Do shares trail comparable companies after a late-report notice when the measure is checked for false alarms?
Study 60 · Changed a rule or plan · recorded 19 October 2026 · answer: Not settled · full paper with figures
The question
Study 50 found shares trailing comparable companies in the month after a late-report notice, but study 56 found the clipped measure it relied on reports differences that are not there too often. We re-test it on study 50's 1,087 notices by 417 companies, against five companies of the same industry and similar trading size and the same company's ordinary months.
Rules fixed before the result
Fixed before any notice month was read: four measures in order, the plain average, the difference from the same company's ordinary months, the middle value and, held in reserve, a reshuffling test of that difference, each tried 4,000 times on the same companies' ordinary months, where no notice sits. The first whose false-alarm rate stayed within 2.5% to 7.5% decides. A range wholly below zero reads as a shortfall; otherwise the answer says whether a three-point shortfall is excluded.
Data and method
Each company's return over the 21 trading days after its notice was measured from adjusted prices, less the mean of five comparable companies; the regulator's filing indexes for 2014 to 2022 supplied the notices. On the same companies' 7,777 on-time months, against a 5% target with 2.5% to 7.5% allowed, the plain average (6.1%), the difference from ordinary months (6.2%) and the reshuffling test (5.0%) held; the middle value (34.5%) and the clipped average (9.4%) failed.
The answer
Not settled. The plain average decided: its false-alarm rate was 6.1% (range 5.4% to 6.8%), within the 2.5% to 7.5% band set in advance around the 5% target, and it reported a false shortfall 3.3% of the time against the 2.5% a fair test would. On it, shares trailed comparable companies by 2.2 percentage points over the month after a notice, range 4.8 behind to 0.7 ahead, so a shortfall is not established and a shortfall of three points is not ruled out either. The clipped average, the measure study 50 relied on, gave 3.8 points behind (1.9 to 5.7), with a false-alarm rate of 9.4%. Against the same company's ordinary months the difference was 6.4 points behind, range 22.8 behind to 3.2 ahead, so a shortfall beyond those months is not established either.
Main estimate
-0.022 fraction (95% range -0.048 to 0.007), from 1087 observations
What it changes
By the rule set in advance, a correction of study 50's published answer goes to the owner as a proposal: its shortfall is not established on a measure whose false-alarm rate holds, and a three-point shortfall is not ruled out. Study 50's page already carries a note, approved by the owner, that its shortfall is not established without the clipping; nothing on it changes by this study. The plan expected this answer, most likely on the plain average.
What remains uncertain
This re-reads study 50's own sample, so it corrects a measurement within the data already studied and is not a test on new data; chance in that sample cannot be ruled out. The check months are the companies' ordinary months, not notice months, and the ranges treat companies as independent, though companies with overlapping dates or shared comparables share shocks the comparison does not remove. Comparables match industry and trading size only, not financial strain.
How it was made
Run took 102 minutes; independent code review: approve; the recorded calculation reruns to the same result.
Fingerprints
plan c468d718987b08b841e5319e63ca0cc92dfc59d59cec571f4366da73d62b1563 result fa4d78fa6266ebf8ace504e950546faf17ab2cd52a456da3e7184037bd9291c2
Exploratory work on development data; not a performance claim.