基于混合Copula-GARCH-EVT模型的外汇相依性研究

发布时间：2018-10-04 21:37

【摘要】：国家宣布2005年7月21日起,人民币汇率实施市场自由调控为主,并参照一篮子货币共同驱动的浮动外汇机制,随着外汇市场机制的逐步改进,中国外汇市场渐渐走向成熟。从2016年10月1日开始,人民币成为继美元、欧元、英镑和日元之后的第五种货币正式加进特别提款权(SDR)货币篮子,代表着中国外汇市场向国际化走出关键的一步。随着外汇市场改革的进行,人民币汇率的波动日趋市场化,而全球化的日益深化也使中国与外国的贸易也越来越频繁。因此,一些涉外投资企业如商业银行、证券公司、基金公司和进出口公司所面临的外汇风险(也称汇率风险)也日益凸显。所以在外贸企业与国外贸易往来过程中,能够准确地测度外汇风险对于防范和规避货币风险来说尤为关键。而市场之间相依性研究是风险控制与测度的重要组成部分,由于Copula函数不仅可以对时间序列的相关性进行刻画,而且能够描述其相依结构,因此Copula函数在风险控制方面常常被作为探究多种资产间相依结构的主要工具。但是,市场中不同资产之间的相关结构并不是一成不变的,不会是固定的模式,因此单纯使用一种Copula函数难以有效地刻画不同资产之间的相关结构。但是混合Copula函数由具有不同性质的Copula函数组合而成,比单单使用一种Copula函数更能准确地表述资产间的这种关系。本文以2007年5月21日到2015年8月10日欧元对人民币与日元对人民币外汇数据为样本,将两种外汇收益率数据采用GARCH-EVT模型进行边际分布处理,实证显示,GARCH-EVT模型处理了残差序列自有的条件异方差及其厚尾特性,更加适合后续Copula拟合。我们选取阿基米德Copula函数中的Gumbel Copula、Clayton Copula 和 Frank Copula来构造混合Copula模型。研究结果表明:由Gumbel Copula、Clayton Copula和Frank Copula组合而得的混合Copula模型能更加准确地表述多种市场间的相关关系。
[Abstract]:The state announced that from July 21, 2005, the RMB exchange rate will be controlled mainly by the market, and with reference to the floating foreign exchange mechanism driven by a basket of currencies, with the gradual improvement of the foreign exchange market mechanism, China's foreign exchange market will gradually mature. Beginning on October 1, 2016, the yuan became the fifth currency after the dollar, euro, sterling and yen to be formally added to the SDR (SDR) basket, marking a crucial step towards internationalisation in China's foreign exchange market. With the reform of the foreign exchange market, the fluctuation of RMB exchange rate is becoming more and more market-oriented, and the deepening of globalization makes the trade between China and foreign countries more and more frequent. As a result, some foreign investment enterprises such as commercial banks, securities companies, fund companies and import and export companies are increasingly exposed to foreign exchange risk (also known as exchange rate risk). Therefore, in the process of foreign trade between foreign trade enterprises and foreign trade, it is very important to measure foreign exchange risk accurately to prevent and avoid currency risk. The study of market dependence is an important part of risk control and measurement, because the Copula function can not only describe the correlation of time series, but also describe its dependent structure. Therefore, the Copula function is often used as the main tool to explore the dependence structure of various assets in the aspect of risk control. However, the correlation structure between different assets in the market is not fixed, nor is it a fixed pattern. Therefore, it is difficult to describe the correlation structure between different assets effectively by using a Copula function. But the hybrid Copula function is composed of Copula functions with different properties, which is more accurate than using a single Copula function to describe the relationship between assets. From May 21, 2007 to August 10, 2015, the data of euro to RMB and yen to RMB are taken as samples, and the data of two kinds of foreign exchange rate are processed by GARCH-EVT model. The empirical results show that the GARCH-EVT model deals with the conditional heteroscedasticity and its thick-tailed characteristics of the residual sequence, and is more suitable for subsequent Copula fitting. We select Gumbel Copula,Clayton Copula and Frank Copula in Archimedes Copula function to construct mixed Copula model. The results show that the hybrid Copula model combined with Gumbel Copula,Clayton Copula and Frank Copula can more accurately describe the correlation between various markets.
【学位授予单位】：浙江工商大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：F224;F832.6

【参考文献】