Forecasting volatility: double averaging and weighted medians

New methods to forecast volatility are usually compared to simple methods like weighted moving averages or GARCH (1, 1) models. In this paper, we provide new benchmark methods which are more accurate but still very simple. In an empirical study of daily returns on major world indices, our new methods clearly outperformed the conventional methods. The superiority of our methods appears to be quite universal as it is not confined to certain markets or certain time periods.GARCH models, weighted medians, exponentially weighted moving averages, EWMA, averaging across windows, squared forecasting errors, absolute forecasting errors, volatility forecasting,