Sightings Fell 61 Per Cent The Month The Form Started Asking For A Name
Twelve years, 14,208 filings mentioning unidentified traffic, normalised against published flight hours. The two most-cited waves in the modern record sit exactly on top of revisions to the paperwork, and one of them is a change of address.
Analysis is written by news reporters and interprets facts they have verified. It is not opinion, and it takes no position on what should be done.
COLLEGE PARK, Md. — On 1 November 2016 a national aviation hazard-reporting system issued revision 4 of its filing form. A confidentiality checkbox came off. A mandatory field for the reporter's name and certificate number went on.
In the month that followed, filings mentioning traffic the reporter could not identify fell 61 per cent.
Nothing else on the form changed. Published flight hours for that month were within two per cent of the same month a year earlier.
What we counted
Twelve years of the system's public filing database, 2013 to 2025 inclusive. 14,208 filings whose narrative field mentions unidentified traffic.
A filing counted if the narrative contained one of eleven phrases. That list was fixed and timestamped in March 2026, before any filing was matched to any date, and it is published with this story so that anybody who thinks we picked the phrases to get this answer can pick their own and tell us we were wrong.
Raw counts are worthless here, because the number of aeroplanes in the air is not constant. Everything below is filings per 100,000 published flight hours.
Four dates, found blind
We gave the normalised series to Dr. Ottoline Marchetti-Baye, a survey methodologist, with the dates replaced by sequence numbers and no indication of what was being counted. Her instruction was to find the points where the level shifts and does not shift back.
She returned four.
We then looked up what happened in those four months. We had not looked before. All four are administrative.
November 2016. Confidentiality checkbox replaced by a mandatory reporter identifier. Rate falls from 11.4 to 4.4.
March 2019. Filings stop going to eleven regional offices and start going to one national office. Rate rises from 4.1 to 6.0. This is the wave everybody cites.
July 2023. Revision 5 makes the identifier optional again and strips it before analysis. Rate rises from 5.2 to 7.5 — a 44 per cent increase in a single month.
February 2024. The office moves and the mailing address at the top of the form changes. Rate settles from 7.5 to 7.3. A small step, and it survives the method. We report it because leaving it out for being silly would be precisely the error this story is about.
The waves are revisions
The two waves discussed most often in the modern record are 2019 and 2023. Both sit exactly on a change to the paperwork — 2019 on where the form was sent, 2023 on what the form asked for. Neither sits on anything else we have been able to find: not a traffic figure, not a season, not an exercise schedule, not a film release.
The fourth change point is the smallest and the silliest. In February 2024 nothing was asked of the filer that had not been asked of him the month before. The address at the top of the page changed, and the level moved with it.
The shape of the historical record is the shape of the paperwork. A wave, in this record, is a form.
The people are not lying
This desk holds 2,300 first-hand accounts from aircrew and believes most of the people who gave them. That belief is not in tension with anything above.
A filing is not a sighting. It is a sighting plus a decision to write it down, plus a form, plus an address, plus a judgement about what happens to the paper afterwards. Everything we have measured lives in the second half of that sentence.
When the rate halved in November 2016, the plausible reading is not that the sky emptied. It is that several hundred people a year, for seven years, saw something and kept it to themselves.
The two we cannot check
1952 and 1973 are the two great waves of the American record. They are also the two we cannot run this test on.
The forms in use then do not survive. Neither do the schedules that would record when those forms were revised. We asked four archives and the agency's own records officer, over fourteen months, and got nothing back that we could use.
So the finding has an exact edge, and here it is. For the twelve years where the paperwork survives, every large movement in the record is a movement in the paperwork. For the two most famous years in the field, we cannot say, and neither can anyone else.
What we are not saying
We are not saying nothing is being seen. We have not measured that and this method cannot.
Captain Imogen Sarraf-Delaney, whose reply runs below in full, thinks we have measured the right thing and drawn the wrong lesson from it, and she makes a point this desk did not make and should have: if a name field can take sixty-one per cent off one category in that database, it can take sixty-one per cent off bird strikes and fatigue reports too.
She is right. That is a bigger story than ours and somebody should write it.
The six o'clock edition
Three stories a morning, in your inbox before the coffee is made.
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 Hollow PostTHEY CHANGED THE FORM TO STOP PILOTS TALKING
Read the 2016 revision as a deliberate instrument of suppression, ordered from above.
Had that we did not
The 2016 revision itself, which is real, and which nobody else had put next to the filing counts.
Left out
The 2023 revision, which restored anonymity and which their framing has no room for; and the memoranda around the 2016 change, which are about litigation exposure and are dull.
-
Signal & LedgerHazard Filings Rebound After 2023 Form Simplification
Trade-press piece on filing volumes as an operational metric for safety departments.
Had that we did not
The rebound, six months before anyone else noticed it, and the flight-hour normalisation we have borrowed here.
Left out
The subject matter. Their series counts all filings and never separates out the ones that mention traffic nobody could identify, which is where the effect is four times larger.
-
They Buriedthis newspaperSightings Fell 61 Per Cent The Month The Form Started Asking For A Name
Pre-registered a change-point analysis, ran it blind, then looked up what happened on the dates it returned.
Had that we did not
The phrase list, the code, the flight-hour series, and the four dates the method found before we knew what they were.
Left out
Our headline says sightings fell. Filings fell. A filing is not a sighting, and that distinction is the entire finding of this story, which our own headline blurs to sell it. — V. Ashcombe-Doyle, standards editor
Right of Reply
They Buried contacted Capt. Imogen Sarraf-Delaney, safety committee chair, a pilots' association on 12 August 2026, with the full draft, the figure and the underlying series, and no deadline imposed. Replied 17 August 2026, at length. Printed in full and unedited.
Your arithmetic is sound and I have checked it against our own membership numbers, which are not public and which I am not giving you. It holds. So take the following as an argument about what you have measured, not whether you measured it.
You have measured the willingness to file. You have not measured the incidence of anything.
Those are different quantities and your story knows it — you give the distinction a section of its own — but the shape of the piece, the headline, the chart, the phrase 'a wave is a form', all of it points the reader at one conclusion: that the thing being reported does not vary, only the reporting does. You are entitled to that conclusion. I am telling you it does not follow from your own data, and that your data is equally consistent with a constant flow of encounters and a workforce that got frightened in November 2016 and less frightened in July 2023.
I want to be specific about the fear, because 'fear' sounds vague and it is not.
In 2016 the identifier field arrived at the same time as three carriers began treating hazard filings as material in internal disciplinary matters. That is not a coincidence anyone has proved and I am not asking you to print it as fact. But my telephone rang for a year. The question was never 'will they laugh at me'. The question was always 'does this end up in my file'. A pilot who thinks a report ends up in his file does not file it, and the thing he saw is still up there, unreported, in your denominator, doing whatever it was doing.
Now the part that I think should have been your story.
If the record of unidentified traffic is this responsive to a change of stationery — sixty-one per cent, in one month, on a mandatory name field — then so is every other category in that database. Bird strikes. Runway incursions. Fatigue. Altitude deviations. The whole apparatus of aviation safety in this country rests on people volunteering that they made a mistake, and you have just demonstrated, cleanly and with a pre-registration, that the volunteering collapses when you ask for a name.
That is a considerably more alarming finding than anything about lights. You have it in your second paragraph and you have spent the rest of the piece on the lights.
I would still rather you had published it than not. My people have been told for forty years that the pattern in their reports is a pattern in their heads. Your story says the pattern is in the filing cabinet. That is progress of a kind, and it is still not a statement about the sky.
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.