因應衍生性商品的蓬勃發展,風險管理受到愈來愈多的重視,而在眾多的風險管理工具中,風險值具有量化風險的優點,以金額呈現風險的方式,使風險的表達更為簡單明確,廣為投資機構及金融機構所採用,並將風險值視為衡量市場風險的標準。 衡量風險值模型如雨後春筍般不斷更新,然而目前針對衍生性金融商品並無一最適之風險值估計模型。本研究採用三種風險值估計模型:歷史模擬法、RiskMetrics 模型、GARCH 模型,針對五種股價指數期貨:道瓊工業股價指數期貨、那斯達克股價指數期貨、恆生股價指數期貨、日經225 股價指數期貨、臺灣加權股價指數期貨進行風險值估測,試圖找尋估計股價指數期貨之最適模型。 本研究之實證結果發現: 一、整體而言:以成功率來比較各模型,歷史模擬法較為適合,若以概似比檢定比較,則以RiskMetrics 模型較佳,若以平均誤差百分比絕對值而言,也是以RiskMetrics 模型之MAPE 值最小,因此,對於股價指數期貨風險值的估計,似乎以RiskMetrics 模型較為適合 二、個別而言: (1)道瓊工業股價指數期貨:估計持有資產天數1 日時,選取RiskMetrics 模型;估計持有天數5 日及10 日時,選取GARCH 模型。(2)那斯達克股價指數期貨:估計持有資產天數1 日、5 日及10 日時,均選取RiskMetrics 模型。(3)恆生股價指數期貨:估計持有資產天數1 日時,選取GARCH 模型;估計持有天數5 日及10 日時,選取RiskMetrics 模型。(4)日經225 股價指數期貨:估計持有資產天數1 日及10 日時,選取RiskMetrics 模型;估計持有天數5 日時,選取GARCH 模型。(5)臺灣加權股價指數期貨:估計持有資產天數1 日及5 日時,選取RiskMetrics模型;估計持有天數10 日時,選取GARCH 模型。
Value at Risk (VaR)is an emerging tool of risk management which is designed to meet the trend of various derivatives financial markets. VaR is a better way of quantifying risk. It makes risk simply and clearly to understand by using money. VaR has become the standard measure of market risk employed by investment and financial institutions. There are more and more models for estimating VaR now, but the best model for the VaR of derivatives has not been found. This study uses three VaR models (Historical Simulation Method, RiskMetrics Model, GARCH Model) to estimate the VaR of five stock index futures (DJ, ND, HSI, SSI225, TX). This study tries to find a best VaR model to estimate the VaR of the VaR of stock index futures. The conclusions in this paper are: Historical Simulation Method has the best performance on proportion of successes and RiskMetrics Model has better performance on LR test and MAPE value. So we can say that RiskMetrics Model seems the best model for stock index futures. For different stock index futures: (1) To estimate DJ, RiskMetrics model is the best model for 1 day VaR, and GARCH model is best for 5 days VaR and 10 days VaR. (2) For ND, RiskMetrics is the best model. (3) To estimate HSI, GARCH model is better for 1 day VaR and RiskMetrics model is better for 5 days VaR and 10 days VaR. (4) To estimate 1 day VaR and 10 days VaR of SSI225, it is better to use RiskMetrics model , and it is best for 5 days VaR to use GARCH Model . (5) To estimate 1 day VaR and 5 days VaR of TX, it is better to use RiskMetrics model , and it is best for 10 days VaR to use GARCH Model .