金融商品投資風險評估之研究 -以VaR模型之歷史模擬法為主-

Study of Financial Products Investment Risk Assessment-Using VaR Model-based Historical Simulation Method

張貿易(Mao-I Chang)

指導教授 ： 簡俱揚 ； 施能仁

中原大學/商學院/會計研究所/碩士(2003年)

https://doi.org/10.6840/cycu200300278

摘要

中文摘要
金融商品交易日益遽增，眾多國內外企業皆因未作好投資之風險管理，導致企業面臨破產或重整，而風險值觀念興起，使得金融商品之市場風險管理進入新的里程碑。國內在相關風險值之揭露規範，僅於財務會計準則第二十七號公報中提及建議揭露。本研究期從企業角度幫助國內金融業如何選定適當風險值模型，依選定之模型估計企業之風險值，至分析風險值之含意，最後檢定風險值模型是否適用，提供金融業對風險值之運用有一整體性之串連，協助金融業作內部控制及評估其投資風險。另一方面則可以幫助市場之參與者，瞭解及實際運用風險值之資訊，使量化之風險值具決策之有用性。最後期藉此補充國內第27號公報之相關規定及幫助國內金融監理機構制訂相關準則之參考。

本研究針對國內上市銀行業及上櫃證券業作實證分析，依參考國內外文獻之研究，選定歷史模擬法為估計風險值之模型，計算及分析量化之風險值資訊，再以回溯測試、超限比率、超限比率值及Kupiec（1995）提出之LR 檢定法，檢定歷史模擬法是否適用於評估國內金融業之風險值。實證結果顯示出歷史模擬法，不論在銀行類股或證券類股中，依四種風險值模型檢定方法下，歷史模擬法對大部分兩類股而言，具有一定估計之效率及精確性。

（一）依回溯測試而言，在99％及95％信賴水準下，民國89年度兩類股整體之超限次數有偏高情形，但民國90年度次數明顯比較少，表示民國89年之極端事件對估計民國90年度之風險值產生影響，顯示出移動窗口選擇上之重要性。

（二）依超限比率值而言，歷史模擬法對大部分銀行業及證券業兩類股之估計效率仍佳，且歷史模擬法在99％信賴水準下，其估計風險之效率最佳。

（三）依超限距離值而言，比較89年及90年之兩類股風險值，顯示出歷史模擬法在民國90年度較民國89年度之估計佳。

（四）依LR 檢定而言，不論在99％信賴水準及95％信賴水準下，此法估計兩產業之風險值均具精確性，但其檢定值偏低，表示歷史模擬法在估計風險值上，顯趨於保守。

關鍵字

風險值 ； 歷史模擬法 ； 回溯測試

並列摘要

Dissertation Summary

Many enterprises at home or overseas either face bankruptcy or reorganization due to failure in practicing sufficient risk control in a time seeing drastic increase of transaction of financial products. Emergence of the concept of VaR, the risk value, has brought the risk control for the market of financial products into a new milestone. Domestic laws subject “disclosure” involved in VaR only to that as proposed in No. 27 Publication of Financial Accounting Standards. On one hand, this study attempts entering from the angle of the enterprise to help domestic financial industry select the proper VaR model to predict the enterprise specific VaR using the selected VaR model, analyze VaR significance, and determine whether the VaR model is appropriate or not; present an overall series of the application of VaR for the financial industry; and assist the financial industry in internal control and assessment of investment risks. On the other hand, this study seeks to help those involved in the market on understanding and practical application of information about VaR in making the quantitative VaR useful for decision making. Finally, this study also serves the purposes of making supplements to those requirements set forth in No.27 Publication and reference by domestic financial control authorities in preparing related standards.

Following the pragmatic analysis on the listed bankers and OTC securities in Taiwan and documentary search, the Historical Simulation Method is chosen as the modal for estimating the VaR to solve and analysis quantitative VaR information before being tested for its feasibility to assess the VaR of domestic financial industry with Back Testing, Uncovered Loss Ratio, Uncovered Loss Distance and LRuc Test proposed by Kupiec (1995). Results indicate that either in the banking stocks or securities stocks the Historical Simulation method gives a certain estimate efficiency and accuracy for most of the banking stocks and securities stocks when tested with four VaR models:

(1)Regarding the Back Testing method, the Uncovered Loss number of both types of stocks is comparatively higher for FY 2000 given with 99% and 95% Confidence Level respectively; however, the Uncovered Loss number significantly drops for FY 2001, showing that the VaR of FY 2001 has been affected by those Extreme Events taking place in FY 2000 and protruding the importance in the selection of the Moving Window.

(2)As for the Uncovered Loss Ratio, the Historical Simulation method maintains good efficiency in the estimate of VaR in both types of banking and securities stocks, and the best efficiency is observed in the risk estimation for the Historical Simulation method given with a Confidence Level of 99%.

(3)In terms of Uncovered Loss Distance, the better estimation efficiency for the Historical Simulation method is found with FY 2001 than FY 2000 by comparing between the risk values of both types of stocks in FY 2001 and FY 2000.

(4)Whether with a 99% or 95% Confidence Level, the LRuc Test method proves its accuracy for both banking and securities industries; however, the test value is comparatively lower to suggest that the Historical Simulation method significantly tends to be conservative in estimating the VaR.

Key words: VaR, historical simulation method, back testing

並列關鍵字

VaR ； historical simulation method ； back testing

參考文獻

財務會計準則公報第27號-金融商品之揭露，財團法人中華民國會計研究發展基金會財務會計準則委員會，民國86年6月。

Taiwan』，working paper。

Alexander,c.o.,and Leigh,c.t.（1997）,“On the Covariance Matrices Used in Value at Risk Model”,The Journal of Derivatives ,p.50-62。

Beder,Tanya Styblo（1995）,” VAR : Seductive but Dangerous”,Financial Analysts Journal vol.51 Sep/Oct ,pp12-24.。

Christoffersen,P.F.（1998）,”Evaluating Interval Forecasts”,International Economic Review, 。