An Application of VaR Method to Forecasting Foreign Exchange Risk:Perspectives of ARCH Models

Liu Jin,Shi Jianhuai
2008-01-01
Abstract:This paper employs several ARCH models with t-distribution residual errors to estimate the foreign exchange risk in China.Based on the empirical study of VaR of USD/RMB exchange rate daily change ratio,we find that(1) the values of VaR estimated by ARCH-like models fit the actual USD/RMB exchange rate data very well,suggesting that there is obvious ARCH effect on USD exchange daily change ratio;(2) the performance of ARCH-like models surpasses the EWMA method used by J.P.Morgan’s Risk Metrics;and(3) among ARCH-like models,TARCH-M(1,1) model is the best one in forecasting the VaR value of USD exchange rate change ratio.The empirical study can be useful for financial institutions and foreign exchange investors to manage their foreign exchange risk.
What problem does this paper attempt to address?