UK Gilt Crisis 2022: When the Safest Asset Broke

Britain’s defined-benefit pension schemes are the most regulated, most conservative institutional investors in the country. At the last count there were 5,051 of them, holding £1.4 trillion for 8.9 million members.

In late September 2022 they were forced to raise over £40 billion in cash and sell over £30 billion of government bonds, in a matter of days. Not because of crypto. Not because of tech stocks. Not because of anything you would recognise as a risky bet.

Because of British government bonds. The safest asset in the country.

The Bank of England had to step in as an emergency buyer to stop it. A senior London banker, quoted at the top of the Bank’s own staff working paper on the crisis, put it about as well as anyone has: “It was not quite a Lehman moment. But it got close.”

And the risk that made it explode is called duration. It sits inside every long-dated bond fund in the world, and it needs no leverage at all to hurt you.

2026 update

Britain's 10-year gilt yield is higher now than at the worst moment of 2022

Worst close of the 2022 crisis

4.65%

10 October 2022

Highest close of 2026

5.43%

8 October 2026, the highest since July 2007

Largest three-day rise

+0.90 · +0.27

percentage points, 2022 against 2026

It isn't only Britain. On October 5, 2026, the 10-year US Treasury closed at 5.31%, its highest close since May 2002. Everyone says that's because Washington borrows too much. I checked the debt, inflation and the Fed against the data.

Why are interest rates so high in 2026? →

The powder keg

To understand what exploded you have to understand what had been loaded, patiently, for two decades, with the approval of every regulator involved.

A defined-benefit scheme makes a promise: fixed payments, for life, decades into the future. When interest rates fall, the present cost of that promise rises. After 2008 rates collapsed and pension deficits blew open across the industry — not because anything went wrong with the investments, but because the arithmetic of discounting future liabilities turned against them.

Two things followed. The first was a wholesale change in what these schemes owned.

The powder keg · UK defined-benefit pension assets, 2006 and 2022

In sixteen years, bonds went from 28% to 72% of what British pension schemes owned

Nothing about the gilt crisis makes sense until you see what British defined-benefit schemes had been holding, and what they had stopped holding.

The lines cross, and that is the point. A sector that had been mostly an equity investor became overwhelmingly a bond investor — deliberately, with regulatory approval, for sound accounting reasons. It also meant that when gilts moved violently, they moved through almost everything these schemes owned.

The second was Liability-Driven Investment, or LDI. The idea is genuinely sensible: use derivatives and repo to hedge the interest-rate exposure so the scheme is no longer at the mercy of the discount rate, and use leverage so most of the capital stays free for growth assets instead of sitting in bonds.

This was not a fringe product sold by cowboys. Liabilities hedged this way grew from £400 billion in 2011 to £1.5 trillion by 2020 — about two thirds of UK GDP — on figures the IMF attributes to the Investment Association. Total UK LDI assets stood at around £1.4 trillion at the end of 2021, and the top three managers held roughly 70% of that market between them.

Concentration on that scale has a specific consequence. When the trigger came, the sector did not behave like thousands of investors making thousands of decisions. It behaved like a handful of firms doing the same thing at the same moment.

And the stress test? The Bank of England’s Financial Policy Committee had run a scenario with a maximum shock of 100 basis points.

Hold on to that number.

The spark

23 September 2022. The Chancellor announces permanent tax cuts that the Treasury’s own estimate put at £45 billion a year by 2026-27. The Institute for Fiscal Studies later read the market’s reaction as a lesson in how quickly “fiscal and institutional credibility” can be lost.

The verdict came within hours. At the Bank of England’s daily 4pm rate, sterling lost 4.6% against the dollar in four trading days, from $1.1265 to $1.0745 on 28 September, its weakest fixing since March 1985. Long-dated gilt yields did something the Bank of England had not modelled: a move of 130 basis points in the 30-year nominal yield in just a few days, on the Bank’s own description, or 140 basis points over three days on the IMF’s. Both institutions reach for the same word for it, which is historic.

The gap that mattered · movement in the 30-year gilt yield, basis points

The market went a third past the worst case anyone had tested, in a few days

One hundred basis points is one percentage point.

What the stress test assumed 100 bp

Bank of England FPC scenario, before the crisis

What the market actually did 130 bp

30-year gilt yield, a few trading days, September 2022

What LDI funds must survive now 250 bp

Minimum resilience recommendation, after the crisis

The stress test was not narrowly missed. The market moved about a third further than the worst case anyone had modelled, and it did it in days. The requirement that now applies is nearly double what actually happened — which tells you how far behind the old assumption had fallen.

