Value at Risk of Exchange Rate Return: A comparison among linear and nonlinear evaluation models

• Main Authors: Hou-Cheng Chen, 陳厚丞
• Other Authors: Po-Chin Wu
• Format: Others
• Language: zh-TW
• Published: 2009
• Online Access: http://ndltd.ncl.edu.tw/handle/71265585884054518513

碩士 === 中原大學 === 國際貿易研究所 === 97 === Abstract Recently the Subprime Mortgage Crisis in the United States has caused a series of impact on the international finance. Although the impact is still going on, each government stresses this issue highly. Under this new environment it is more difficult to predict the exchange rate and exchange rate risk. Although some literature has evaluated the VaR of exchange rate and compared their performance, few of them focused on the forecasting of exchange rate. Theoretically the more forecasting performance the exchange rate forecasting model is, the less VaR error it will. This study attempts to find out the more optimal model for predicting exchange rate return, and verifies whether its VaR measure is more better. This paper employs the STAR family models (STAR and STR), advocated by Teräsvirta and Anderson(1992), to test the nonlinearities of monetary fundamentals (MF) model and AR model; forecast exchange rate return, and further to evaluate exchange rate risk by VaR model. Sample period spans from January1999 to November 2008. Sample objects are the exchange rate returns of USD/GBP and USD/JPY and USD/EUR. Empirical study shows that USD/JPY satisfies the logistic STR model in the linearity tests of MF model; USD/GBP and USD/EUR satisfy the logistic STAR model and exponential STAR model in the linearity tests of AR model. ST(A)R models all provide better goodness in fit than linear models. Besides, this study also compares the out-of-sample forecasting performance of linear and ST(A)R models. USD/GBP and USD/JPY are appropriate to adopt nonlinear model in evaluating VaR, but USD/EUR is appropriate to adopt MF model in evaluating VaR.