A New Value at Risk Estimation under the phenomenon of volatility clustering and heteroscedastic volatility.

• Main Authors: Hsieh-Heng HUNG, 洪榭亨
• Other Authors: Shu-Hui Yu
• Format: Others
• Language: zh-TW
• Published: 2005
• Online Access: http://ndltd.ncl.edu.tw/handle/66715039150405369290

碩士 === 國立暨南國際大學 === 財務金融學系 === 93 === How to develop a method for measuring and managing the market risk effectively has been discussed a lot since risk management became an important issue in financial management. According to the regulation of Basel Committee on Banking Supervision in 1996, the financial institutions can evaluate the value at risk in accordance with their model built internally to calculate the market risk capital. Thus many big financial institutions look for their own suitable evaluation models. Among them, the most important landmark method is the concept about the value at risk. However, it is found in this study that if we rely on the value at risk calculated through some existing models great loss or too high risk-based capital adequacy ratio are inclined to happen very possibly. Though how important the existence of the extreme value to the calculation of the risk value is has been emphasized in relevant literatures, it is found often in actual financial data that its regression variance is not fixed and will change with time, which thus causes the effect of volatility clustering. Therefore, this article is planned to submit a new method for estimating the value at risk since those old methods submitted in literatures still need to be improved. The five methods mentioned in literatures, such as Equally Weighted Moving Average, Exponentially Weighted Moving Average, Historical Simulation Method, Jump-diffusion Process, and Extreme Value Theory, are combined considering the volatility clustering frequently found in the financial assets to calculate the value at risk. The method submitted in this study for evaluating the weighed value at risk has various advantages of the above-mentioned models facing different market situations. And through the weighed values given by the models for volatility (GARCH and ARCH models), this method decides what kind of value at risk should be measured when the market faces different situations. The estimation for the weighed value at risk has changed some limits of old models, e.g., limit on the independence of the hypothesis events under the Generalised Pareto Distribution and the Jump-diffusion Process, or the problem that the occurrence of extreme values under the normal distribution is neglected. In this research, it is found that when the volatility is bigger the model for the weighed value at risk performs better apparently, considering the error frequency and capital efficiency. So this kind of method offers better evaluation for the value at risk when there occurs volatility clustering and heteroscedastic volatility in the financial data. Or it offers an estimation better than the existing ones that evaluating the value at risk, even though there is no heteroscedastic volatility in the financial data when the financial market faces the event of enormous crisis.