基于组合预测方法的国债利率期限结构实证研究

发布时间：2018-01-10 07:05

本文关键词：基于组合预测方法的国债利率期限结构实证研究　出处：《东北财经大学》2012年硕士论文　论文类型：学位论文

【摘要】：本文研究的对象是利率期限结构,利率期限结构是指不同期限的利率与该期限之间的关系,它需要用债券的即期利率与其剩余到期期限之间的关系来表示。研究利率期限结构具有重要的理论和实践意义,它可为确定基准利率提供理论支撑,为债券和金融衍生品的定价提供理论基础,同时可以为金融风险控制和利率风险管理提供操作依据,有利于设计合理的金融产品,有利于发现市场中的套利机会,提高金融市场的效率。利率期限结构相关理论的研究大体经历了传统的定性研究阶段和现代定量精确研究阶段。传统的定性研究大体有预期理论和市场分割理论。后来人们更多的借鉴数学模型,从而更精确的预测和模拟利率,提出的模型归纳起来主要有逐点递推法、样条函数法和简约模型法等。基于不同利率期限结构模型所采用的具体数学模型不同,因此各模型存在不同的优缺点,而实际经济环境中利率期限结构则复杂多变,这导致各模型对实际利率期限结构的拟合优度会出现阶段性的波动,不存在拟合优度稳定不变的单一利率期限结构模型,因此仅采用单一利率期限结构模型对我国利率期限结构进行估计,效果并非最好。为了提高利率的拟合预测效果,本文借助了组合预测方法,组合预测方法是把各种模型组合起来,按模型的拟合优度大小分配给各模型适当的权重,拟合优度高的模型会分配给较大的权重,从而分散单一预测模型的不确定性,以提高预测整体的精确度和稳定性。本文对各利率期限结构模型对国债的理论定价能力进行了实证比较,每一样本交易日随机选取部分样本用于各模型参数的估计,先用单一利率期限结构模型对国债利率期限结构进行实证研究,选用了多项式样条函数模型、Nelson-Siegel模型和Nelson-Siegel模型的扩展Svensson模型三个单一模型。在完成单一模型的估计后,本文利用国债价格拟合误差最小这个组合优化准则,对三个单一模型进行组合,估计出三个模型的组合权重。利用当日估计出的组合权重,借助组合预测方法形成组合预测模型,并利用该模型对当日剩余国债样本进行理论定价。在评价各利率期限结构模型对真实利率曲线的拟合优度上,本文选取的准则是对国债价格的预测能力,即通过贴现求出附息国债的理论定价,并与实际交易价格进行比较,用定价的精确度和稳定性评价各利率期限结构模型的优劣。在进行详细的统计分析对比后,本文最后得出结论：本文提出的基于组合预测方法的利率期限结构模型不管是在国债理论定价精确度,还是在理论定价稳定性上都优于其他三个单项模型(多项式样条函数模型、Svensson模型和Nelson-Siegel模型),组合预测方法对各单项模型的组合达到了预定的效果,其组合了各单项模型对实际利率期限结构曲线反映的特点,对拟合优度高的模型赋予了较高的组合权重,使组合在一起的模型强于任一单项模型。本文在最后提出了今后研究的方向。在今后的研究过程中,会更加注重单项模型的选择,在组合模型中加入更多的单项模型以提高组合模型的对实际利率期限结构曲线的拟合优度。同时也会更加注重组合权重参数估计算法的研究,从而使估计出的权重更精确。
[Abstract]:The object of this paper is the term structure of interest rate, the interest rate term structure refers to the relationship between the interest rate and the different period of time, it needs to use the spot rate of bonds between the remaining maturity of the form. It has important theoretical and practical significance of the study on the interest rate term structure, which can provide theoretical support for the determination of the benchmark interest rate and provide a theoretical basis for the bonds and financial derivatives pricing, and can provide operation basis for the control of financial risk and interest rate risk management, is conducive to the rational design of financial products, is conducive to the arbitrage opportunity in the market, improve the efficiency of the financial market.
Study on the related theory of the term structure of interest rates has undergone qualitative research stage of traditional and modern precise quantitative research stage. Traditional qualitative research has largely expected theory and market segmentation theory. Then more and more people from the mathematical model, thus more accurate prediction and Simulation of the interest rate, the proposed model can be summed up by recursive method the spline function method, and the simple model.
The mathematical model of the structure model based on different interest rate term is different, so the models have different advantages and disadvantages, and the term structure of interest rate in actual economic environment is complex and changeable, which leads to the goodness of the actual model of the term structure of interest rates will appear periodic fluctuations, there is no single interest rate term structure model fitting is stable, so only by a single interest rate term structure model to estimate the term structure of interest rate in China, the effect is not the best. In order to improve the rate of fitting prediction results based on the combined forecasting method, combination forecasting method is the combination of model, the weights assigned to each model according to the goodness of fit of proper size the model fitting degree is high, the model will be assigned a larger weight, in order to disperse the single prediction model of uncertainty, in order to improve the prediction accuracy of the overall And stability.
In this paper, each term structure model are empirically compared to the bond pricing theory, estimation of each trading day samples randomly selected part samples for each model parameter, the term structure of interest rates for empirical research with a single interest rate term structure model, choose the polynomial spline function model, Nelson-Siegel model and Nelson-Siegel model the extended Svensson model three single models. In the estimation of single model, the minimum of the combinatorial optimization criterion of bond price fitting error of three single model group, estimate the combination weights of the three models. The combination weights estimated by the day, the combination forecasting method of formation combination forecasting model and on the day of the remaining sample bonds pricing theory by using the model. In the evaluation of the interest rate term structure model of the real interest rate curve The goodness of fit, the criterion is the ability to predict bond prices, namely the discount for treasury bond pricing theory, and compared with the actual transaction price, the evaluation model of interest rate term structure with the merits of pricing accuracy and stability. In the process of detailed comparative analysis, this paper concludes conclusion: Based on the model of interest rate term structure combination forecast method in bond pricing accuracy, or in the theory of pricing stability is better than that of the other three individual models (polynomial spline function model, Svensson model and Nelson-Siegel model), a combination model of each single model to achieve the intended effect. The combination of the characteristics of each single model of actual interest rate term structure curve reflects the combination of the right fit high model gives higher weight to group The combined model is stronger than any single model.
At the end of this paper, the future research direction is put forward. In the future research, will pay more attention to the selection of a single model, adding more individual models to improve the structure of the combined model actual interest rate curve fit in the combined model. At the same time will be more filling recombination algorithm of weight parameter estimation. In order to make the estimated weight more accurate.

【学位授予单位】：东北财经大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：F810.5;F820;F224

【参考文献】