The Rehire Is the Audit

The interesting thing about the Geddes Axe is not that it was wrong. Parts of it were plainly right. Postwar Britain was carrying inflated wartime establishments and somebody was going to have to say so out loud. The interesting thing is the accounting.

The savings were denominated in pounds sterling and dated to the 1922–23 fiscal year. The costs were denominated in things that do not sum: teacher-to-pupil ratios, dry-dock capacity, the number of people inside a ministry who know how a particular thing works and whom to call at two in the morning. Those costs came due in later years, under other names, credited to whatever was in the newspaper that week. Nobody filed a variance report. There was no line item that read deficiency, capacity, previously deleted.

That is not a moral complaint. It is a units problem, and the units problem is the whole mechanism. An efficiency drive is a trade in which one side of the ledger is legible, signed, and dated, and the other side is a fog. Which means the trade will always look good at the moment it is booked, and will always look good in the summary, and will only ever be evaluated by an event.

And that is a property of the instrument, not of anyone's judgment. An institution whose entire output is a non-event — a flood forecast issued in time, an intrusion that never landed, a case closed before it became a crisis — cannot be scored by a system that counts artifacts, because it produces none. Which means the accounting selects. Not the Chancellor, not the committee, not the administrator: the measurement itself decides what kind of capacity a state is permitted to keep, and it decides against prevention every time, in every era, regardless of who is holding the pen.

The United States has now had eighteen months of this at scale, and the useful thing about the American version is that one of the cuts has already been reversed. That reversal is the only clean instrument we have.

i · the rehire is the audit

Here is the test, and it is a test the argument can fail.

If a cut found genuine waste, the function does not come back. Nobody repurchases a redundancy. The position stays empty, the org chart absorbs the gap, and five years later nobody can name what used to be there.

If a cut removed load-bearing capacity, the organization buys the capacity back. Not because anyone admits error — admission is not required and rarely offered — but because the function was doing something and the something stopped.

The instrument only runs in one direction, and it is worth being exact about why, because the reason is the same pathology the rest of this piece is about. A rehire is hard evidence: the function was load-bearing. An empty chair is evidence of nothing at all. And what the requisition actually detects is not load-bearingness — it is the political visibility of failure, whether the gap produced something a camera could point at inside a news cycle. An agency whose failures are non-events will never trip the instrument no matter how much weight it was carrying. Which is why the audit only ever fires on hurricanes.

So: at the National Oceanic and Atmospheric Administration, the cuts landed as roughly a ten percent budget reduction and the loss of about seventeen percent of staff, on the order of 2,500 people. At the National Weather Service specifically, after the Texas floods, the agency was authorized to add 450 employees back. By the end of June it had managed about 275 of them, and hoped to reach the full 450 by September.

Four hundred and fifty meteorologists, repurchased — because water moved through Texas on television.

That is the audit. It was not conducted by an inspector general and it did not produce a report. It was conducted by the hiring requisition, which is a more honest document than anything an oversight committee will generate, because nobody writes a requisition for a job that isn't being done.

And note what the repurchase does not restore. Headcount is a number; a forecast office is not. The person who left had eight years of knowing which radar return over which ridge means the water is about to move, and that knowledge does not arrive with the replacement. It reassembles slowly, from experience, over seasons — and a season is exactly the interval during which the thing you are watching for happens. The rehire closes the staffing gap immediately and the competence gap on a delay nobody has costed, because there is no unit for it.

The buy-back is also priced at retail. Recruiting, training, and seasoning an entry-level meteorologist costs more than not firing an experienced one, and the difference is not recorded anywhere as a loss attributable to the original decision. It shows up as ordinary hiring expense in a later budget, under a different heading, signed by different people. The ledger closes clean on both ends. This is the Geddes accounting, reproduced exactly, a hundred and four years later, with better software.

ii · the vector that doesn't hold

Three vectors are usually offered — hurricanes, cyber intrusion, identity theft. Intellectual honesty requires saying that they are not equally strong, and that the weakest one is the most instructive.

Start with the weak one. At the Internal Revenue Service the staffing fell from roughly 102,000 to roughly 74,000 in a year, a twenty-seven percent reduction, on top of an appropriations cut of $1.1 billion that leaves the base budget some forty percent below its 2010 level in real terms. More than 500,000 victims of tax-related identity theft are currently waiting on case resolution, at about twenty months per case. The identity theft victim assistance line answers about nineteen percent of calls.

