Can the earlier priced-prospectus study's one-year test be made able to detect the effect size set in advance?
Study 46 · Changed a rule or plan · recorded 10 October 2026 · answer: Yes · full paper with figures
Rules fixed before the result
Fixed before any of these figures was computed: the test counts as precise enough when it can detect a one-year effect of 10 percentage points. Price faults would be the answer if at least a quarter of the 16 most extreme price series failed the check; otherwise the first summary, in a fixed order, to reach 10 points is named, and failing that the same companies' ordinary years decide, needing at least 500 of them from 100 companies.
Data and method
We rebuilt the earlier study's 2,190 sales by 745 companies exactly and split each one-year difference into the company's return and the mean of its five comparables' returns. We ran a check of each extreme price series for a one-day adjustment break, a spike that reverts, a dead quote carried forward, a long gap or a sub-penny price, tried four summaries that limit extremes, and compared the same companies' ordinary years, with no sale within about two years before.
The answer
Yes. The 16 most extreme differences came mostly from the selling company's own rise (14 of 16), and only 1 of them failed the price-series check, against none of the 44 least extreme. Comparing the issuer's and the comparables' year on a logarithmic scale, so that a tenfold rise does not dominate, cuts the smallest detectable effect from 19 to about 5.8 percentage points, inside the 10 set in advance; pulling the most extreme 1% at each end in to the 1% value gives 7.4, while the comparables' median in place of their mean leaves 18.5. The same companies' ordinary years were about half as spread. The plan expected this answer with low confidence, but expected ordinary years to be as wide.
What it changes
By the rule fixed in advance, the earlier study's one-year question can be planned again on the same sales with years compared on a logarithmic scale, which measures a difference in growth rather than in returns. Because these companies' ordinary years were narrower, later plans will also take the spread for selling companies from their ordinary years.
What remains uncertain
These sales were studied before, in study forty-five, and this plan was fixed after their spread was seen. Only 16 sales are extreme, the price check's limits were set by judgement, and companies selling on the same dates are not treated as dependent beyond an assumed correlation among one company's sales. Nothing after a sale was measured, so no effect is shown either way.
How it was made
Run took 72 minutes; independent code review: approve; the recorded calculation reruns to the same result.
Fingerprints
plan 1762cc3c77481a84bb86cb73d92864933d05e3af076e5fcb1b8e7ce01d5135d1 result d626a015a23c22f3117e86a0db0292db2df9dd93baa7164afb159195443c8c26
Exploratory work on development data; not a performance claim.