The Yield Curve as a Growth Signal
The curve is a forecast of the Fed, and the Fed is a forecast of the economy — which makes the slope a secondhand growth forecast with a firsthand transmission channel through bank credit.
The yield curve is the most famous recession indicator in finance, which is precisely why it is so widely misused. The common failure is not ignorance of the signal but a category error about what kind of signal it is: investors treat an inversion as a timer — sell on the cross — when it is a regime statement with long and variable lags, and they treat it as an oracle when it is something more interesting: a price with a mechanism attached. Understanding why the slope carries growth information is what separates using the curve from merely quoting it.
The framing we use: the curve is a forecast of the Fed, and the Fed is a forecast of the economy. Long yields embed the market's expected path of future short rates; the short rate is where policy sits now. The slope between them is therefore the market's live answer to one question: how far is current policy from where policy will need to be? A steep curve says the market expects rates to rise toward or beyond neutral — an economy with room to run. An inverted curve says the market expects meaningful cuts — and central banks do not deliver meaningful cuts into strong economies. Inversion is the bond market pricing a future in which the Fed is forced to ease, which is a growth forecast wearing a policy forecast. The slope is a secondhand growth signal — and it comes with a firsthand transmission channel that most users of the indicator never price.
What the slope actually measures
Strip the curve to its expectations core and the slope is a statement about policy relative to neutral — the unobservable rate at which policy neither stimulates nor restricts. Nobody knows where neutral is, including the Fed; it is estimated with wide error bands and revised after the fact. But the long end of the curve embeds the bond market's working estimate of where short rates settle once current conditions wash out — a market-priced proxy for the destination. When the current short rate sits well below that destination, the market is saying policy is easy relative to where it belongs. When the short rate sits above long yields, the market is saying policy is restrictive — currently tight relative to its own long-run home, and tight in a way the market expects will have to be unwound. That is why inversion is the interesting state: it is the market's declaration that restriction is being applied, made by the only institution that continuously prices the question with money at stake.
Three layers of why
The bank-margin mechanism deserves one more beat, because it is the part that converts the curve from folklore into an input. It implies testable intermediate evidence: if an inversion is doing real work, it should show up in lending standards tightening, in loan growth decelerating, in the economics of term finance visibly worsening — before it shows up in employment or output. Sterling's Embedded Intelligence treats those credit-channel readings as the verification layer for any curve signal: an inversion that is confirmed by tightening credit is a different object from an inversion the credit data ignores. Sterling reads the slope, then audits the mechanism — because a signal with a mechanism can be checked mid-flight, and folklore cannot.
The record, and the two escapes from it
The historical record, stated qualitatively, is the reason anyone cares: across the post-war era, sustained inversions of the curve have preceded recessions with remarkable consistency, often while consensus forecasts and equity markets remained sanguine. The bond market's implicit growth forecast has repeatedly beaten the surveyed one. But the record's instructive cases are the escapes — the times the simple rule would have misled.
The first escape is the soft landing. The mid-1990s offered the canonical version: the curve flattened hard as the Fed tightened, the slowdown arrived, policy adjusted with a few well-timed cuts — and no recession followed, with the expansion running for years afterward. A flat or briefly inverted curve is the market pricing restriction; whether restriction ends in recession depends on how quickly it is withdrawn and what it collides with. The curve prices the stance, not the outcome with certainty.
The second escape runs the other way: an inversion vindicated by an event it could not possibly have foreseen. The curve inverted in 2019; a recession followed in 2020, delivered by a global pandemic. Whether that inversion would have called a recession on its own is unknowable — the shock arrived first and settled the question by force majeure. Honest use of the indicator counts that episode as neither victory nor failure, and the humility generalizes: the curve is one price, summarizing expectations that can themselves be wrong, distorted, or overtaken. A track record is not a law of nature — it is a mechanism that has usually operated, observed over a sample small enough to count on two hands.
The strongest case against this framework
The serious objection is that the slope is not a clean expectations object at all. A long yield is expectations plus a term premium — the extra compensation for bearing duration risk — and the premium component moves for reasons that carry no growth information: central bank asset purchases, regulatory demand for long bonds, foreign reserve flows, pension hedging. In an era when policy itself has deliberately compressed term premia, a flat or inverted curve may partly reflect the plumbing of the bond market rather than a priced recession. The indicator's celebrated record was compiled mostly in regimes with different plumbing; extrapolating it uncritically into a distorted-premium world is exactly the kind of error a good desk should refuse.
We accept this objection almost entirely — it is the strongest one in the cluster, and this library treats the decomposition of yield moves into expectations and premium as its own framework. The response is not to defend the raw slope but to refine the reading: the growth signal lives in the expectations component of the slope, and the practical fix is to weigh the slope alongside estimates of the premium, to prefer confirmation from the credit channel — the mechanism does not care why the margin compressed, only that it did — and to demand more corroboration in regimes where premia are visibly distorted. Note the asymmetry the mechanism preserves: even a plumbing-driven inversion still compresses lending margins. The forecast layer can be distorted; the transmission layer cannot be faked. That is one more reason to audit the mechanism rather than worship the sign of the spread.
Reading slope changes, not just levels
A final refinement separates practitioners from quoters: the curve's re-steepening after inversion has historically been the more proximate signal, and its composition matters. When the curve steepens because the front end is collapsing — the market pricing imminent, substantial cuts — that is the expectations component declaring that the restriction phase is ending because something is breaking. When it steepens because the long end is rising while the front holds, the message is different: reflation, fiscal supply, or a repricing of the destination rate. Same steepening, opposite growth content. The slope is not one signal but a grammar — level, direction, and composition each carry a clause — and reading only the level is reading one word of the sentence.
How to apply this framework
- Use the slope to set regime priors, never entry or exit points. Inversion says restriction is underway by the market's own estimate; it says nothing usable about dates. Let it raise the weight on late-cycle scenarios and tighten your standards for credit-sensitive and cyclical exposure — as an input, not an instruction.
- Audit the mechanism, not just the sign. A working inversion should be visible in tightening lending standards and slowing credit growth before it is visible anywhere else. An inversion the credit data ignores is a weaker signal; one the credit data confirms deserves full weight.
- Decompose the slope before trusting it. The growth information lives in the expectations component; term-premium distortions — QE, regulatory demand, hedging flows — can flatten or invert the curve without pricing any recession. In distorted-premium regimes, demand more corroboration.
- Watch the re-steepening and its composition. Steepening led by a collapsing front end has historically been the proximate warning — the market pricing forced cuts. Steepening led by a rising long end is a different sentence entirely. Level, direction, and composition together are the signal; the level alone is a headline.