Nine Per Cent, Plus Or Minus Eleven. The Insurers Renewed Anyway.
Two statisticians who have never met were paid to test thirty-nine seasons of hail suppression over the same prairie. They agree on the number. The consortium's own actuaries reached it in 2011 and filed it under vehicle routing.
A months-long reporting project. Documents cited below are held in The Vault and available to readers.
OLDS, Canada — Nine per cent. That is the reduction in hail kinetic energy over 26,000 square kilometres of southern Alberta, measured across thirty-nine seasons of flying. The confidence interval runs from minus two to plus twenty.
Minus two means the seeded ground may have been hit slightly harder than the ground next door. Plus twenty means the aircraft may be doing everything the brochure claims. Both of those are inside our number.
The Chinook Hail Project has flown silver-iodide burners into the same storms every summer since 1988. No government pays for it. Eleven crop and property insurers pool the cost between them under the name of the Prairie Hail Pool, and they have renewed the contract eight times since 2011.
This newspaper wanted to know what they think they are buying.
How we got the file
One of the eleven is a mutual, owned by its policyholders.
In February a member of that mutual — acting on her own account, months before she had heard of us — used a statutory right to inspect the pooled-expense and loss-experience schedules filed on her behalf. She was sent 1,140 pages. She gave them to this newspaper, asked for nothing, and does not want her name printed.
The project supplied the rest without argument: thirty-nine seasons of C-band volume scans, the flight logs, and the readings from 214 hailpad sites, which are squares of dense foam that keep a record of whatever dented them.
Two strangers, the same question
We paid two statisticians who have never met to do one job separately. Compare hail kinetic energy inside the seeded target area against the unseeded control corridor to its east, season by season, adjusted for storm count and track.
Dr Ingrid Salcedo-Bramwell, in Winnipeg, returned 9.4 per cent, interval minus 1.8 to plus 20.6. Dr Renata Oyibo-Hallgren, in Uppsala, returned 8.1 per cent, interval minus 3.0 to plus 19.4. They disagree about four seasons in which the control corridor took the worse storms, and they disagree at length, in reports we publish entire.
Nine per cent, plus or minus eleven.
This is not an outlier. Our own table of thirty-one randomised seeding trials, PX-1986, has the same shape wherever you look: wide intervals with zero inside them. That is what an honest measurement of weather modification looks like, and the interval is the column that gets dropped everywhere else.
They already had the number
Here is the part we did not expect.
In August 2011 the Pool's own pricing actuary ran the comparison herself. Fewer seasons, a coarser pad network, a different adjustment for storm track: 7 per cent, interval minus 4 to plus 19.
Materially our number, fifteen years early.
It was filed. It went to the renewal committee that October. The committee renewed, and has renewed every two years since: eight renewal papers, eight votes, and not one dissenting vote recorded in any minute we were given.
Section 6, vehicle routing
The 2011 review runs to fifty-one pages. Section 6 is headed "Vehicle routing and adjuster deployment", and one sentence inside it accounts for the whole of the renewal.
The radar and the crews' hail reports advance first notice of loss by approximately two hours, which on its own recovers the programme fee through deployment savings; the suppression effect is not required for the renewal to price.
The aircraft fly into the storm and radio what they see. The C-band scans reach the claims desk in near real time. All of it arrives, at a median of 117 minutes, before the first policyholder telephones.
Adjusters who are already moving reach the farm sooner. Inside the radar footprint, files close in a median of eleven days; outside it, nineteen. Claims are reopened at 6.1 per cent inside and 14.8 per cent outside. Flattened canola gets appraised while it is still standing, before rain puts it under water and it stops being appraisable at all.
That is the saving. It is not hail that never fell.
What we cannot settle
We have not shown that the seeding does nothing. Nine per cent is not zero, and thirty-nine seasons over one target area is the largest sample anybody has assembled for hail anywhere on earth.
We have also not shown that it works.
And there is a harder problem, which probably belongs higher in this story than we have put it. The radar could be run without the aeroplanes. Nobody has tried it. No member proposed it in any minute we read. Part of the two-hour advantage is a crew member's voice saying the stones are the size of a thumbnail, and no unattended scan says that.
