The predictive value of temporally disaggregated volatility: evidence from index futures markets

Abstract

This paper examines the benefits to forecasters of decomposing close-to-close return volatility into close-to-open (nighttime) and open-to-close (daytime) return volatility. Specifically, we consider whether close-to-close volatility forecasts based on the former type of (temporally aggregated) data are less accurate than corresponding forecasts based on the latter (temporally disaggregated) data. Results obtained from seven different US index futures markets reveal that significant increases in forecast accuracy are possible when using temporally disaggregated volatility data. This result is primarily driven by the fact that forecasts based on such data can be updated as more information becomes available (e.g., information flow from the preceding close-to-open|nighttime trading session). Finally, we demonstrate that the main findings of this paper are robust to the index futures market considered, the way in which return volatility is constructed, and the method used to assess forecast accuracy. Copyright © 2008 John Wiley & Sons, Ltd.

Asymmetry of cross-correlations between intra-day and overnight volatilities

“…One of the interesting and attractive areas in economy is predictability. The general consensus seems that overnight volatility is useful for predicting subsequent intra-day volatility [16][17][18]. This is an important result.…”

“…Since research in economy is mostly driven by the hope for practical applications, it is not surprising that the majority of the above references were concerned with prediction. The general consensus seems that overnight volatility is useful for predicting subsequent intra-day volatility [12][13][14]. This is an important result.…”

A general property for time aggregation

“…This question has already been thoroughly explored for linear models in financial econometrics. See Taylor [2008], Hafner [2009] and Kole et al [2017] amongst many others. But non-linear models are more difficult to handle and very few attempts to derive aggregation results have been successful.…”