On daily stock volatility predictability

This paper considers 100 stocks from 10 sectors over the period 2000 to 2010 to examine the impact of market conditions, sector type and horizon on the forecasting performance of various volatility models. Out-of-sample forecasts are generated with standard daily GARCH and with ARFIMA/HAR models fi…tted to intraday realised measures. We …find that forecast performance is sensitive to sector type, market conditions and forecasting horizon. ARFIMA generates the best daily volatility forecasts across all sectors for the typical, 1-day ahead forecasting whereas HAR leads for longer horizons from 5 to 22-day ahead. Energy stocks are the most forecastable and the only to show higher forecastability during crisis. Health Care<br/>stocks are the least forecastable irrespective of market conditions.