Correction, 30 September 2026. Technical wording rewritten in shorter sentences under the readability rule; the meaning is unchanged.
Correction, 19 October 2026. This text was reviewed after the study finished, in a catch-up of the earliest studies; the result, outcome and dates are unchanged.
The question
An earlier study found, on our development data, that five-session returns after such splits ran 1.75 percentage points below the same shares' ordinary sessions. We read that gap with the test our planning rules chose since, which judges it against reshuffled ordinary sessions of the same companies, and ask whether timing or a small group of companies explains it.
Rules fixed before the result
The rules for each possible answer were written down and fixed before this study computed anything. The earlier finding on these same split events was already known: it was the result being read again.
Data and method
For each of 128 splits, at 94 companies, we compared the five-session return with the average of the same share's ordinary sessions, and read the gap against a test that puts reshuffled ordinary sessions in the splits' place. For timing, we compared each split with the same share's sessions 64 to 252 sessions away. For spread, we planted the earlier gap on all the companies or on smaller groups, down to 6, and asked which pattern the gap matched.
The answer
Yes on development data only: the stricter test still finds the gap and timing does not explain it, though whether few or many companies carry it is unsettled.
What it changes
The earlier finding, that returns in the five sessions after a split that adds shares trail the same shares' ordinary sessions, still stands on our development data only. Only new splits from outside that data can settle it.
What remains uncertain
Whether a small group of companies or many carry the gap is still unknown. These are the same split events the earlier study used, and returns of companies on the same dates move together in a way the test does not fully treat. Only the five-session window was tested, on one development sample from 2013 to 2022.