The stress-test ceiling was not approached and then narrowly exceeded. It was shattered on the first day.

The ten-year yield tells the same story, and the Bank publishes it daily over a much longer record. In the Bank of England’s daily series it rose 90.1 basis points in the three trading days from 22 to 27 September, 34.3 of them on the 23rd alone. No three-day window since the series begins in November 1993 comes close: the previous record, set in October 1998, was 58.1. Three more places in the top six all fall in the spring of 1994, at 55.9, 50.8 and 42.8.

10-year gilt yield · largest three-day rises since 1993

September 2022 beat the old record by more than half again

Every three-trading-day window since November 1993, ranked by how far the 10-year yield rose. Red: the 2022 crisis. The dashed line is the largest three-day rise of 2026 so far.

Two of the five largest moves in more than thirty years came within a fortnight of each other, in autumn 2022. The first started the cascade; the second, ending on 10 October, came as the Bank's purchase programme was entering its final week.

Two of the five largest moves in more than thirty years came within a fortnight of each other. The second, 47.9 basis points from 5 to 10 October 2022, came while the Bank was already buying.

The spark

The 30-year yield rose 270 basis points in ten weeks, and 140 of them came in three days

The 30-year gilt is the longest promise the British state makes. This is what happened to what it costs, between the first of August and the middle of October 2022.

• 23 Sep Mini-budget: £45bn of unfunded tax cuts, no OBR assessment
• 28 Sep Bank of England intervenes
• 10 Oct Daily auction cap raised from £5bn to £10bn
• 14 Oct Operations end. Yields peak above 5%

The cascade, in five steps

Fifty people are standing on chairs, trying to stay above rising water. One chair breaks. That person grabs the next and pulls them down. More chairs break. Everyone grabs, everyone falls — and the water never rose any further. The panic destroyed the room.

Here is that mechanism, in the gilt market, in order.

One, the price shock. Gilt prices crashed. LDI funds held large leveraged positions in exactly those bonds, so the leverage multiplied every point of the move into a mark-to-market loss.

Two, the collateral calls. The counterparties on the other side of those repo and derivative contracts demanded cash. Not next month. Overnight.

Here is the structural trap almost nobody appreciates. The schemes could not move at that speed. Raising cash inside a pension scheme means trustee approval, and the IMF puts that process at one to two weeks. They were facing calls due before the next business day. They were not insolvent, and in most cases they were not even short of assets. They were short of cash, against a deadline they were constitutionally incapable of meeting.

Three, the fire sales. With no time and no cash, the managers sold the only thing that could be liquidated immediately. More gilts.

Four, the doom loop. Those sales pushed prices lower, lower prices triggered new calls, new calls forced new sales. No individual participant could exit, because exiting meant defaulting. The general pattern has a name, forced deleveraging, and it is the same loop that turned March 2020 into a fire sale of everything.

Five, the market breaks. Between 23 September and 14 October, net gilt sales by the LDI and pension sector came to over £36 billion. Three firms accounted for over 70% of that sector’s sales to primary dealers. The Bank of England’s assessment was that multiple LDI funds were likely to fall into negative net asset value, and at zero the lenders seize the collateral and liquidate it into a market that already has no buyers.

To see the scale problem, hold two numbers next to each other. Roughly £200 billion sat in pooled LDI funds. The traded gilt market they were selling into is around £1.4 trillion, and it is the foundation the UK financial system rests on, underlying some £2 trillion of lending. A seller that size does not find a price. It moves the price, and then has to sell into the price it just moved.

The part that surprised the researchers

The Bank’s own transaction-level study found something the headlines missed. The selling was heaviest not in ordinary gilts but in index-linked ones, which are only about a quarter of the market — because index-linked gilts were what the funds had pledged as repo collateral. The strongest single predictor of how much a firm liquidated during the crisis was how much index-linked repo exposure it had gone in with.

The damage did not stay in that corner either. Transaction costs rose across the gilt market, including for investors who had nothing to do with LDI, and including on small trades of the kind a retail investor makes. Aggregate price dispersion more than doubled within days, and the increase in costs was worse at smaller dealers than at large ones.

Hedge funds, for their part, did provide liquidity — modestly, and profitably. They timed it to maximise their return, which is exactly what you would expect and exactly why a market of willing buyers is not the same thing as a backstop.

Two things almost everyone got wrong

“Government bonds are risk-free.”

Institutionally, risk-free means one thing only: no credit risk. Britain will pay you back in 2052. That is a genuine and valuable property, and it held throughout the entire episode.

