Case: 2022, When Both Legs Fell

The year, decomposed

The mechanism is the one the last three lessons built. Policy rates rose faster than in any cycle in four decades, from near zero to above four percent, and the move was driven by inflation rather than by growth. Bonds fell by duration times the change in yields — an aggregate index with a duration around six years losing roughly 13% — and equities fell by the same arithmetic applied to much longer cash flows, with the compression concentrated in the longest-duration part of the market. The two legs therefore moved in the same direction, driven by the same variable, and the portfolio lost in proportion to its weights. What made the year unusual is not that both fell, which happens, but that the safe leg fell by nearly as much as the risky one and with less chance of recovery. A bond that falls because its yield rose has a higher yield thereafter, which is the arithmetic reason a longer holding period repairs it: the coupon and the reinvestment are better, and the yield to maturity at purchase is a reasonable expectation over the fund's duration. That is a real comfort and it is not immediate — for a fund with a six-year duration, the repair takes roughly six years, which is longer than most investors' patience and longer than most performance reviews. The equity leg had no analogous mechanical repair, because a lower multiple is not a promise of a higher future return in the same way. Which produces the asymmetry that defines the whole case: the asset that was supposed to protect the portfolio lost almost as much as the asset it was protecting, and it is also the one whose losses require the longest patience to undo. A balanced portfolio is a bet on a correlation, and in 2022 the bet lost. One million dollars, one year — Equities: 60% at −18.1%: −10.86 percentage points · Bonds: 40% at −13.0%: −5.20 percentage points ← · Portfolio: −16.06%, or −$160,600 · Equity return that would have broken even: +8.67%, with a risk-free rate moving from 0.06% to 4.4% The loss was not caused by holding bonds, and the conclusion is not that a balanced portfolio is broken. It is that the correlation between the legs is a regime variable, and the portfolio constructed on the assumption that it is a constant was mis-specified rather than unlucky.

What the case teaches, and what it does not

The first lesson is about inherited assumptions. Most portfolio construction for two decades used a negative stock-bond correlation estimated from a sample dominated by demand shocks, and reported it with a statistical confidence that was really a statement about the sample. A correlation estimated from twenty years of one kind of shock is not evidence about the opposite kind, and the absence of a supply shock in the sample is not a small omission — it is the whole regime the estimate came from. The remedy is not to abandon the framework but to use it with assets chosen for what they do in each quadrant rather than for their average correlation (MR12 for the quadrants themselves). The second lesson is about the difference between a nominal hedge and a real one. Bonds hedge nominal liabilities and demand shocks; they do not hedge inflation, and in an inflation shock they compound it. What hedges an inflation shock is a claim on real cash flows — real assets, businesses with pricing power, and to a lesser extent commodities and inflation-linked debt. None of those are risk-free, and all of them behave badly in the other quadrant, which is why the answer is a spread across quadrants rather than a substitution. The third lesson is about horizon. A duration loss on a bond is repaired by time and a coupon, which is a real property, and it is only a property for an investor whose horizon exceeds the duration. A portfolio with a five-year liability funded partly by a six-year-duration bond fund is not hedged at all in the intervening period; it has taken a duration bet and called it safety. The composition of the liability is what decides which of those it is, and that is the same point the real-returns lesson made about what safety means. • A correlation estimated in one regime is not evidence about the opposite regime. • Bonds hedge nominal liabilities and demand shocks; they do not hedge inflation. • A duration loss repairs over roughly the duration if you have the horizon. • Real-asset hedges are paid in the quadrant where nominal assets fail, and they fail in the others. The arithmetic of repair is worth stating once: a bond fund that loses 13% on a yield move from 4.2% to 5.4% has a yield to maturity of 5.4% afterwards, so the return from that point is higher. The loss is not forgiven, it is compensated over the duration — which is why the horizon of the investor and the duration of the asset have to be matched.

