【摘要】：隨著金融市場的迅猛發(fā)展,金融衍生工具頻頻出新,市場的多元化對風險度量提出了更高的要求,對風險的定量分析顯得尤為重要。 本文在介紹譜風險度量理論及風險厭惡度量理論的基礎上,給出了雙曲型風險譜函數(shù)等三種風險譜函數(shù)的形式,得到了譜風險度量的估計量,從而構(gòu)造出投資組合的優(yōu)化模型。采用樣本外數(shù)據(jù)對模型的有效性進行Kupiec檢驗。實證部分計算了單個資產(chǎn)及多個資產(chǎn)投資組合的譜風險度量值。實證結(jié)果表明,風險厭惡因子和置信水平的選取均對單一資產(chǎn)的雙曲型譜風險度量值產(chǎn)生影響,風險厭惡因子可以作為譜風險度量的數(shù)值表征；對給定的置信水平和風險厭惡因子,隨著期望收益率的增加,高收益的股票所占權重逐漸增大。 將Copula函數(shù)運用到投資組合的譜風險度量模型中是本文的一個重要創(chuàng)新點。通過Copula函數(shù)研究資產(chǎn)之間的相依結(jié)構(gòu),可以提高SRM估計的準確性。核密度估計對樣本的擬合度高,本文選用其確定邊緣分布,選擇Copula函數(shù)描述尾部相依性。用極大似然估計和非參數(shù)方法估計Copula函數(shù)的參數(shù),結(jié)合經(jīng)驗Copula函數(shù),運用平方歐氏距離對參數(shù)的估計進行評價。最后,通過Monte Carlo模擬方法得到一種新的Copula-SRM算法。實證部分得到了五種Copula函數(shù)的參數(shù)估計值及Kendall秩相關系數(shù)和Spearman秩相關系數(shù)。實證結(jié)果表明上證指數(shù)和深證指數(shù)的日對數(shù)收益率存在較強的正相關,t-Copula模型能更好地擬合原始數(shù)據(jù),且Copula-SRM算法比傳統(tǒng)的SRM算法得到的結(jié)果更準確。
[Abstract]:With the rapid development of the financial market, the financial derivatives frequently come out new, the diversification of the market put forward higher requirements for risk measurement, the quantitative analysis of risk is particularly important. On the basis of introducing the theory of spectral risk measurement and the theory of risk aversion, this paper gives the form of three kinds of risk spectrum functions such as hyperbolic risk spectrum function, obtains the estimator of spectral risk measurement, and constructs the optimal model of investment portfolio. The validity of the model is tested by Kupiec with the data outside the sample. The empirical part calculates the spectral risk measures of individual assets and multiple asset portfolios. The empirical results show that the selection of risk aversion factor and confidence level have an effect on the hyperbolic spectral risk measure of a single asset, and risk aversion factor can be used as a numerical representation of spectral risk measurement. For a given confidence level and risk aversion factor, with the increase of expected rate of return, the weight of high yield stock increases gradually. It is an important innovation of this paper to apply the Copula function to the portfolio spectral risk measurement model. The accuracy of SRM estimation can be improved by studying the dependent structure of assets by Copula function. The kernel density estimation has a high fitting degree to the sample. In this paper, the edge distribution is determined and the Copula function is chosen to describe the tail dependence. The parameters of Copula function are estimated by maximum likelihood estimation and nonparametric method, and the estimation of parameters is evaluated by square Euclidean distance combined with empirical Copula function. Finally, a new Copula-SRM algorithm is obtained by Monte Carlo simulation. In the empirical part, the parameter estimates of five kinds of Copula functions and the Kendall rank correlation coefficients and Spearman rank correlation coefficients are obtained. The empirical results show that there is a strong positive correlation between the daily logarithmic returns of Shanghai Stock Exchange Index and Shenzhen Stock Exchange Index, and the t-Copula model can better fit the original data, and the Copula-SRM algorithm is more accurate than the traditional SRM algorithm.
【學位授予單位】：北京化工大學
【學位級別】：碩士
【學位授予年份】：2012
【分類號】：F224;F830.9

【參考文獻】

相關期刊論文 前10條

1 石媛昌,韓立巖;金融風險度量方法的新進展[J];首都經(jīng)濟貿(mào)易大學學報;2005年04期

2 史道濟;李t，

本文編號：2338223