But a 30-year gilt, levered three times, loses roughly 60% of your equity on a one-point move in rates. The label says nothing whatsoever about the risk that nearly broke Britain. It says nothing about duration.

“The pensions nearly went bust. Retirees almost lost everything.”

Backwards, and the tabloids printed it anyway. Rising rates cut the present value of future obligations faster than asset values fell, so schemes came out of it better funded on paper than they went in. By 31 March 2023 the Pension Protection Fund put total scheme assets at £1,404.4 billion against section 179 liabilities of £1,045.5 billion — a funding ratio of roughly 134%.

This was never a solvency crisis. It was a pure liquidity crisis in the leveraged structures sitting in front of the pensions.

The rescue, and the ultimatum

On 28 September the Bank announced it would buy long-dated gilts at up to £5 billion a day for 13 business days, a ceiling of about £65 billion. Daily turnover in that market was around £12 billion, so the promise was roughly five days of the entire market.

It spent £19.3 billion of it: £12.1 billion of conventional gilts and £7.2 billion of index-linked. Under a third of the ceiling. Every one of those gilts was later sold back into the market.

The rescue · Bank of England gilt purchases, 28 September to 14 October 2022

The Bank offered £65bn and needed £19.3bn: the promise did most of the work

The Bank of England committed to buy up to £5 billion of long-dated gilts a day for thirteen business days. Here is that promise, and inside it, what it actually cost.

The empty space is what stopped the panic. A seller facing a market with no bid behaves very differently from a seller who knows a buyer of last resort is standing there with £65 billion of permission. Most of that permission was never used, and never needed to be. The Bank later sold every gilt it bought back into the market.

What actually broke the panic was not the volume. It was the credible announcement that a buyer existed. The ten-year yield fell 41.1 basis points on 28 September, the day of the announcement — more than it had risen on any single day of the sell-off.

The Bank was blunt about the terms. Its own statement, published three days after the programme closed, records that it terminated the operations and ceased all bond purchases on Friday 14 October, and that the purpose had been “to provide time for LDI funds to address risks to their resilience from volatility in the gilt market, not to provide a permanent backstop.” It had raised the daily ceiling from £5 billion to £10 billion on 10 October, and it stopped on the date it had named from the beginning.

That is not reassurance. That is the lender of last resort telling the market to recapitalise now or face this alone. It worked: LDI funds raised over £40 billion and sold over £30 billion of gilts during the operations, and the programme ended on schedule.

The bond market removed the government

The market did not merely price the policy. It destroyed the administration that produced it. On 14 October the Chancellor was dismissed and his successor reversed almost every measure. On 20 October the Prime Minister announced her resignation, and on 25 October she left office, having served 49 days — the shortest premiership in British history.

2026: higher than 2022, and nothing like it

Here is the part the story usually leaves out, because it happened after everyone stopped watching.

The worst daily close of the 2022 crisis, for the ten-year gilt, was 4.65% on 10 October 2022. The yield first closed above that level on 6 July 2023. Since 12 March 2026 it has closed above it on every trading day, 151 trading days up to 15 October, and on 8 October 2026 it closed at 5.43%, the highest close since July 2007. It ended September at 5.37%, a rise of 0.91 percentage points since the last close of 2025 (4.47%), and closed at 5.31% on 15 October, the latest Bank of England figure when this was updated. The Bank of England’s own rate, meanwhile, has sat at 3.75% since 18 December 2025.

So by the measure most people would reach for, the level, 2026 is the worse year. And nothing like September 2022 has happened in the gilt market. The largest three-day rise of 2026 was 0.27 percentage points, to 20 March. In September 2022 it was 0.90.

That gap is the whole lesson of the crisis, measured. A yield level does not break a market. Speed does, when it meets leverage that was sized for a slower world. A rise spread over nine months gives every pension trustee time to meet, every fund time to raise cash and every margin call time to be paid. A rise compressed into three days gives nobody time for anything, which is how a 100-basis-point stress test met a move several times its size and lost.

The other difference is who moved.

10-year government bond yields · rise over the period

In 2022 Britain moved alone. In 2026 everyone moved together

Two different periods on two different scales. Read the shape of each panel, not the heights across them.

September 2022: Britain alone

3 trading days, 22 → 27 Sep · basis points

2026: everyone together

9 months, end of 2025 → 30 Sep 2026 · percentage points

In 2022 Britain moved alone, on a budget. In 2026 the whole rich world moved together, and America moved the most. A move that is shared, and spread over months, is a repricing. A move that is British and compressed into days is the kind that finds the leverage.

