Long-Term Volatility Forecasting

A variety of historical-volatility, peer-historical-volatility, implied-volatility and blended estimators of stock price volatility are developed and tested for a group of large U.S. companies over roughly a thirty-year window. Longer-term historical estimators (up to fifteen years) are found to outperform shorter-term estimators as forecasts of five- to seven-year realized volatility. Inclusion of implied volatility into forecasts at low weightings is found to have little discernible effect on overall results; at higher weightings, implied volatility appears actually to detract modestly from forecast accuracy. Nevertheless, certain correlations show that implied volatility may contribute strongly toward forecasting volatility in some situations. Finally, patterns of apparently-cyclical variation in historical forecast-errors are presented for exploration and inclusion in potential future modeling