The Pool may be paying C$3.4m a season for a dispatch feed. It may equally be paying C$3.4m a season for a dispatch feed that only exists because somebody is up there flying into the hail.
The six o'clock edition
Three stories a morning, in your inbox before the coffee is made.
Sources & Method
We did not set out to test the seeding. We set out to read the expense side, and paid two statisticians who have never met to test the seeding separately so that we would not be the ones deciding what the effect was. Both were paid the same fee in advance, both were told their reports would run whether they agreed or not, and neither saw the consortium's own 2011 number until after she had filed.
Who we spoke to
- Prairie Hail Pool — pooled-expense and loss-experience schedules, 1994–2025, 1,140 pages, including the 2011 loss-and-expense review and eight renewal papers. Obtained from a policyholder of the member mutual, who inspected them under a statutory right, on her own initiative, and gave them to us unconditionally February 2026 She declines to be named and was paid nothing. We have verified her membership and the date stamp on the disclosure letter.
- Dr. Ingrid Salcedo-Bramwell, Statistician, Winnipeg; extreme-value methods, insurance loss data. Commissioned to reconstruct the seeded–unseeded comparison from the raw hailpad and radar archive, with no sight of the expense files March–July 2026 Returns 9.4 per cent, interval −1.8 to +20.6. Believes the control corridor is defensible. Report published entire.
- Dr. Renata Oyibo-Hallgren, Statistician, Uppsala; second reconstruction, blind. Given the identical archive with no contact with Salcedo-Bramwell and no knowledge of who else had been hired March–July 2026 Returns 8.1 per cent, interval −3.0 to +19.4. Thinks the control corridor is contaminated in four seasons and argues it for nine pages.
- The Chinook Hail Project, Radar archive, hailpad network and flight logs, 1988–2026. Released in full after a written request; the director, Emmanuel Thorsby-Aduba, sat with our correspondent for two days while we copied 39 seasons of volume scans April 2026 214 hailpad sites. The project asked for no conditions and imposed none.
- Solveig Amundsen-Baptiste, Pricing actuary to the Prairie Hail Pool, 2004–2016; retired. Wrote the 2011 review. Sent the full draft, both reconstructions and a photograph of her own section 6, and given five weeks July–August 2026 Her reply is printed unedited at the foot of this page and is the most serious objection to this story we received.
Documents
- PX-1986 — Effect sizes and confidence intervals from 31 randomised cloud-seeding trials, 1961–2024 accepted
What we could not confirm
- Whether the target area and the control corridor are comparable at all. Hail climatology can change materially over sixty kilometres of prairie. Our two statisticians adjusted for it by different methods, disagree about four seasons, and neither can demonstrate that her adjustment is the right one. If the corridor is systematically hailier, the nine per cent is partly geography.
- Whether nine per cent is an effect. The interval includes zero and it also includes twice what the operator claims. We have not shown that seeding works and we have not shown that it does not, and thirty-nine seasons over one target area is the largest sample assembled anywhere for hail.
- Whether the deployment saving would survive without the aeroplanes. The radar could be run on its own. Nobody has tried it, no member proposed it in any minute we were given, and part of the two-hour advantage is a crew member's voice describing stone size, which no unattended scan supplies. Our deflation may depend on the thing it deflates.
How Others Covered This
The same events, as reported elsewhere on the same day. We list what each outlet had that we did not, as well as what we had that they did not — including where we come off worse. Why we print this.
-
The Meridian TelegraphINSURERS SPENT $130M SEEDING CLOUDS THEY KNEW DID NOTHING
Read an interval that crosses zero as a proof of fraud, and the eight renewals as eight cover-ups.
Had that we did not
The interval itself, quoted accurately from our figure, which is more than most will manage.
Left out
That an interval crossing zero is not a finding of nothing, and that the same file explains the renewal in one sentence anybody may read.
-
Signal & LedgerFirst Notice Of Loss: What Eleven Insurers Bought When They Bought The Weather
Took it as an operations lesson for claims executives and wrote the deployment numbers up better than we did.
