Predicting Realized Volatility: The Effects of Extended Hours Trading

This paper discusses the effects of extended hours trading, i.e. trading that occurs between market close at 16:00 and reopen the next business day at 09:30, on models of daytime realized volatility. Focusing on the impact of data enrichment, rather than model specification, the in-sample and out-of-sample effects are demonstrated in the context of an AR(1) model. The model is estimated using 24-hour high frequency data for Apple Inc., one of the Nasdaq’s most actively traded equities. The in-sample results show that both after hours and pre-market trading contribute significantly to the explanation of daytime realized volatility movements. Moreover, out-of-sample forecasting shows that the inclusion of extended hours realized volatility leads to statistically significant reductions in meansquared error of approximately 28% for a one-day horizon. The results also show that in a hypothetical options market with two traders, one who uses extended hours data and one who does not, the trader who incorporates the extended hours information into his volatility forecast outperforms the trader who does not.