The Research on the Estimation of Exchange Rate Risk by Using the Stochastic Volatility Model

• Main Authors: Shi-Zhe Lin, 林世哲
• Other Authors: Ai-Chi Hsu
• Format: Others
• Language: zh-TW
• Published: 2009
• Online Access: http://ndltd.ncl.edu.tw/handle/19373113111382383878

碩士 === 雲林科技大學 === 財務金融系碩士班 === 97 === Taiwan economy depends heavily on international trade, and hence exchange rate fluctuations can affect economic developments. A series of financial turmoil has led foreign exchange market to take on am unprecedented level of fluctuations and left business exposed to tremendous amount of exchange rate risk. This research takes New Taiwan Dollar to US Dollar as our empirical data. We consider a computational Bayesian framework for full parametric forecasting of Value-at-Risk (VaR) thresholds from the stochastic volatility model. Adaptive and efficient Bayesian Markov chain Monte Carlo（MCMC）sampling methods are designed for in-sample estimation and need to construct Bayesian estimators of the general l-period forecast VaR. In order that stochastic volatility model better describe the leptokurtic of asset returns and persistence of volatility, model is expanded to two parts. Besides, we also use variance-covariance approach, generalized autoregressive conditional heteroscedasticity, historical simulation method and Monte Carlo simulation method. In this study, we use three likelihood ratio tests to examine the accuracy the VaR calculated by the models mentioned above. The VaR models’ performance assessments are based on a range of measures that address the conservatism, accuracy and efficiency under 1% and 5% confidence level. From the empirical results we show that stochastic volatility model based on MCMC methods give more robust VaR forecasting estimates than others.