Had that we did not
The eleven-day close, the reopen rate, and the canola window, laid out cleanly for people who run adjusting panels.
Left out
The flying. Thirty-nine seasons of crews going into hail appear in their piece as a data-acquisition strategy.
-
They Buriedthis newspaperNine Per Cent, Plus Or Minus Eleven. The Insurers Renewed Anyway.
Obtained the loss files through a policyholder, bought two blind reconstructions of the effect, and then read the expense side.
Had that we did not
Both reconstructions entire, including where they disagree, and the 2011 memo photographed whole rather than quoted.
Left out
Our headline puts the interval in the subordinate clause and the renewal in the main one, which invites a reader to think eleven companies were being cynical. Nothing in 1,140 pages shows cynicism. It shows an organisation that went on paying for a measurement it had already taken, because a by-product turned out to be useful, and we could not find one document in which anybody decided that. — V. Ashcombe-Doyle, standards editor
Right of Reply
They Buried contacted Solveig Amundsen-Baptiste, pricing actuary to the Prairie Hail Pool, 2004–2016 on 3 July 2026 in writing, with the full draft, both commissioned reconstructions, and a photograph of the page carrying her own section 6; no deadline was set. Replied 11 August 2026 and asked that it run entire. Printed unedited, including the parts about us.
You have taken one sentence out of fifty-one pages, put a heading on top of it, and made the heading your headline. I want to be careful here, because I do not think you have been dishonest. I think you have been literary.
A review has headings because a review has sections. Section 6 was about vehicle routing because section 6 was about vehicle routing. It is not a hiding place. Sections 2 through 5 are about the suppression effect, they run to thirty pages, and you have not quoted a line of them.
Now the substance, which is where I think you are actually wrong.
You write that the renewal was not a decision to buy hail suppression but a decision to buy a radar feed. There was no such decision available to make. The operator sold, and still sells, one product: aircraft, crews, generators, radar, forecasters. I could not have bought the feed alone in 2011 and my successor could not buy it alone this morning. You have discovered that a bundled contract had a valuable component, and you have written it up as though a committee sat in a room and chose the component. It did not. It renewed the contract, and the contract came with everything in it.
On my number. Seven per cent, minus four to plus nineteen. That is very close to what your two statisticians got with fifteen more seasons and a better pad network, and I am glad of it, because it suggests I was not careless. But I never wrote that seven per cent was nothing. I wrote that the suppression effect was not required for the renewal to price. Those are two different sentences and you have printed the second in a way that retracts the first. An interval that includes zero also includes nineteen. Nineteen per cent less hail energy over 26,000 square kilometres is an enormous quantity of unbroken glass and standing crop, and I priced a contract in the knowledge that we might be getting it and might not. That is not cynicism. That is what it is to buy under uncertainty, which is the entire trade I spent thirty years in.
And here is my real complaint. You have written a story in which eleven companies save money, an actuary writes a memo, and a newspaper is clever. Nobody in it lives there. Two hundred and fourteen hailpads is two hundred and fourteen points across an area larger than Wales, and underneath them are people who have watched those aircraft go up every July since they were children and who have opinions about whether it does anything. Some of them are furious about it. Some of them plan around it. You did not ask one of them. Your figure has canola in it as a line item and no farmer anywhere in the piece.
One thing I will give you, and I have thought about it since your letter arrived, and I do not have a good answer. You are right that nobody ever wrote the decision down. I looked for it too, after you wrote. There is no minute in which anyone says we are keeping this for the dispatch. There is a renewal paper, and a vote, and a renewal paper, and a vote, eight times. Committees are very good at continuing to do a thing, and very bad at recording why. I would rather that were not true of the committee I advised, but it is.
Published unedited under our right-of-reply guarantee.
How was this story?
We publish the result, whatever it is. Reader verdicts appear on the front page and in our newsroom metrics.
Readers' Letters 0
Printed at once under the name you give and read by the desk afterwards; anything unfit is removed, with a note saying so, and nothing else is ever deleted — only corrected. Letters that changed something in the story carry a mark saying so, and there are 5 of those across the archive.