In the same three trading days of September 2022, the ten-year US Treasury yield rose 27 basis points and the German Bund 28. Britain rose 90. The Bank’s December 2022 Financial Stability Report makes the same point for long-dated gilts: after the fiscal announcement they rose more sharply than equivalent US and German bonds. In 2026 the picture is reversed. From the end of 2025 to 30 September 2026 the ten-year yield rose 1.11 percentage points in the United States, 0.99 in Japan, 0.91 in Britain and 0.70 in Germany. Britain is in the middle of the pack, not out on its own.

A shared repricing across every large bond market is a different animal from a run on one of them. It has its own causes, and they deserve their own article rather than a paragraph here. We tested six common explanations for it in a research note: Why US Bond Yields Rose in 2026.

What it does not do is make duration irrelevant. A long gilt fund loses the same amount on a 0.9-point rise whether the rise takes three days or nine months; the slower pace only removes the forced seller, not the loss. Which is why the next section matters more in 2026 than it did when this article was first written.

One limit on all of this, because I would rather state it than imply the opposite: the data show the absence of a 2022-style spike. They do not show that no fund anywhere was under strain, and I have no source that says so either way.

What this means for your portfolio

You do not run a pension scheme and you are not levered three to one. So why does this matter? Because duration does not need leverage to hurt you. Leverage only decides how fast.

CHECK — one number, one factsheet

Open the factsheet of any bond fund you own and find modified duration.

What modified duration costs you · yields up one percentage point

One point on yields takes about 15% from a long-dated gilt index

Approximate fall in value. The rule is as blunt as it looks: duration is roughly the percentage you lose.

Short-dated bond fund −2.0%

modified duration 2.00

Typical aggregate bond fund −7.0%

modified duration 7.00

FTSE Actuaries UK Conventional Gilts Over 15 Years a real index −15.1%

modified duration 15.07

Very long-dated or zero-coupon −20.0%

modified duration 20.00

The third row is not a worked example. It is the published duration of an index that British pension funds and ordinary investors actually hold, and it means a single percentage point on yields takes roughly fifteen percent of the value — with no default, no credit event and no leverage anywhere in sight.

That is your actual exposure. Not default risk. Duration risk, the same risk that nearly broke Britain, printed on a document most people scroll past on their way to the fee.

AVOID — the phrase “conservative bond fund”

Here is what that exposure has actually done. The FTSE Actuaries UK Conventional Gilts Over 15 Years Index, the benchmark for long-dated British government debt and the safest borrower in the country, was down 48.1% over the five years to July 2026, with a maximum drawdown along the way of 52.6%.

No default. No credit event. No issuer in any distress whatsoever. Just the safest asset in Britain, halving.

The Pension Protection Fund’s own market table gives the mechanism in one line. Over the year to 31 March 2023, ten-year gilt yields rose 1.91 percentage points and five-to-fifteen-year index-linked yields rose 2.56. Multiply a move like that by a duration in the teens and you have the entire story, with no crisis required.

A long-duration bond fund in a rising-rate environment is not a conservative holding. It is a concentrated directional bet that rates will stay low, and it should be held, if at all, by someone who knows that is the bet they are making. (What a government bond is safe from, and the two risks the word hides, has its own page.)

BUILD — a ladder, and a buffer

Instead of a single long-duration fund, build a bond ladder: short, medium and long maturities, each held to maturity. When you hold to maturity the mark-to-market swings become noise rather than loss. You collect the coupon, and you receive the principal at par.

It is more administrative work than one fund, and it will not shield you from opportunity cost. What it removes is the mechanism that turned a price move into a catastrophe: being forced to sell.

Hold a cash buffer for exactly the same reason. Every institution in this story had assets. What nearly killed them was needing cash on somebody else’s schedule.

Never be forced to sell. That is the whole of it.

The thing worth remembering

British pension schemes held £1.4 trillion, and they used the safest asset in the country as collateral for leveraged positions. When the market turned, the safety itself became the weapon, because the safety label is precisely what persuaded everyone to stop looking at the leverage.

Regulators responded seriously. LDI funds are now expected to withstand a 250 basis point shock, well above the 130 to 140 basis points in the 30-year yield that nearly ended the system, and the structure is meaningfully safer than it was. Four years later the ten-year yield went higher than it ever did in 2022, slowly, and nothing like the 2022 spike came with it.

The principle underneath has not moved an inch. Leverage on a safe asset is the most dangerous trade in finance, because nobody sees it coming until the chairs start breaking.