宏观经济变量引入均衡模型；混合模型对以上模型取长补短,主要的研究包括Diebold等(2002)(2006)建立的动态Nelson-Siegel模型。国内对于利率期限结构主要研究包括朱世武等(2003)和余文龙等(2010)通过Nelson-Siegel模型对交易所利率期限结构的研究。静态模型和动态模型的存在的共同问题是,尽管对利率期限结构拟合效果较好,但相对于宏观金融模型和混合模型,其预测能力较差；国内学者尝试采用宏观金融模型结合国内经济变量对利率期限结构进行解释和预测,但这些研究主要采用交易所或银行间债券收盘价格拟合利率期限结构,没有针对银行间报价的研究。本文主要采用银行间双边报价数据,基于动态Nelson-Siegel模型,计算出对利率期限结构具有代表作用的三个因子,并通过误差修正模型(VECM)与宏观经济变量相结合,构建出适合我国利率期限的混合模型。在对模型各变量提出有意义的经济解释的同时,发现其样本外预测能力高于普通动态Nelson-Siegel模型。全文共6个部分,各部分内容安排如下。第一部分是引言,介绍论文研究的背景和意义、研究方法和研究工具、实证主要内容及详细步骤,最后指明了文章的创新之处。第二部分是对国内外利率期限结构研究进行文献综述。从传统的利率期限结构理论出发,到现代利率期限结构模型包括期限结构静态估计方法、动态模型、混合模型和宏观金融模型等。对国内利率期限结构也从静态模型,动态模型,Nelson-Siegel模型、银行间利率期限结构等不同角度进行了综述。第三部分是介绍模型与检验的原理和实现方法。首先介绍了Nelson-Siegel模型的模型设定以及水平因子,斜率因子,曲率因子的数学含义。然后给出了状态空间模型的概念以及卡尔曼滤波器的概念、滤波的过程以及如何对模型中的参数进行估计。再介绍了本文所采用的的数值计算方法。最后介绍了文中使用的一系列时间序列模型,包括向量自回归模型,向量误差修正模型,Johansen协整检验的基本思想。第四部分是数据来源及处理。首先对选择银行间债券市场双边报价数据作为本文研究的对象的原因进行说明。然后对该数据进行预处理,并且通过Fama-Bliss方法将银行间国债双边价格数据转化为零息票债券收益率日数据。第五部分是实证研究,主要工作包括对模型参数的估计以及预测能力的研究。第一步是在前序工作的基础上对NS模型三因子以及规模参数λ进行估计。第二步对三因子是建立向量自回归模型并检验预测能力。第四步将三因子与宏观经济变量结合起来建立误差修正模型考察因子与宏观经济数据的关系并改进了模型的预测能力。最后,对研究得到的结论进行总结,给出建议,提出不足之处并对进一步研究的方向进行展望。本文通过研究发现,第一,水平、斜率、曲率三因子与居民消费价格指数(CPI)和工业增加值(IP)存在一个长期稳定的协整关系。第二,发现加入宏观变量后的误差修正模型(VECM)在短、中期限的预测能力得到了明显改善,强于无约束的向量自回归模型(UVAR)模型。本文的创新之处在于以下三个方面。首先,本文利用Nelson-Siegel系列模型估计出来的三个因子与宏观经济变量通过误差修正模型建立起稳定的模型并提高了普通向量自回归模型的预测能力。其次,本文不同于一般的文章选择交易所债券数据对利率期限结构进行研究,转而选择了银行间市场双边报价数据进行研究。最后,文章采用数值方法结合最小二乘法估计的三因子的平均值作为卡尔曼滤波器三因子的初值,提高了估计的准确性。本文未来发展方向有以下两个方面。第一,从本文中构建的模型出发,沿着文中的建模思路继续寻找合适的宏观经济因子提高模型的预测能力。第二,参照Koopman(2010)中的方法,首先将规模参数λ考虑为第四个可变因子进入到状态空间模型之中,通过扩展的卡尔曼滤波(Extended Kalman)对模型参数进行估计,然后采用本文中的方法,将规模参数也引入整体模型设计之中,提高模型的拟合效果和预测能力。
[Abstract]:In the interest rate market environment, financial institutions have become more and more important for the control of interest rate risk and asset allocation on demand makes the research on the term structure of interest rates. The term structure of interest rates can not only provide portfolio management recommendations for investors, but also provide the macro-control ideas for supervision. In practical application. Study on interest rate term structure fitting effect and prediction ability is the core interest rate model.
