CAJ | 학술논문

以广义帕累托分布为边缘分布函数,引入de Haan矩估计和Bootstrap抽样方法定量选取阈值,进而运用三种Copula簇方法研究了我国台湾和韩国股票市场之间的相关关系,然后比较了单参数与双参数Copula的拟合效果,测算了两市场遭遇极端市场风险的条件概率,探讨了双参数Archimedean Copula函数在构建联合分布中的应用.研究结果表明：BB7 Copula较好地刻画了两市场尾部相关的非线性、非对称特征,且相关结构拟合度较好,表明两市场在低迷时期的相关性明显高于其活跃时期的相关性,但通过对两市场构建组合难以有效降低风险.同时回测检验显示Copula-GPD模型能有效测度两市场组合的极值风险.另外,当韩国指数日波幅出现下降超过7％时,台湾加权指数也出现同样日波幅下跌的概率为5.72％.
이엄의파루탁분포위변연분포함수,인입de Haan구고계화Bootstrap추양방법정량선취역치,진이운용삼충Copula족방법연구료아국태만화한국고표시장지간적상관관계,연후비교료단삼수여쌍삼수Copula적의합효과,측산료량시장조우겁단시장풍험적조건개솔,탐토료쌍삼수Archimedean Copula함수재구건연합분포중적응용.연구결과표명：BB7 Copula교호지각화료량시장미부상관적비선성、비대칭특정,차상관결구의합도교호,표명량시장재저미시기적상관성명현고우기활약시기적상관성,단통과대량시장구건조합난이유효강저풍험.동시회측검험현시Copula-GPD모형능유효측도량시장조합적겁치풍험.령외,당한국지수일파폭출현하강초과7％시,태만가권지수야출현동양일파폭하질적개솔위5.72％.
The relationship between financial markets is growing in complexity because of the rapid development of financial economic globalization and financial internationalization.TWII and KOSPI stock markets could represent the continuous growth of emerging economies in Asia,and their complex relationships.Although the correlation between financial assets in Asian markets has been wildly discussed,there are several drawbacks on existing studies.In order to overcome these problems,the Copula functions are employed to provide a flexible and useful statistic tool to construct the multivariate joint distribution and to analyze the multivariate dependency structure.The present study aims to apply the Copula theory in the financial field including modeling and analysis of financial assets portfolio and risk measurement.This paper applies several types of Copula to investigate the dependence structure between logarithmic rate of return of TWII and KOSPI.The Copula models can capture nonlinear asymmetric and tail dependence.In addition,this study discusses the dependency of indicators derived from the Copula function.Our discussion results show that the joint generating function method,which describes the tail correlation by combining the coefficient of tail dependence and slowly changing function,is better than the common tail correlation coefficient method.Our analysis is based on the stock database from Yahoo Financial Data which includes 3,175 pairs of valuable data from November 5,1998 to September 30,2011.The first part of this study estimates the parameters of the Copula function.We employ the margin inference method which makes estimation more feasible by estimating the parameters for the univariate marginal distribution and the Copula function.We discuss the key issue of applying the Copula model into the data fitting and selection of Marginal distribution function.The computation result shows that the de Haan and bootstrap methods are effective for quantities＇ threshold selection of GPD model.These methods avoid the threshold selection uncertainty of MEF and SHAPE estimator.GPD model can describe the fat-tail and asymmetrical distribution features in addition to estimating the extreme risk prior to normal distribution.Diagnostic plot of the GPD fit is Tail of Underlying.This plot indicates that GPD appears to fit the distribution of threshold excess fairly well.The second part discusses the rationale of selecting the Coupla function.Based on AIC and BIC minimum criterion,the Gumbel Copula function shows that a correlation between these two markets exists in only upper tail for single parameter Copula.However,the two-parameter BB7 Copula has the good Goodness-of-fit test and practical value among the selected Copulas.Through the relationship analysis of the TWII stock market and the KOSPI stock market,we found that these two stock markets have larger lower tail dependence than upper tail dependence.These findings illustrate that two different market returns are more likely to correlate with each other during market downturns than upturns.These findings need careful interpretations if selected Copulas with one or two parameters need to be simulated simultaneously.Consequently,it is necessary to consider the asymmetry tail dependence structure when studying risk management and portfolio investment.In summary,the accuracy of Copula models largely depends on how well the marginal distribution functions fit the data.The GPD model needs to estimate the threshold values in order to exactly fit the margin distribution of Copula functions.Our work analyzes a portfolio composed of the TWII and KOSPI indices with daily returns and estimates of the one-day ahead of VaR position following the flexible Copula-GPD model.Furthermore,the testing results show that Copula-GPD model is suitable for measurement of tail risk of TWII and KOSPI portfolio.It should be noted,however,that creating an investment portfolio of two different stock markets won＇t be necessarily helpful for reducing investment risk.Finally,this study also shows that the conditional probability of the TWII will fall 5.72％ if more than 7％ fall occurs at the KOSPI stock market.