Stats › Intraday and market microstructure › Intraday periodicity
The half hour that repeats
There's a finding in the academic literature that sounds too neat to be true. A stock that beats its peers between 10:30 and 11:00 this morning tends to beat them again between 10:30 and 11:00 tomorrow morning. Same half hour, next day. And the day after that.
It's a real published result, in the Journal of Finance, with t statistics that most anomalies would envy. I put this page together because it's also a very good case study in what happens to a published anomaly afterwards, and this library already has a page saying published predictors lose 58% of their return after publication. So I wanted to check whether this one did too.
Short answer: it decayed, one independent out of sample study puts the shrinkage at about 75%, and it never survived transaction costs in the first place. The original authors say so themselves. All the numbers and links are at the bottom.
Heston, Korajczyk and Sadka split each NYSE trading day into 13 half hour intervals and found return continuation at lags that are exact multiples of 13, meaning exactly one trading day. Over the first week the smallest t statistic at those daily lags was 9.62. A top minus bottom decile trade earned 3.01 basis points in the matching half hour the next day, and the pattern lasted at least 40 trading days. The catch is in the internet appendix: the same trade goes from +3.03 bp before costs to -23.78 bp after bid-ask costs. An out of sample study covering Aug 2021 to May 2023 found the US pattern had weakened by about 75% and now shows up mainly in the last half hour of the day.
Members only
The rest of this page is for members
Below this point there are 8 sections, 1 chart, 1 table and 6 named sources, roughly 2050 words of it. Every figure carries the source it came from and the date the data is from.
You can keep browsing every statistic in the library for free. The intro and the summary are always open.