What the regime did pay, and how to tell which one you are in

A year in which both stocks and bonds fall is not a random event; it is the signature of a specific kind of shock. When the dominant shock is to **demand**, output falls and inflation falls, so policy eases, bond prices rise and equities fall for earnings reasons — the two legs move in opposite directions and the diversification in a balanced portfolio works. When the dominant shock is to **supply** or to inflation itself, output falls while prices rise, policy has to tighten into a weakening economy, and both legs fall at once. The correlation between stocks and bonds is not a constant of the financial system; it is a readout of which shock is dominant, and 2022 was the clearest supply-side readout in two decades. What a portfolio holds that is paid in that regime is a short list, and it is worth naming because it is the practical answer to “what do I own instead”. Cash and short-dated government paper became the hedge as the policy rate rose, since their yield moves with the policy rate and their price barely moves. Commodities and producers of them are the direct beneficiaries of the supply shock. The dollar strengthened through the period as global capital sought the funding currency. Value and energy equities, whose earnings are nearer to the cash flows the inflation is priced in, outperformed long-duration growth. None of these is a permanent hedge; each is what the *inflation-shock* branch of the tree pays for, and their usefulness lasts exactly as long as the branch does. The transferable skill is the diagnosis rather than the positions. Before deciding what hedges you, ask what shock would hurt the two things you already own, and whether the same event hurts both. If the answer is yes — if the hedge and the core asset fall on the same news — the diversification is present in the weights and absent in the behaviour. That is the assumption this case study is built around, and the lesson’s point is not that 60/40 failed but that a correlation estimated over a demand-shock sample was being used to size a portfolio that would have to live through a supply shock. • Demand shock: stocks and bonds diversify; the classical balanced portfolio works. • Supply or inflation shock: the same two legs fall, and diversification fails at the worst moment. • In that branch, cash, commodities, the dollar and near-term cash-flow equities are what is paid. • Test the hedge by behaviour, not by weight: would the same event move both holdings?

The UK LDI episode: when the safe asset was levered

The 2022 repricing taught one lesson about holding duration when inflation surprises. A different event in the same year, in the same month, taught a second and more durable one: what happens when the supposedly safe asset sits inside a leveraged structure. The episode was the collapse of the UK liability-driven investment funds in late September 2022, and its mechanism generalises far beyond one country’s pension system. The structure first. UK pension schemes held long-dated government bonds against long-dated liabilities, which is the textbook match. To improve the funding position without selling the bonds, many entered into derivative overlays — interest-rate swaps and repo arrangements — that replicated a longer duration than their cash assets provided. The collateral behind those overlays was, in the first instance, more government bonds. The result was a portfolio that looked conservative on every conventional measure and was, in fact, long duration with leverage on top of it. Then the trigger, which was not a credit event. A fiscal announcement moved long gilt yields sharply higher over a few sessions, and the move was large enough that the collateral behind the swaps was no longer sufficient. The funds had a choice between posting more collateral and closing the positions, and closing them meant selling the gilts that served as collateral — into a market where the marginal buyers were the very funds doing the selling. Yields rose further, which demanded more collateral, which forced more sales. That is a margin spiral, and it is the same shape as the one in the 1998 case where a highly levered fund was unwound. Three observations generalise. The first is that **leverage converts a price move into a margin call**, so the relevant question about any position is not only how much it can fall but what happens to its financing if it falls quickly. The second is that **liquidity and collateral are the same object in stress**: the bonds serving as collateral are also the assets being sold, so a structure that appears to hold two things — an asset and a source of financing — is holding one. The third is that the intervention that stopped it was a central bank buying bonds temporarily, not a change in policy: the function was to interrupt the spiral long enough for collateral to be posted in an orderly way. The lesson for a private portfolio is unglamorous and worth repeating. A position is not safe because the instrument in it is safe; the safety is a property of the structure, the financing, and what happens under a large, fast move. A government bond held outright is a different object from the same bond pledged as collateral against a duration overlay, and the difference is invisible in the yield. • Derivative overlays made a conservative portfolio long duration with leverage. • The trigger was a yield move, not a default — no credit event was required. • Collateral and the asset sold in the unwind were the same bonds, so there was one object. • Safety is a property of the structure and its financing, not of the instrument alone. The generalisable question for any leveraged structure: what is the collateral, who can demand more of it, and what would I have to sell to raise it? Where the answer names the same asset in all three parts, the position is one asset with a loop around it.

Sources

• The 2022 repricing of both asset classesContemporaneous index returns: broad equity indices about −18%, aggregate bond indices about −13%, in 2022
• Stock-bond correlation in inflation and disinflation regimesStandard empirical literature on the time-varying stock-bond correlation
• Inflation shocks and the failure of nominal diversificationMR12 on regimes; standard literature on inflation and asset valuations