Study on the modern term structure mainly includes the static model, no arbitrage model, the equilibrium model, the macro finance model, hybrid model. The static model to estimate the term structure of interest rates a point in time, related research mainly include McCulloch (1971), Vasicek (1982) of the spline method and Fama (1987). Bootstrap method; no arbitrage model and equilibrium model starting from the stochastic process of interest rate, the main models include Vasicek model, CIR model and HJM model; the combination of macro financial model of macroeconomic variables and the term structure of interest rate, the main research include Wu (2003) positive impact of macroeconomic shocks, the yield curve, Ang etc. (2003) the macroeconomic variables into the equilibrium model; hybrid model of the above models complement each other, the main research including Diebold (2002) Nelson-Siegel (2006) dynamic model. For the domestic interest rate period The main studies of the limited structure include the study of the term structure of the exchange rate through the Nelson-Siegel model (2003) and Yu Wenlong (2010), including (2003) and Yu Wenlong.
There is a common problem of static model and dynamic model is better, although on the interest rate term structure fitting effect, but compared to the macro financial model and mixed model, the prediction ability is poor; domestic scholars try to use the macro financial model combined with the domestic economic variables of interest rate term structure to explain and predict, but these studies mainly by the exchange or bank the bond between the closing price fit the term structure of interest rates, not on the inter-bank offer.
This paper mainly uses bilateral quotation data between banks, based on dynamic Nelson-Siegel model, calculate the three factor has effect on the term structure of interest rate, and the error correction model (VECM) combined with macroeconomic variables, build a hybrid model for China's interest rates. In the interpretation of meaning of the economy at the same time the variable model, the sample forecasting ability is higher than the general dynamic model of Nelson-Siegel.
The full text is divided into 6 parts. The contents of each part are arranged as follows. The first part is the introduction, which introduces the background and significance of the research, research methods and research tools, main contents and detailed steps. Finally, it points out the innovation of the article.
The second part is the literature review on the structure of domestic and foreign interest rate term. Starting from the traditional theory of term structure of interest rates, to the modern interest rate term structure model including static term structure estimation method, dynamic model, hybrid model and macro finance model. On the domestic interest rate term structure also from the static model, dynamic model, Nelson-Siegel model, bank the term structure of interest rates between different angles are reviewed.
The third part is to introduce the principle and test model and realization method. Firstly introduces the model of the Nelson-Siegel model and the level factor, slope factor and the curvature factor. Then the mathematical meaning gives the concept of state space model and the concept of Calman filter, filter process and how to estimate the parameters of the model are introduced. The numerical calculation method used in this paper. Finally introduced a series of time series model used in this paper, including the vector auto regression model, vector error correction model, the Johansen cointegration test base.
The fourth part is the data sources and processing. The choice of the inter-bank bond market quotations data as the research object of the paper reasons are explained. Then the data were preprocessed by Fama-Bliss method and bilateral inter-bank bond price data into zero coupon bond yields on the data.
The fifth part is the empirical research, the research work mainly includes the estimation of model parameters and prediction ability. The first step is based on pre order on the three factor NS model and scale parameter estimation. The second step of the three factor is the establishment of vector autoregressive model and test the predictive ability. The fourth step will be the three factor and the macroeconomic variables are combined to set up error correction model to examine factors and macroeconomic data and improve the prediction ability of the model.
Finally, the conclusions of the research are summarized, the suggestions are given, the shortcomings are put forward and the direction of further research is prospected.
The study found that, first, the level, slope, curvature factor three and the consumer price index (CPI) and industrial added value (IP) there is a long-term stable cointegration relationship. Second, found that the error correction model after adding macro variables (VECM) in the short and medium-term prediction ability are significantly improved, stronger than the unconstrained vector autoregressive model (UVAR model).
The innovation of this paper lies in the following three aspects. Firstly, the three factors and macroeconomic variables by Nelson-Siegel model estimated model to establish a stable model and improve the prediction ability of ordinary vector autoregressive model through error correction. Secondly, this paper from the general choice of the exchange bond data of interest rate the term structure is different, in favor of the inter-bank market quotations data for research. Finally, the numerical method combined with the average value of three factor least squares method to estimate the initial value as Calman filter three factors, to improve the estimation accuracy.
The future development direction of this paper are the following two aspects. First, starting from the construction of the model in this paper, to find appropriate macroeconomic factors improve the prediction ability of the model along the modeling method in this paper. Second, according to Koopman (2010) in the first method, the scale parameter is considered as the fourth variable factors into the state the space model, the extended Calman filter (Extended Kalman) to estimate the parameter of the model, and then using the method in this paper, the scale parameter is also introduced to the whole model design, improve the model fitting effect and prediction ability.

【学位授予单位】：西南财经大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：F832.51;F224;F812.5

【共引文献】