Forecasting Volatility - An Empirical Investigation of Implied Volatility and Its Information Content

Abstract: The purpose of this thesis is to evaluate volatility forecasts by testing the predictive power of implied volatility vis-à-vis model based forecasts. Furthermore we test if implied volatility contains any additional information beyond that captured by the model based forecasts. A number of time series models are fitted to historical data. The fitted models are then used to forecast volatility. The procedure is repeated to produce a series of forecasts. The forecast are evaluated against out-of-sample realized volatility through regression analysis. Finally we test for additional information in implied volatility through GMM and OLS estimation. We find that volatility can be predicted to some extent. Tests indicate that implied volatility is the superior forecast of future realized volatility when compared bilaterally against time series models. Implied volatility does not contain any additional information about future realized volatility in levels when orthogonalized to all model based forecasts. There is however some incremental information regarding changes in future realized volatility.