基于M-SCAD方法的利率期限结构构造的研究

发布时间：2018-02-22 01:37

本文关键词： 利率期限结构样条函数节点选取惩罚分位数回归 M-SCAD　出处：《中国科学技术大学》2014年硕士论文　论文类型：学位论文

【摘要】：利率期限结构,它是用来描述在一个时点上,具有相同风险度的条件下,对应的收益率与它对应的时间长度之间的关系,在金融领域当中,该课题一直是一个重要的研究热点。随着中国利率市场化进程的不断推进,这使得开展利率期限结构的相关研究越发重要。因此,构造一条可靠稳定完整的利率期限结构曲线越来越重要。本文第一部分首先回顾债券市场的基本理论,包括债券定价公式和多种期限结构曲线的基本知识。继而,详尽归纳概括了国内外学者对于利率期限构造的一些静态模型以及对应的优缺点,国内外学者的一些模型有：多项式样条函数模型,指数样条基模型,叠加指数函数模型,B-样条基模型,平滑样条基模型。通过理论及已有实证模型的分析,分别对各个模型的优缺点进行了总结。本文第二部分包括第三章和第四章,为本文的重点,主要探讨如何构造利率期限结构,以及如何对构造模型的参数进行估计。虽然不少学者对利率期限构造的实证研究已日臻完善,可是仍需要我们不断完善对于它的研究。第三章,在综合考虑分位数回归与LAD-LASSO的基础上,将Wu(2009)提出的惩罚分位数回归方法引入到利率期限构造中,该模型采用非凹惩罚函数使得模型比较稳定,同时保留了LAD-LASSO的一些优点,可以同步实现参数估计与节点选择,并且最小一乘回归是分位数回归的一种特例。第四章,在Fan(2001)提出的一个非凹惩罚函数SCAD(Smoothing Clipped Absolute Deviation)的基础上,采用大M估计方法(Huber1981),构造利率期限结构模型。该模型拓展了LAD-LASSO且可以同步实现节点选择和参数估计。我们在第四章采用这种方法来构造上海证券交易所发行债券的利率期限结构,实证分析表明,该方法有较好的稳健性,并且在样本外预测结果显示,与传统方法相比该模型可以选择合适的模型,提高预测的精度。文章的结尾,对本文主题内容和重要的结果进行了概括归纳,并提出了两个日后可以作进一步研究的方向。
[Abstract]:The term structure of interest rates, which is used to describe the relationship between the corresponding rate of return and the corresponding length of time at a time point with the same degree of risk, in the field of finance. This subject has always been an important research hotspot. With the development of interest rate marketization in China, it is more and more important to carry out the research on the term structure of interest rate. It is more and more important to construct a reliable, stable and complete term structure curve of interest rate. The first part of this paper reviews the basic theory of bond market, including the basic knowledge of bond pricing formula and various term structure curves. Some static models of interest rate term construction and their advantages and disadvantages are summarized in detail. Some models of scholars at home and abroad include polynomial spline function model, exponential spline base model, and so on. The superposition exponential function model is composed of B-spline basis model and smooth spline basis model. The advantages and disadvantages of each model are summarized through the analysis of theory and existing empirical models. The second part of this paper includes the third chapter and the 4th chapter, which is the focus of this paper, mainly discusses how to construct the term structure of interest rate. And how to estimate the parameters of the construction model. Although many scholars' empirical research on term construction of interest rate has been improved, we still need to improve the research on it. On the basis of considering the quantile regression and LAD-LASSO, the penalty quantile regression method proposed by Wu Ying-2009) is introduced into the term construction of interest rate. The model uses non-concave penalty function to make the model more stable, while retaining some advantages of LAD-LASSO. Parameter estimation and node selection can be realized synchronously, and the least multiplicative regression is a special case of quantile regression. Chapter 4th, on the basis of a non-concave penalty function, SCAD(Smoothing Clipped Absolute selection, is proposed in chapter 4th. A term structure model of interest rate is constructed by using large M estimation method, which extends LAD-LASSO and synchronously implements node selection and parameter estimation. In Chapter 4th, we use this method to construct bonds issued by Shanghai Stock Exchange. Interest rate term structure, The empirical analysis shows that the proposed method is robust and the prediction results outside the sample show that the model can select a suitable model and improve the prediction accuracy compared with the traditional method. At the end of the paper, the main contents and important results of this paper are summarized, and two possible directions for further study are put forward.
【学位授予单位】：中国科学技术大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：F832.51;F224

【参考文献】