發(fā)布時間：2019-06-04 21:10

【摘要】：隨著金融市場的不斷變化,金融資產(chǎn)之間的相關(guān)關(guān)系越來越復(fù)雜,呈現(xiàn)出非線性、非對稱性和尾部相關(guān)的特性,基于線性相關(guān)關(guān)系的分析方法有的時候不能準(zhǔn)確反映金融市場的相關(guān)關(guān)系,同時現(xiàn)實(shí)中金融資產(chǎn)收益率存在尖峰厚尾的特征,明顯具有非正態(tài)特性與非線性相關(guān),因此有時候采用傳統(tǒng)VaR計(jì)算方法不盡合理,這時有必要采用合理的方法描述收益率的實(shí)際分布與相關(guān)性。而運(yùn)用COPULA函數(shù)方法可以構(gòu)造靈活的多元分布函數(shù),很好的描述金融資產(chǎn)收益率的真實(shí)分布與相關(guān)關(guān)系,從而可以建立起更為有效的風(fēng)險度量模型,所以運(yùn)用COPULA函數(shù)研究金融資產(chǎn)風(fēng)險價值具有重要的理論價值與運(yùn)用意義。 論文首先介紹了GARCH模型族,并且對GARCH模型殘差的分布進(jìn)行了研究,引入了兩種厚尾分布t分布與GED分布；然后給出了廣義帕累托分布(GPD)的定義和閥值的選擇方法；接著本文詳細(xì)的介紹了COPULA函數(shù)的定義、性質(zhì)以及常用的五種COPULA函數(shù),并且給出了COPULA函數(shù)的估計(jì)方法,以及最優(yōu)COPULA函數(shù)的選擇方法。在此基礎(chǔ)上,引入了GARCH-COPULA、GPD-COPULA和GARCH-GPD-COPULA三種計(jì)算風(fēng)險價值的VaR模型。在實(shí)證部分首先運(yùn)用歷史模擬法與分析法計(jì)算出資產(chǎn)組合在不同分位數(shù)下的風(fēng)險價值(VaR)。然后,運(yùn)用蒙特卡洛模擬法計(jì)算出三種模型GARCH-COPULA、GPD-COPULA、GARCH-GPD-COPUL對應(yīng)的風(fēng)險價值(VaR)。最后,對五種結(jié)果在1%,2%,3%,4%,5%,10%分位數(shù)下的VaR運(yùn)用失敗頻率法加以檢驗(yàn),并且進(jìn)行比較,實(shí)證結(jié)果表明基于GARCH-GPD-COPULA方法計(jì)算的VaR在樣本內(nèi)失敗率是最低的,說明它估計(jì)的風(fēng)險價值最接近真實(shí)風(fēng)險價值。 本文的創(chuàng)新主要體現(xiàn)在以下幾個方面：(1)系統(tǒng)全面的總結(jié)了COPULA函數(shù)的定義,分類以及估計(jì)方法。(2)在運(yùn)用GARCH模型對邊緣分布進(jìn)行擬合時,考慮了殘差在正態(tài)分布,t分布與廣義誤差分布(GED)的不同情況,最后選擇出最優(yōu)的GARCH模型。(3)在前人對COPULA函數(shù)研究的基礎(chǔ)上,將CARCH-COPULA和GPD-COPULA進(jìn)行結(jié)合,提出了GARCH-GPD-COPULA函數(shù),并進(jìn)行了相應(yīng)的實(shí)證分析。
[Abstract]:With the continuous change of financial markets, the correlation between financial assets is becoming more and more complex, showing nonlinear, asymmetric and tail-related characteristics. The analysis method based on linear correlation sometimes can not accurately reflect the correlation of financial market. At the same time, in reality, the rate of return on financial assets has the characteristics of sharp peak and thick tail, which obviously has non-normal characteristics and nonlinear correlation. Therefore, sometimes it is unreasonable to use the traditional VaR calculation method, so it is necessary to use a reasonable method to describe the actual distribution and correlation of the rate of return. By using the COPULA function method, a flexible multivariate distribution function can be constructed to describe the real distribution and correlation of the rate of return on financial assets, so that a more effective risk measurement model can be established. Therefore, it is of great theoretical value and application significance to use COPULA function to study the risk value of financial assets. In this paper, the GARCH model family is introduced, and the residual distribution of GARCH model is studied, and two kinds of thick tail distribution t distribution and GED distribution are introduced, and then the definition of generalized Pareto distribution (GPD) and the selection method of threshold value are given. Then, the definition and properties of COPULA function and five kinds of COPULA functions are introduced in detail, and the estimation method of COPULA function and the selection method of optimal COPULA function are given. On this basis, three VaR models, GARCH-COPULA,GPD-COPULA and GARCH-GPD-COPULA, are introduced to calculate the risk value. In the empirical part, the risk value (VaR). Of asset portfolio under different quartile is calculated by using historical simulation method and analysis method. Then, the risk value (VaR). Corresponding to GARCH-COPULA,GPD-COPULA,GARCH-GPD-COPUL of three models is calculated by Monte Carlo simulation method. Finally, the failure frequency method was used to test and compare the five kinds of results in 1%, 2%, 3%, 4%, 5%, 10% of the quartile, and the failure frequency method was used to test and compare the five kinds of results at 1%, 2%, 3%, 4% and 5%, respectively. The empirical results show that the failure rate of VaR calculated by GARCH-GPD-COPULA method is the lowest in the sample, which indicates that the estimated risk value is the closest to the real risk value. The innovation of this paper is mainly reflected in the following aspects: (1) the definition, classification and estimation method of COPULA function are summarized systematically and comprehensively. (2) when the GARCH model is used to fit the edge distribution, the residual error in the normal distribution is considered. According to the difference between t distribution and generalized error distribution (GED), the optimal GARCH model is selected. (3) on the basis of previous research on COPULA function, CARCH-COPULA and GPD-COPULA are combined to propose GARCH-GPD-COPULA function. And the corresponding empirical analysis is carried out.
【學(xué)位授予單位】：天津財(cái)經(jīng)大學(xué)
【學(xué)位級別】：碩士
【學(xué)位授予年份】：2012
【分類號】：F224;F830.9

