The Corrections Log.
Show the working · Track the promises · Say when I’m wrong
Errors get logged here: dated, not quietly edited away. That is what “Say when I’m wrong” means on this site — hold every candidate to it.
What earns an entry: any factual error in a published claim, on this site, on the Observatory, or in campaign materials — and any change of position, so you can see what moved and why. Typos don’t get entries. Numbers, claims, verdicts and positions do. An entry that changes a conclusion says so.
Corrected 1 September 2026 · Braddan Woods: 80 affordable homes, not 82
- Error: the first online version of the manifesto said “only 82 of 320 homes” at Braddan Woods are affordable. 82 was the figure in the original July 2022 application, for 328 homes.
- Correction: the scheme as approved (August 2023; appeals dismissed July 2024) and as branded Braddan Woods in June 2026 is 320 homes with 80 affordable — one in four. Corrected online and in the printed booklet before it went to press.
Position changed 1 September 2026 · Braddan Woods
- What I said before: “I back Braddan Woods, and I’ll publish the exact wording of every commitment attached to it.”
- What I say now: I’m concerned by how little of it is affordable — 80 of 320 homes, one in four, when affordability is the whole problem — and by the infrastructure that isn’t there: the access, the school that was only ever “approval in principle”, and the Peel Road crossing. Right-size what we build, and track every promise in public.
- Why it changed: the doorstep, and reading the approved scheme rather than the headline. Not an error of fact, but a change of mind should be visible, so it goes here.
Withdrawn 1 September 2026 · “59 known actions, only 9 completed”
- What I said: that the ten health reviews produced 59 known actions of which 9 were complete (“15% done”), on this site and in the interim manifesto.
- Why it’s withdrawn: the count was produced with AI assistance from the published reviews. On checking, the way actions were counted and marked complete isn’t robust enough for me to stand behind a precise number. The Observatory’s tracker stays up with its method shown, but the headline figure comes off this site and out of print.
- What replaces it: “Ten reviews in seven years, and barely any progress reported” — which the published record supports.
Corrected 17 August 2026 · Jersey data blended into an Observatory narrative by Fable
- Error: an Observatory scrollytelling piece (Act 6 of the Isle of Man economy system dynamics model, iom_economy_story.html) went live with Jersey statistics pulled in and blended with genuine Isle of Man dissertation data while Fable (an AI model) built the narrative. Asked others to check the logic and numbers before publishing; it sat live for about two weeks.
- Correction: the piece rewritten and re-checked line by line against source data. The Act 6 system dynamics model re-validated and republished as an interactive tool (build rates, house prices, etc.)
- Process fix: AI is no longer asked to build the story itself — narrative construction stays with me; AI use limited to tasks that don’t risk mixing jurisdictions or inventing connective tissue between real figures.
- Verdict: no reviewer or reader flagged the blended data during the two weeks it was live; caught by me, not by review.
Corrected 14 August 2026 · Two data errors, found in my own files
- Error: a Statistics Jersey real-output dataset was filed as Isle of Man data (jurisdiction prefix stripped). Fed the home-page growth chart, the Seven predictions scorecard, an Economy sector figure, and an Observatory analysis.
- Correction: on the Isle of Man’s own constant-price accounts, real output rose strongly to a 2019/20 peak and has fallen in every year but one since. Every affected chart and figure rebuilt or removed.
- Process fix: every data file now requires a catalogue entry naming publisher, source, and jurisdiction. Uncatalogued files are quarantined by default.
- Error: electricity comparison (£994 Manx vs £760 UK, sourced to smartisland.im) mixed a standing-charge-inclusive figure against a units-only one, pre-VAT against post-VAT rates, and an outdated consumption benchmark.
- Correction: like-for-like, VAT-inclusive: £872 (IoM) vs £861 (UK) — a 1.2% difference. The comparison has been removed from the site and replaced with the inflation chart, which does hold up.
- Also corrected: deaths have exceeded births since 2016, not 2017; the Steam Packet’s 25% rise applied to the freight fuel surcharge, not freight costs overall; reserves are forecast to grow, not run low — each year’s deficit is closed by drawing on them.
Corrected July–August 2026 · One misfiled Guernsey workbook, two corrections
- Error: a States of Guernsey census workbook, misfiled in the Isle of Man archive, contaminated the census employment series used in two pieces of work.
- Affected: the seven-predictions scorecard issued 18 July (corrected 22 July — true Isle of Man figures: employment rose 2001–21, flat 2011–21); an early version of an Observatory piece, withdrawn within a day and republished once corrected.
- Correction: affected figures corrected everywhere the series was used; every model result built on the bad series withdrawn pending re-estimation; every remaining figure in both pieces re-verified against primary documents.
- Verdict: no conclusion flipped. Quantified fit results stayed withdrawn until re-estimation cleared.
- Addendum, 17 August 2026: the Observatory piece referenced above was superseded by the 14 August correction (above) and retitled The Island That Stopped Growing. Fit results withheld pending re-estimation were restored 17 August 2026, run against the corrected, independently re-verified series.
An empty corrections log should worry you more than a full one. About 1,300 source files sit behind the Observatory and this campaign, mostly extracted from Government PDFs — error-prone by nature. Found one? Tell me, or email james@cochrane.im. If you’re right, it goes here with a date on it.