Those numbers are appalling and they are not evidence of what they are usually offered as evidence of. In fiscal 2023 — well before any of this — the same backlog stood at about 484,000 cases at about nineteen months. The dysfunction predates the cuts. If you are building a causal case that DOGE created the identity theft crisis, the data will not carry you, and you should stop.

What the data does carry is worse and less quotable. The backlog was a known, named, documented failure with an identified remedy, and the remedy was staff. The National Taxpayer Advocate had been saying so in writing, annually, for years. What was removed was not the function. What was removed was the repair capacity for a failure already in progress — the only mechanism by which the 484,000 was ever going to become zero. Twenty months instead of nineteen is not the harm. The harm is that the number is now structurally unfixable, and there is no year on the calendar where anyone will notice this happening, because nothing will have changed. That is the signature of the whole exercise: the damage that consists of removing the thing that was going to fix the damage.

The strong structural case is the cyber one, and it is strong precisely because it will never generate a number like 500,000. The Cybersecurity and Infrastructure Security Agency went from about 3,400 people to about 2,400 — roughly a third of the agency. The election security program was frozen and its team eliminated. The FY27 proposal would remove approximately $700 million in gross program spending, a net reduction near $360 million, and formally zero out the Election Security Program: fourteen positions, about $39.6 million. In August, five House Democrats asked the Government Accountability Office to examine what, specifically, has been lost.

They will not get a satisfying answer, and this is not GAO's fault. Ask what a cyber defense agency prevented and you are asking for a count of intrusions that did not occur, coordination calls that were not needed because someone was already patched, county clerks who did not get a call because the call had nowhere to originate. Prevention has no artifact. It produces an absence, and an absence has no invoice. An agency whose entire output is non-events will always, in every fiscal year, appear to be the safest thing to cut — and the appearance is generated by the measurement instrument, not by the agency.

CISA has not been rehired, and it will not be. Not because the cut found waste. Because the instrument that found the waste at the Weather Service cannot be pointed at a building where nothing has visibly gone wrong.

iii · the rule that renews itself

In August 1919 the British Cabinet adopted a planning assumption: the empire would not face a major war for ten years. It was reasonable at the time. In 1928, at the urging of the Chancellor of the Exchequer — a man named Winston Churchill, who would later be given considerable credit for noticing the opposite — the rule was made self-renewing. Each year, the ten-year horizon rolled forward another year. Defense estimates were built on it.

The genius of the Ten Year Rule, and it was a genius, is that it validated itself annually by the absence of the event it was disarming against. Every quiet year was evidence the assumption held. There was no year in which the rule was falsified until the year in which it was catastrophically falsified, and by then the manufacturing base needed most of a decade to respond. The rule was abandoned in March 1932. The rearmament it made necessary was still incomplete in September 1939.

That is the structure now operating, and it is running on institutions rather than shipyards. The FY27 proposals will be scored against a year in which no election was successfully attacked, no hurricane forecast failed catastrophically, and no federal network breach became a household name. Each of those absences will be read as headroom. Each will make the next reduction look like the previous one — prudent, overdue, obviously fine.

The absences are not headroom. They are the product being delivered.

I do not think the reckoning arrives as a discrete event with a budget line attached, and that is my actual prediction. There will be a storm, and it will have a first name, and the postmortem will discuss its unusual rapid intensification. There will be an intrusion, and it will be attributed to a foreign service with a cute animal codename. There will be a filing season where a great many people simply do not get their money and conclude the government is like that. Nobody will connect these to a spreadsheet from 2025, because the spreadsheet from 2025 was denominated in dollars and these will be denominated in weather, adversaries, and the background hum of a state that does not work.

The Grace Commission promised $424 billion in three-year savings in 1984. Congressional scorekeepers found a small fraction of it was real, most of that requiring legislation nobody passed. The report is on a shelf. The number is in nobody's memory. What people remember from that era is a general sense that the federal government had become somewhat worse at things, which is true, and which is attributed to a mood.

Four exercises across a century, run by men with nothing in common, no knowledge of one another's arithmetic, and no shared politics — arriving at the same place by the same route. That is not a recurring failure of judgment. That is what a measurement instrument does when you leave it running for a hundred years.

The Geddes Axe fell in 1922. British schools were still short of the capacity in 1930. Nobody called it the Geddes deficit. They called it the schools.