The Gilt Crisis

In the Autumn of 2022, Liz Truss managed to stumble into one of the most dangerous financial crises to hit Britain since the 2007-8 financial crisis, with disaster averted only by quick-thinking intervention. Households still nevertheless felt the brunt of higher rates afterwards. So what happened, and why? How could a Prime Minister, a Chancellor with a Treasury full of advisors, not have seen the risk and avoided it in the first place?

Markets aren't very good at describing risk - we use the word in too many different ways. To the average person, it means something like "the chance of a bad thing happening". To some investors it means 'the chance I'll lose money', but to most institutions it tends to just mean "volatility" now - how much a price wobbles over a period of time. The UK gilt crisis is a great lens to unpick some of these and work out what went wrong and why we nearly stumbled into a disaster.

Frank Knight, an academic in the 1920s, set out a sensible division between 'risk' and 'uncertainty'. Risk, he thought, boils down to a measurable probability - bookies take risk when they offer odds on horses, insurers take risk when they underwrite your home against fire. Uncertainty, on the other hand, is all the stuff we simply can't know because we don't have enough information, or the system is too chaotic. There's a risk of rain in London this afternoon, but it's uncertain whether I'll get wet.

Government Bonds are one of the safest assets an investor can own - it's a formalised loan to the government. I pay them £100 today, they give me my money back in 10 years and pay me perhaps £4 interest for my trouble every year. And, because it's just a contract, I can sell that to other people, which makes it a very useful financial asset. That £4 interest attracts me to give them my money, and not stick it in a cash account - for very big investors, it's even safer than cash (which has to be kept somewhere, like a bank) because it's assumed that the risk of the British government is nearly zero.

Over a few decades, many British pension funds decided to reduce their 'risk' - their exposure to volatile equity markets with less certain payouts, by moving into much safer government bonds. The problem was that, if lots of people want bonds, then the government doesn't have to offer as much interest, so 'interest' rates fall and your annual payout gets smaller. With lower payouts, they decided to just borrow money and invest that in gilts too, juicing up their returns in these extremely safe assets. Add in a quirk - that if the 'interest' rate falls, old bonds are worth more because they pay out more, and you had a virtuous circle of assets that kept going up, soothing the fears about interest rates. So what went wrong?

Well, COVID came along first, and triggered a global wave of inflation as prices went up. With higher inflation, interest rates started rising. But as interest rates start rising, the value of bonds starts falling - why buy the old one when you can get a new one? Then Liz Truss came along, and announced a raft of policies that (markets believed) would lower tax revenue in the short term and increase inflation even more. Never mind whether it works, higher inflation means your key asset's price is going to fall - so they started selling. And when lots of people sell, the price falls faster; something that banks will notice when what's falling is your collateral for your borrowing.

What followed was a death spiral. These funds owned huge numbers of gilts, which were used as the super-safe collateral for their borrowing, which was used to buy gilts on leverage. As the prices collapse they have to sell into a collapsing market - and others started to notice, piling pressure by borrowing gilts and selling them to buy them back at lower prices while banks demanded their money back forcing the funds to sell even faster. In theory it's a cycle that could have gone on until all those funds went bust, perhaps taking a few major banks with them, but the Bank of England stepped in at the heeding of investors to become a 'Buyer of Last Resort', putting a floor on prices and stopping the collapse. There are reasonable parallels to the unravelling of the housing securities market in the US back in 2007-8.

What risks are there?

The reason all this could happen is that markets aren't really very good at 'tail risk' - the unlikely, rare, but disastrous events that could make it all collapse. There are two reasons for that. Firstly, the mathematical models we generally rely on massively underestimate how frequent they really are. Secondly, markets operate in a world with high expectations and very limited patience - if a crisis comes along every ten years, but your conservatism means you lag your peers for five in performance, you may not exist by the time the crisis comes along, anyway. It's easier (and ironically safer) to act as if the tail events won't happen, and then hope that you'll navigate the next crisis slightly better than your peers.

Let's break risk down into some more detailed categories to see what was going on here:

Measurable Risk

The most comfortable variety - things that we can plug into a spreadsheet, like interest rate exposure, volatility of prices, correlations between asset classes. These risk are quantifiable, have historical data, and show up in charts. In general we've designed systems to handle these pretty well.

Model Risk

This is where our tools and framings start to mislead us. Perhaps our assumptions are wrong, or our data is incomplete. Perhaps the world is changing subtly without our notice - value stocks aren't reverting to fair value any more, or a value-at-risk model wrongly estimates the risk of a collapse.

