FORECASTING VOLATILITY:: EVIDENCE FROM SWEDISH STOCK MARKET

Abstract: This study evaluates the performance of alternative models for predicting stock price volatility on Swedish market. The model set contains various methods for producing volatility forecasts ranging from simple ones (Random Walk, Moving Average and Exponentially Moving Average) to non-linear group of models (GARCH (1,1) and EGARCH) and Implied volatility from OMX S30 option prices (IV). Overall model performance is evaluated using RMSE and MAE measures. The main results are the following: (1) Forecasts based in implied volatility produce the most accurate results under both measures, while GARCH (1,1) model gain the highest overall error statistics. (2) Allowing asymmetry in variance and non-normal error distribution, the EGARCH (1,1)-GED models perform much better than GARCH (1,1), especially for 20- and 40-day forecasts. (3) Further tests applied to IV confirm that indeed it is an unbiased and efficient estimate of future volatility. The results suggest that the findings in the OMX S30 are mainly in line with the majority of recent evidence, although the study can be extended by the inclusion of other models.