Expected Shortfall Estimation: An Application on Stock Market

[[abstract]]由於金融商品發展蓬勃及全球金融環境瞬息萬變，使得市場參與者逐漸重視所面對的經營風險；然而在眾多風險測量工具中，期望短損為在給定充足資訊和損失大於風險值下，捕捉隱含損失的範圍，以彌補風險測量工具─風險值之不足。 許多文獻提出不同的期望短損估計模型，但對金融商品而言，尚未找出最適合的期望短損模型來測量所面臨的風險。本研究使用六種期望短損估計模型，並針對九個主要股市之股價指數進行期望短損估計並評估其結果之好壞，以找出適合用於估計股價指數之期望短損模型。 本研究實證結果發現：就整體表現來看，不管在持有一日或二日之期望短損下，所有模型在各種股價指數皆不因信賴水準的不同而改變其評估結果。從各個估計模型來觀察，不論在往前預測一期或是二期之期望短損下，QML GARCH模型在所有股價指數中表現最好，可以用來估計股價指數之風險情形；然而極值理論只有在往前預測一期之期望短損下表現最好，在預測兩期之期望短損下卻表現最差。最後GARCH-Normal模型和RiskMetrics模型不管在持有一日或二日之期望短損下皆表現不好，最不適合用此模型來測量股價指數的風險情況。[[abstract]]Because the financial commodities develop vigorous and the global financial environment is fast changing, and let the market participant gradually value management risk they face. In numerous risk measuring tools, the expected shortfall provides some information about the size of the potential losses given that a loss bigger than Value at Risk (VaR) has occurred, and it makes up insufficiency of the risk measuring tool - VaR. Many papers introduce different methods for estimating the expected shortfall, but they don’t find out the most suitable method that uses in measuring the risk of financial merchandise. The research uses 6 kinds of the expected shortfall and 9 sorts of main stock price index to estimate the expected shortfall and evaluate which method works better and it tries to discover the most appropriate way for estimating the risk of stock price index. The evidence result of the research can detect that all methods for horizons of 1 or 2 days are not because of different coverage rate and change its result. On the other hand, QML GARCH model have best performance for horizons of 1 or 2 days, and it can use in estimating the risk of stock price index. Extreme value theory can be a good model in 1-day horizons, but not for 2-day horizons. Finally the performance of GARCH-Normal model and RiskMetrics approach for horizons of 1 or 2 days are very bad, and they are the most inappropriate methods for estimating the risk of stock price index.[[note]]碩