基于Copula理论的外汇市场投资组合分析

发布时间：2018-01-20 11:48

本文关键词： 投资组合 Pair Copula Gaussian DAG Copula 混合C藤Copula VaR　出处：《天津科技大学》2013年硕士论文　论文类型：学位论文

【摘要】：对于投资者来说,投资于相关程度较低的多品种金融资产组合,可以有效降低投资风险,那么如何选择最优的投资组合便成为投资者关注的焦点问题。为了克服传统投资组合风险管理模型中风险度量方法、正态及线性相关假设的不足,我们不仅应该运用更科学的风险度量方法,还应该构建更合理的联合分布,这对于投资组合风险的度量以及最优投资策略的选择都有至关重要的作用。在风险测度方面,VaR已成为应用最为广泛的风险度量方法,是金融风险管理研究的重点；而Copula理论则为构建投资组合的联合分布函数提供了一条便捷、科学的途径,该联合分布可以满足金融资产非正态、非线性相关的假设,因此可以克服传统风险管理模型的不足。本文研究内容主要是多金融资产投资组合的风险分析、度量问题。基于Copula理论,以VaR作为风险度量指标,将Copula函数、GARCH模型、SV模型、VaR以及Monte Carlo方法有机结合,解决了多金融资产的非正态、非线性相关问题,为最优投资组合的选择与投资风险的度量提供了一种新的解决方法。本文第一部分实证研究对象是中国外汇市场四种主要外汇资产的投资组合,先通过GARCH与SV类模型的比较研究确定了单个风险资产收益率的边际分布波动模型；然后运用PC算法估计了表示资产间相关结构的DAG,并基于Pair Copula建模思想构建了Gaussian DAG Copula模型,该模型能更好的描述资产组合间的相依结构；在Gaussian DAG Copula模型的基础上利用Monte Carlo方法计算了投资组合的VaR,并通过返回检验证明了模型的有效性；最后,赋予四个外汇资产不同投资权重,计算了不同组合在不同置信水平下的VaR值。第二部分的实证研究是以中国外汇市场上七种外汇资产投资组合为对象,在该部分的研究中采用基于Pair Copula高维建模方法的混合C藤Copula模型,构建了反映多个资产收益实际分布和相依性的联合分布函数。该模型根据变量间相关性强弱确定变量顺序,并对各Pair Copula都选择最优函数族,故能更好描述资产组合的相依结构。在此基础上,利用Monte Carlo方法计算了七种外汇资产投资组合的VaR,并通过实证分析验证了该模型的有效性,最后,同样赋予七种外汇资产不同投资权重,计算了不同组合在不同置信水平下的VaR值。两部分的实证研究结果都表明,所建立的风险分析模型都能更好的刻画资产间的相关结构,在风险度量方面更加的科学、准确。
[Abstract]:For investors, investing in a variety of financial asset portfolios with low correlation can effectively reduce the investment risk. In order to overcome the shortcomings of the traditional portfolio risk management model, the deficiency of the normal and linear related assumptions is overcome in order to solve the problem of how to choose the optimal portfolio. We should not only use more scientific risk measurement methods, but also build more reasonable joint distribution. This plays an important role in portfolio risk measurement and the choice of optimal investment strategy. In risk measurement, VaR has become the most widely used risk measurement method. Is the focus of financial risk management research; The Copula theory provides a convenient and scientific way to construct the joint distribution function of the investment portfolio. The joint distribution can satisfy the assumption that the financial assets are not normal and nonlinear. Therefore, we can overcome the shortcomings of the traditional risk management model. This paper mainly focuses on the risk analysis and measurement of multi-financial asset portfolio based on Copula theory. Taking VaR as the risk measure index, the Copula function GARCH model and the SV model are combined with the Monte Carlo method. Solve the non-normal and nonlinear related problems of multi-financial assets. This paper provides a new solution for the choice of optimal portfolio and the measurement of investment risk. The first part of this paper is the empirical study of the four major foreign exchange portfolio in China's foreign exchange market. Firstly, through the comparative study of GARCH and SV models, the marginal distribution volatility model of return rate of individual venture assets is established. Then we use PC algorithm to estimate the structure of correlation between assets, and build Gaussian DAG Copula model based on the idea of Pair Copula modeling. The model can better describe the dependent structure of portfolio. On the basis of Gaussian DAG Copula model, the Monte Carlo method is used to calculate the portfolio VaR, and the validity of the model is proved by the return test. Finally, four foreign exchange assets are given different investment weights. The VaR values of different combinations at different confidence levels are calculated. The second part of the empirical study is based on the seven foreign exchange asset portfolios in China's foreign exchange market as the object. In this part of the study, a hybrid Copula model based on Pair Copula high dimensional modeling method is used. The joint distribution function which reflects the actual distribution and dependence of multiple asset returns is constructed. The model determines the order of variables according to the correlation between variables, and selects the optimal function family for each Pair Copula. Therefore, it can better describe the dependent structure of portfolio. On this basis, the VaR of seven kinds of foreign exchange portfolio is calculated by using Monte Carlo method. And through the empirical analysis to verify the effectiveness of the model, finally, the same gives seven kinds of foreign exchange assets with different investment weights. The VaR values of different combinations under different confidence levels are calculated. The empirical results of the two parts show that the established risk analysis model can better describe the correlation structure between assets. More scientific and accurate in risk measurement.
【学位授予单位】：天津科技大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：F224;F832.6

【相似文献】