基于时变Pair Copula-GAS的期货组合动态保证金设定

发布时间：2018-03-18 11:36

本文选题：时变Pair　切入点：Copula　出处：《中国科学技术大学》2017年硕士论文　论文类型：学位论文

【摘要】：本文从期货交易所的保证金设置问题出发,首先介绍了国内外期货交易所的保证金设置方法,包括国际上应用最为广泛的SPAN系统和TIMS系统。接着重点介绍了 Copula函数的相关理论、性质和特点,常见的椭圆Copula族与阿基米德Copula族,Pair Copula应用最广泛的两种藤结构C藤和D藤,一些较为常见的Copula参数估计方法,以及传统的时变相关Copula模型。随后本文重点介绍了广义自回归分数模型(GAS模型),以及该模型的几种特殊形式,并由该方法衍生出新的时变Copula模型。最后介绍了本文计算保证金的方法,利用EGARCH-t模型估计边缘分布密度函数,采用时变Pair Copula-GAS模型计算相关系数,使用Monte Carlo模拟计算VaR即为保证金水平。实证方面我们选取黄金、锌、螺纹钢作为研究对象,使用2012年7月30日至2015年7月28日的期货合约指数收盘价的对数收益率,总共2184个数据。从样本的基本统计信息中,我们发现样本具有高峰、厚尾特征,随后我们采用EGARCH-t模型求出相应边际分布,通过Pair Copula的C藤分解,运用GAS时变Copula来描述两两期货合约之间的尾部相依结构。最后针对期货收益率的相关性会随市场波动而发生变化的现象,用时变Pair Copula来估计组合风险值,并结合蒙特卡洛模拟的方法,计算出VaR,从而求出相应保证金数量,并与常相关模型下计算的保证金数量进行比较,得出时变Pair Copula-GAS模型更为准确和优越的结论。
[Abstract]:Starting from the issue of margin setting in futures exchanges, this paper first introduces the margin setting methods of futures exchanges at home and abroad. It includes the most widely used SPAN system and TIMS system in the world. Then, the related theories, properties and characteristics of Copula function are introduced, and the two most widely used rattan structures, C rattan and D rattan, are commonly used in the elliptical Copula family and the Archimedes Copula family. Some common Copula parameter estimation methods, as well as the traditional time-varying correlated Copula model, are introduced in this paper. Then, the generalized autoregressive fractional model gas model and several special forms of the model are introduced. A new time-varying Copula model is derived from this method. Finally, the method of calculating margin is introduced. The edge distribution density function is estimated by using EGARCH-t model, and the correlation coefficient is calculated by time-varying Pair Copula-GAS model. Using Monte Carlo to simulate and calculate VaR is the margin level. In the empirical aspect, we choose gold, zinc and rebar as the research objects, and use the logarithmic yield of the closing price of futures contract index from July 30th 2012 to July 28th 2015. From the basic statistical information of the sample, we find that the sample has the characteristics of peak and thick tail. Then we use the EGARCH-t model to calculate the corresponding marginal distribution, and decompose it by the C vine of Pair Copula. GAS time-varying Copula is used to describe the tail dependent structure between pairwise futures contracts. Finally, the time-varying Pair Copula is used to estimate the portfolio risk value. Combined with Monte Carlo simulation method, the corresponding amount of margin is calculated, and compared with the amount of margin calculated under the constant correlation model, it is concluded that the time-varying Pair Copula-GAS model is more accurate and superior.
【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】：2017
【分类号】：F224;F724.5

【参考文献】