关键词: 极端金融风险, 极值理论（EVT）, VaR, CVaR-EVT, POT, BMM

Abstract: The Basel Accord II has suggested that Value-at-Risk (VaR) is the advanced measure model of Risk management and it is widely used by the financial institutions and regulation bodies worldwide. Howerver, VaR doesn’t satisfy the coherent risk measure, especially it couldn’t be used to measure the fat-tail distribution of risk such as the operational risk and extreme financial risks. So, this paper takes the Extreme Value Theory（EVT）and Conditional Value at Risk (CVaR)，and return level of Block Maxima Models(BMM) comparing with VaR to measure extreme quantile estimate of financial risk on Standard and Poor’scomposite index (SPI), Hang Seng index(HSI), Shanghai composite index(SHI) and Shenzhen sub-index (SZI) in stock markets. The findings are: (1) The estimate values of CVaR-EVT at 99% quantile on Point estimate and 95% maximum likelihood(ML) and bootstrap(Bs) confidence intervals are better than VaR by using Peak Over Threshold (POT), while BMM shows that the higher the return level (daily loss) is, the longer the period of block is; and (2) The Extreme risk in Chinese stock market is far higher than international stock markets, but it is becoming stable and convergent as its counterparts by using ML and BS to estimate the daily loss of models from 1991 to 2008.

Key words: Extreme, Financial, Risk, Extreme, Value, Theory, Value-at-Risk, Conditional, Value-at-Risk-EVT, Peak-Over-Threshold, Block, Maxima, Models