Do recording errors in the payout amounts in our price data carry study 77's gap between payout-yield thirds?
Study 80 · Inconclusive · recorded 1 November 2026 · answer: Not settled · full paper with figures
The question
Study 77 found that, on a share's first day without its payout, the gain from prior close to opening, recorded payout added back and net of comparable shares, is 0.1421 percentage points larger for the third of largest payouts relative to price than the smallest. Checking recorded payouts against payers' filings, we ask whether recording errors carry half that gap.
Rules fixed before the result
Fixed before any amount was compared: a random 600 payout days from the highest third and 300 from the lowest, from payers with filings in the securities regulator's public filing archive. The answer reads the difference between the thirds' recording errors, on days whose declared amount is stated for that day, against half of study 77's gap, 0.0710 percentage points, with a tenth, 0.0142, beside. More than three disagreements in 30 hand-checked days would stop it.
Data and method
A filing-reading program, repaired after the first attempt and fixed before the sample was read, found each payout's declared amount per share and its date in the payers' own reports; two independent readers checked 30 sampled days, a third settled their differences, and every large mismatch was read again blind beside 40 other days. The interval for the difference resamples whole securities and allows for the program's own reading errors.
The answer
Not settled. On the payout days we could compare, 235 of 600 sampled in the highest third and 145 of 300 in the lowest, the recording error averaged 0.018 percentage points more in the highest third. Allowing for possible errors in our own reading of the filings, the range for that difference runs from -0.656 to 0.575 percentage points; without that allowance it runs from -0.031 to 0.097. Either way it neither rules out nor shows a carry of half of study 77's gap, 0.0710 percentage points. Within two percent of the declared amount were 224 of 235 compared payouts in the highest third and 142 of 145 in the lowest (0.953 and 0.979, against 0.95 set in advance), but the ranges reach 0.883 and 0.890, under the 0.90 floor, so accuracy too is unsettled; declared and recorded dates matched.
Main estimate
0.018 percentage points (95% range -0.656 to 0.575), from 380 observations
What it changes
Nothing changes in study 77's published reading or in its note on the timing of payout days, since the declared dates matched. The question stays open: settling it at half of the gap needs more compared payout days than this sample held.
What remains uncertain
The answer covers only the compared payout days, under half of each third's sampled days, so recording errors on the rest are unchecked. The recording errors varied more than projected in advance. The years 2018 to 2022 were used before, by study 77 on these same days, so this is a check inside our development data, not an independent test.
How it was made
Run took 117 minutes; independent code review: approve.
Exploratory work on development data; not a performance claim.