除了買賣價上下限及資本適足性外，期貨保證金制度的設計更確保了期貨交易的安全性與期貨市場的完整性。如何在價格極端波動下設定一最適保證金水準，一方面考慮可以有效降低券商或期交所的違約風險；另一方面，又不會對市場交易機制或是市場流動性產生過度不利的影響是本研究訴求的核心。所以，本研究分別運用Generalized Extreme Value Distribution和Generalized Pareto Distribution去檢視當價格極端波動情況下之謹慎保證金水準為何及透過前向測試(Back testing)以確定不同模型成功率高低之比較。本文實證模擬結果發現: (1) 常態分配假設會造成保證金水準的低估。(2) 以損失期望值(Expected Shortfall)當做風險測量之檢視基準，Nikkei225在保證金水準的適足性最不足；S&P500、CAC40和DAX則較足夠。 (3) Generalized Pareto Distribution之損失期望值法對於補捉價格極端波動強況之能力來的比Generalized Pareto Distribution之保證金水準來的較佳。 (4) 運用Hillplot當做最適Threshold之指標並對於價格極端波動下之保證金水準設定為一個相當理想的風險測度工具。 Abstract Along with a price limit and capital requirement, the existence of a margin decrease the likelihood of a customer defaulting, a broker going bankrupt and systemic instability of the futures market. This paper applies two sub-theories of generalized extreme value distribution and generalized Pareto distribution inherited from extreme value theory to examine the prudent margin policy for price extremal movements. The theoretical framework focuses explicitly on tail returns, thereby properly computing prudent margin level for large levels of risk, This paper finds： (1) the assumption of normality to impose a smaller margin level since the presence of a fat-tail. (2) on the basis of margin insolvency using an expected shortfall, the margin requirements of stock index futures across contracts with a Nikkei225 contract being the more risky, and S&P500, CAC40, and DAX futures indexes are the least risky. (3) the ability to capture extreme price movements using expected shortfall is more suitable than the approach of the VaR based on generalized extreme value distribution. (4) the proxy of the appropriate threshold using an expected shortfall can capture well the extreme price movements and can be an excellent risk measure instrument to set the prudent margin level.