【參考文獻(xiàn)】

相關(guān)期刊論文 前10條

1 李鋒;劉澄;;基于極值理論的金融風(fēng)險研究[J];商業(yè)研究;2010年05期

2 歐陽資生,龔曙明;廣義帕累托分布模型:風(fēng)險管理的工具[J];財(cái)經(jīng)理論與實(shí)踐;2005年05期

3 余萍,龔金國;描述金融市場相關(guān)結(jié)構(gòu)的一種新工具——Copula[J];東莞理工學(xué)院學(xué)報;2005年05期

4 韋艷華,張世英;金融市場的相關(guān)性分析——Copula-GARCH模型及其應(yīng)用[J];系統(tǒng)工程;2004年04期

5 程炳巖;丁裕國;張金鈴;江志紅;;廣義帕雷托分布在重慶暴雨強(qiáng)降水研究中的應(yīng)用[J];高原氣象;2008年05期

6 司繼文,蒙堅(jiān)玲,龔樸;國內(nèi)外股票市場相關(guān)性的Copula分析[J];華中科技大學(xué)學(xué)報(自然科學(xué)版);2005年01期

7 任仙玲;張世英;;基于核估計(jì)及多元阿基米德Copula的投資組合風(fēng)險分析[J];管理科學(xué);2007年05期

8 韋艷華,張世英,孟利鋒;Copula理論在金融上的應(yīng)用[J];西北農(nóng)林科技大學(xué)學(xué)報(社會科學(xué)版);2003年05期

9 單國莉,陳東峰;一種確定最優(yōu)Copula的方法及應(yīng)用[J];山東大學(xué)學(xué)報(理學(xué)版);2005年04期

10 桂文林;韓兆洲;潘慶年;;POT模型中GPD“厚尾”性及金融風(fēng)險測度[J];數(shù)量經(jīng)濟(jì)技術(shù)經(jīng)濟(jì)研究;2010年01期

相關(guān)博士學(xué)位論文 前1條

1 花擁軍;極值理論在中國股市風(fēng)險度量中的應(yīng)用研究[D];重慶大學(xué);2009年

相關(guān)碩士學(xué)位論文 前2條

1 丁林榮;GPD模型與巨災(zāi)保險[D];華東師范大學(xué);2008年

2 王皓;極值理論在測度中國股市VaR中的應(yīng)用與比較[D];浙江大學(xué);2008年

，

本文編號：2493004