基于GARCH-Copula模型的投资组合在险价值测度应用研究

发布时间：2018-02-13 06:36

本文关键词： Copula函数 GARCH模型投资组合VaR GED分布　出处：《天津财经大学》2013年硕士论文　论文类型：学位论文

【摘要】：金融风险测度一直以来都是现代金融风险管理的重要内容。量化金融资产的风险,对金融风险控制具有很重要的意义。鉴于金融资产数据大多具有尖峰、厚尾以及波动集聚性等特征,在对其进行建模量化资产风险时,就需要根据资产数据特征来选定合适的模型和分布假定。现阶段主要使用VaR方法量化金融资产或者资产组合风险值,使用的模型大多是GARCH族模型、Copula函数以及两者的结合,并在不同的分布假定下,求解VaR值。尤其是在分布假定方面,学者们在进行深入研究的基础上,使用了厚尾的t分布、GED分布以及极值分布等分布形式来更好的拟合数据,构建模型。在资产组合中,每一资产数据都可能拥有各自不同的特征,因而需要对每一资产收益数据进行分布拟合,确定其所更接近的分布模型,最终由Copula函数进行组合得到联合分布函数。由此可以得到更加充分考虑资产组合中各个资产数据特征的GARCH-Copula模型,进而得到的在险价值VaR将更加有效。针对金融资产组合中各个资产数据特征,论文使用具有厚尾特性的t分布和GED分布来对样本数据进行拟合,之后构建基于t分布和GED分布的GARCH-Copula模型,得到资产组合模拟收益率数据的VaR；作为对比,论文同时分别使用基于t分布的GARCH-Copula模型和基于GED分布的GARCH-Copula模型,用于测度资产组合的VaR；论文也引入基于GARCH模型和不同分布假定的计算VaR的简单方法。论文正是在较为充分考虑资产组合中单个数据特征的前提下,选用更适合资产分布假定以及结合GARCH模型,描述数据所特有的波动集聚性的同时,构建一个稳健的模型,用于测算资产组合的VaR。同时结合每一资产数据服从的分布,使用合适的Copula函数对其进行组合,得到基于t分布和GED分布组合的GARCH-Copula模型,以测度资产组合的VaR。
[Abstract]:Financial risk measurement has always been an important part of modern financial risk management. Quantifying the risk of financial assets is of great significance to financial risk control. When modeling and quantifying asset risk, the characteristics of thick tail and volatility agglomeration, It is necessary to select appropriate models and distribution assumptions according to the characteristics of asset data. At present, the VaR method is mainly used to quantify the risk value of financial assets or asset combinations. Most of the models used are Copula functions of GARCH family model and their combination. Under different distribution assumptions, the VaR value is solved, especially in the distribution assumption. On the basis of in-depth research, the authors use the t distribution and extreme value distribution of the thick tail to fit the data better. Build models. In a portfolio, each asset data may have its own different characteristics, so it is necessary to fit each asset income data in order to determine the distribution model that is closer to each asset income data. Finally, the Copula function is combined to get the joint distribution function, which can get the GARCH-Copula model which fully considers the characteristics of each asset data in the asset portfolio, and then the riskier value VaR will be more effective. Aiming at the characteristics of each asset data in the financial asset portfolio, this paper uses the t distribution and GED distribution with thick tail to fit the sample data, and then constructs the GARCH-Copula model based on t distribution and GED distribution. As a comparison, GARCH-Copula model based on t distribution and GARCH-Copula model based on GED distribution are used in this paper. This paper also introduces a simple method to calculate GARCH based on GARCH model and different distribution assumptions. On the premise of fully considering the characteristics of individual data in the asset portfolio, this paper chooses a more suitable asset distribution assumption and a GARCH model to describe the unique volatility agglomeration of the data, and constructs a robust model at the same time. At the same time, combined with the distribution of each asset data, the Copula function is used to combine it, and the GARCH-Copula model based on the combination of t distribution and GED distribution is obtained to measure the value of the portfolio.
【学位授予单位】：天津财经大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：F830.59;F224

【参考文献】