model based on multivariate conditional normal dis-distribution is greater than the tail of a normal distri-tribution assumptions, finding that Copula model bution, indicating their yield series present "spikes based on extreme value theory is dominant. and fat tail". Meanwhile, JB statistics are greater than the critical value of 5.9915 at 5% significant 3 EMPIRICAL RESEARCH level, refusing the null hypothesis that the yield se-ries obey normal distribution, that is, the two index 3.1 Selection and pre-processing of sample data yield series are not normally distributed, and we cannot use traditional mean - variance model to ana-This paper selects the daily closing price of Shang-log. hai Composite Index (SH) and S&P500 as study sample. The time period is from January 4, 2000 to 3.3 Stationary test May 28, 2008, and there are a total of 2023 and 2111

doi:10.1201/b18660-107

Chapter

model based on multivariate conditional normal dis-distribution is greater than the tail of a normal distri-tribution assumptions, finding that Copula model bution, indicating their yield series present "spikes based on extreme value theory is dominant. and fat tail". Meanwhile, JB statistics are greater than the critical value of 5.9915 at 5% significant 3 EMPIRICAL RESEARCH level, refusing the null hypothesis that the yield se-ries obey normal distribution, that is, the two index 3.1 Selection and pre-processing of sample data yield series are not normally distributed, and we cannot use traditional mean - variance model to ana-This paper selects the daily closing price of Shang-log. hai Composite Index (SH) and S&P500 as study sample. The time period is from January 4, 2000 to 3.3 Stationary test May 28, 2008, and there are a total of 2023 and 2111

Chapter

model based on multivariate conditional normal dis-distribution is greater than the tail of a normal distri-tribution assumptions, finding that Copula model bution, indicating their yield series present "spikes based on extreme value theory is dominant. and fat tail". Meanwhile, JB statistics are greater than the critical value of 5.9915 at 5% significant 3 EMPIRICAL RESEARCH level, refusing the null hypothesis that the yield se-ries obey normal distribution, that is, the two index 3.1 Selection and pre-processing of sample data yield series are not normally distributed, and we cannot use traditional mean - variance model to ana-This paper selects the daily closing price of Shang-log. hai Composite Index (SH) and S&P500 as study sample. The time period is from January 4, 2000 to 3.3 Stationary test May 28, 2008, and there are a total of 2023 and 2111

ABSTRACT

This paper selects the most representative Chinese Shanghai Composite Index and the United States S&P500 as the research object, establishes the marginal distribution through EGARCH model, depicts the distribution of market’s closing price indexes, describes the skew, spikes, and fat tail of market data and indicates the reason why it cannot establish appropriate model assuming the traditional normal distribution. And then we construct flexible multivariate distributions through Copula to adapt to the actual distributions of the portfolio investment in stock market, do a quantitative research about risk assessment, and make an analysis of VaR, volatility rate of portfolio risk under the premise that inter-dependence structure of assets are taken into consideration.