Regime Risk

The risks underneath us, where the rules of the game can change entirely. Interest rates that suddenly spike after a global pandemic. A government banning short-selling overnight and destroying investors. A war and a sudden shock to global supply chains. It's not just bad outcomes, it's unexpected new paradigms - think back to Kuhn in Philosophy.

Reflexive Risk

Risk introduced into the system by us, the investors, as a consequence of our beliefs and actions which - in turn - change the reality of the markets themselves. This realisation was Soros's great breakthrough theory, and underscored much of his success.

Meta-Risk

My own innovation, a metric of our own physiological stress and capacity for rational thought and attention allocation. This can be driven by our personal circumstances, but is often heavily shaped by the cultures of the institutions in which we work. It's the interaction of this with the outside world that helps shape reflexive behaviours at market level.

In the case of the gilt crisis, we all had the measurable risk nailed down. We knew the volatility of gilt prices and interest rates, their correlations, their liquidity in normal conditions (Measurable Risk). But rapidly rising inflation as we starting come out of COVID upended it, driving faster changes than we expected (Model Risk). Truss's announcement, perhaps benign if it'd been in 2018 or 2012, became the spark that ignited a panic (Regime Risk). Asset prices started falling too quickly, causing a self-fulfilling (reflexive) prophecy of collapse. Once hedge funds, banks, and other investors spotted it the selling became a stampede for the doors as people try not to be the last one out (Meta & Reflexive Risk), and the leverage underlying it started collapsing like a house of cards. Perhaps there's a floor it would have naturally rested at, but the old wisdom of 'don't try to catch a falling knife' might mean there are few brave enough to take the other side of the trade.

Why do risk models get it so wrong?

These sorts of episodes are more common than our models often assume, and they're baked into the structure of markets. What really distinguishes a crash is the disappearance of liquidity. People always want to buy or sell things in markets, and the price adjusts to reflect the balance of buyers and sellers. But sometimes there's no buyer (or seller) and someone who is being forced to trade, perhaps because of a regulatory requirement (short-covering), or because their own clients are demanding their money back. In those instances prices can go into freefall, and speculators might help give those prices a little nudge to take advantage. It's a hard problem to fix, given we can't just order private individuals or institutions to just step in - if we care about property rights, at least.

More broadly, we can't forget the human element of markets. As John Coates points out, the participants in markets during a crash start to resemble a 'clinical population', as they freeze up and panic. Normally calm, measured individuals may see their career flash before their eyes and think of mortgages and school fees, job loss and humiliation. They may rush into decisions, and these sorts of panics are contagious - we register the panic of others around us and flood our brains with adrenaline and cortisol.

These models also tend to assume that markets are smooth, continuous, and ultimately self-correcting - that if we can just sit tight, things will revert to the mean. But that assumption only holds in what mathematicians call ergodic systems - where the average outcome over time is the same as the average outcome across many parallel worlds. Most of finance isn’t like that. In the real world, you only get one path - you can’t spread your risk across infinite timelines. You either survive the crash or you don’t.

This is why time, not just probability, matters. Risk isn’t just a question of what might happen - it’s a question of whether you can survive long enough to benefit from the upside. Averages can hide the damage along the way, and the path can be more dangerous than the destination. You can have a perfectly good long-term strategy that fails simply because it exposes you to ruin at the wrong moment; and when that ruin is system-wide - when everyone’s holding the same asset, relying on the same liquidity, drawing from the same capital pool - the whole structure can break.

So the real failure of risk models isn’t that they don’t run the numbers - it’s that they often ignore the structure those numbers exist within. They assume a world of independence and equilibrium, where prices reflect fundamentals and volatility is just noise. But the world we invest in is made of feedback loops, institutional incentives, emotional pressure, and single-path outcomes.

Where that leave us

So perhaps the real task of risk management isn’t prediction, it’s preparation. Not just for the known risks we can model, but for the unknowable interactions between structure, behaviour, and belief. The crashes that matter aren’t caused by volatility - they’re caused by misplaced confidence, misplaced liquidity, and misplaced assumptions about what others will do. A good investor doesn’t just track positions. They map fragilities. They ask: what am I depending on? Who else is depending on the same thing? And what happens when we all try to move at once?

In the next piece, I’ll turn to that question of movement—how we forecast, how we update our beliefs, and how we try to stay upright in a world we can’t predict. Because if risk teaches us anything, it’s that survival depends not just on having the right model - but on knowing when the model might break.