On the Specification of Duration Between Price Changes and the Predictability of High Frequency returns: an application to the French CAC 40

This paper focuses on the predictability of the duration between intra-day price changes of stocks from the CAC 40, as well as on the predictability of the returns generated by these price changes. It is argued that traders with different time horizons will look at series of price changes recorded at different time intervals. Small price variations will be of higher importance to short term traders than to longer term traders. We generate various time series to reflect these different time horizons. The duration between these price changes is specified by a model inspired by the Autoregressive Conditional Duration (ACD) model of Engle and Russell (1997). We also introduce threshold versions for the duration and return specifications, where the estimates depend on a state-of-the-world defined by thresholds. We find that the expected conditional durations are highly significant in predicting future returns, even when a three months’ out-of-sample period is used. Finally, a realistic trading strategy taking into account transaction costs is tested out-of-sample and found profitable in some cases.