P. Sadorsky (Canada)
Forecasting volatility, GARCH, stochastic volatility, value at risk
This paper compares the forecasting performance of several well known volatility forecasting models. Each forecasting model is applied to a financial data set that includes the S&P 500, ten year U.S. government bond series, crude oil prices, and the foreign currency exchange rate between the Canadian and U.S. dollar. Forecasts are evaluated using MSE, MAD, U, DM, regression test, LINEX, and Value at Risk (VaR). Overall these forecast summary statistics show that for each financial series, the GARCH models and TARCH model in particular to be better than the stochastic volatility model (SV) in almost every category except MAD. Value at risk calculated from the SV model does not reject independence in three of the four financial series studied but does reject unconditional coverage in all of the series studied. All of the parametric models reject conditional coverage. The conditional empirical density model does not reject conditional coverage in three of the four series studied. In general, the conditional empirical density model is preferred to many other VaR models.
Important Links:
Go Back
