Idiosyncratic Volatility Forecasting in the Stock Market of Saudi Arabia

We test the forecasting ability of two sets of models, one containing historical volatility-based models and the other conditional volatility-based models, on estimates of idiosyncratic risk of individual Saudi Arabian stocks. While the rankings of forecasts are sensitive to the choice of error statistics, historical volatility-based models appear to be superior, unless the model employed to generate the underlying idiosyncratic return series incorporates higher moments. Exponential smoothing models, with a seasonal component in particular, display superior forecasting performance regardless of whether the idiosyncratic volatility estimates are generated at the local (Saudi Arabian) level or the regional (Gulf Cooperation Council [GCC]) level. The results are of particular interest to investors that are not mean variance optimizers.