The yield curve plots the interest rate (yield) on otherwise-identical government bonds — almost always U.S. Treasuries — against their time to maturity, from overnight bills out to 30-year bonds. A normal curve slopes upward: long-term debt yields more than short-term debt, the usual state of affairs. An inverted curve slopes downward: long rates sit below short rates, an unusual configuration that historically has been one of the most reliable single advance warnings of U.S. recession. The core tension is that a downward slope is economically "wrong" — you'd normally demand more to lend for longer — so when it happens it is telling you the bond market expects the central bank to be forced to cut rates sharply in the future, i.e. that the economy is headed for trouble.

How it's formed

Two forces set the shape of the curve, per the standard term-structure framework:

• Expectations of future short-term rates. Under the (pure) expectations hypothesis, a long-term yield is roughly the average of expected future short-term rates. If the market expects the Fed to cut rates over the next few years, far-dated yields get dragged down and the curve flattens or inverts (Wikipedia, Expectations Hypothesis; St. Louis Fed).
• The term premium. Investors normally demand extra yield — the term premium (sometimes called the liquidity premium) — to lock money up longer and bear the price/inflation uncertainty of distant maturities. A positive term premium is what gives the curve its default upward tilt (Brookings Hutchins Center).

So the math of inversion is simple: it requires the expected decline in short rates to be large and persistent enough to overwhelm the term premium (St. Louis Fed). In practice this happens when the Fed has hiked the policy rate (lifting the short end) while the market simultaneously prices in future cuts and low long-run inflation/growth (pulling the long end down).

Two spreads dominate the literature:

• 10-year minus 2-year ("2s10s") — the most widely watched by traders and media.
• 10-year minus 3-month — the spread the New York Fed uses in its official recession-probability model, and the one Estrella & Mishkin (1996) found has the strongest predictive record. Their probit model maps the spread to a recession probability via P(recession) ≈ Φ(−0.5333 − 0.6629 × spread), published monthly by the NY Fed.

Inversion is when the relevant spread goes negative.

How it's used in practice

The inverted curve is treated as a leading macro indicator, not a trading trigger. The empirical regularity most analysts rely on:

• Inversion typically precedes recession by roughly 6–18 months, with averages commonly cited around 11–12 months — but the lead time is highly variable across cycles (multiple sources; see caveats below).
• Because of that lag, the curve says little about the next few months of equity returns. Markets and the economy have often kept rising for a year-plus after the first inversion.
• Professionals lean on the NY Fed 10y-3m probability model for a quantified read rather than eyeballing the slope, and frequently note that deeper and longer inversions historically carried more weight than brief, shallow ones (commonly cited; treat specific accuracy percentages from blog sources with caution).

A second, often-overlooked signal is the re-steepening (un-inversion): historically recessions have tended to begin after the curve climbs back above zero, not at the moment of inversion — because the un-inversion usually comes via the Fed cutting the short end in response to weakening data.

Adoption, debate & evidence

This is a genuinely respected indicator, not folklore. The headline track record, from Federal Reserve research: an appropriately defined inversion has preceded every U.S. recession since the late 1960s — eight of the last eight — with essentially one false alarm in over fifty years (NY Fed; Brookings; CAIA). That is an unusually clean record for any single economic variable.

But the honest caveats are large and have grown louder:

• Variable, unreliable lead time. "Recession follows" is well supported; when is not. Lead times have ranged widely, making the signal nearly useless for timing.
• The 2022–2024 episode is the cautionary tale. The 10y-3m curve inverted on 25 October 2022 and stayed inverted until 13 December 2024 — the longest inversion in roughly 45 years — yet no recession arrived in that window; the U.S. economy grew solidly through 2023 (real GDP roughly 2.5% on a full-year basis per the BEA, ~2.9% Q4-over-Q4) with strong 2024 quarters, and the S&P 500 rose sharply (BEA; CAIA; U.S. Bank; Lombard Odier). Whether this is a true false signal or merely an unusually long lead is still debated.
• Why it may have misfired: a depressed/negative term premium means small rate moves now flip the curve, weakening the signal (Brookings). Locked-in low mortgage/corporate debt also made the 2022–24 economy less interest-rate-sensitive than past cycles.
• Inversion alone may be insufficient. Recent academic work (Reassessing the inversion…, Journal of International Money and Finance, 2024) argues inversion is a far stronger recession signal when accompanied by falling house prices and widening corporate credit spreads.

Strengths & limitations

Works best as one input in a basket of leading indicators, read for direction and rising risk over a 1–2 year horizon, and given more weight when the inversion is deep, sustained, and confirmed by credit spreads and housing.

Fails as a market-timing tool (lead time too noisy), and is degraded whenever the term premium is abnormally low or QE/policy distorts the long end. The #1 misuse is treating "the curve inverted" as a sell-equities-now signal — historically a costly mistake, since stocks have frequently risen for a year or more after inversion. A close second is fixating on 2s10s while ignoring the NY Fed's better-performing 10y-3m model.

Sources

• Brookings, Hutchins Center — The yield curve: what it is, and why it matters (term premium, caveats, shift toward shorter-maturity spreads).
• Federal Reserve Bank of New York — The Yield Curve as a Predictor of U.S. Recessions (Estrella & Mishkin) and the monthly recession-probability model (10y-3m probit).
• Federal Reserve Board, FEDS Notes (Mar 2018) — Predicting Recession Probabilities Using the Slope of the Yield Curve.
• St. Louis Fed — The Yield Curve as a Forecasting Tool (expectations hypothesis, term-premium offset).
• Wikipedia — Expectations hypothesis, Yield curve (definitions; cross-checked against Fed sources).
• CAIA, Far From Perfect (May 2024); U.S. Bank; Lombard Odier — the 2022–2024 inversion and the false-signal debate.
• Reassessing the inversion of the Treasury yield curve…, Journal of International Money and Finance (2024) — inversion + housing/credit conditioning.

Disputed/soft: exact "accuracy %" figures for short vs long inversions circulate on commercial blogs without primary backing — treated as unverified. Whether 2022–24 was a false signal or a long lead remains genuinely contested.