Why the curve slopes upward
A Treasury yield is the price of parking money for a fixed term. Longer terms normally pay more — lenders demand a term premium for locking money up through more inflation surprises, more rate cycles, more unknowns. So the resting state of the curve is upward-sloping: 3-month bills yield less than 10-year notes.
The short end is pinned by the Federal Reserve's policy rate. The long end is set by the market's expectations of growth, inflation, and future policy over the next decade. The spread between them — commonly 10-year minus 3-month or 10-year minus 2-year — compresses when the Fed tightens faster than the market believes the economy can bear.
What inversion means
An inverted curve says the bond market expects short rates to be lower in the future than they are today — which, historically, means it expects the Fed to be cutting. And the Fed cuts aggressively for one reason: the economy is rolling over. That's the causal chain behind the famous track record: the 10y−3m spread inverted ahead of every US recession since the late 1960s, typically with a 12–18 month lead, with essentially one debated false positive (1966). The academic work formalizing this (Estrella & Mishkin's probit models at the New York Fed) made the 10y−3m spread the canonical recession-probability input.
Reading it in practice
- Depth and persistence matter. A one-week dip a few basis points below zero is noise; months below zero at meaningful depth is the historical signal.
- The re-steepening is the late-cycle tell. Recessions have typically begun not while the curve is inverted but after it un-inverts — specifically on a bull steepening, where short rates collapse as the Fed starts cutting into weakness. Un-inversion is not the all-clear; it's often the acceleration point.
- It is not an equity-timing tool. The lead time is long and variable, and equities have often rallied substantially between inversion and the eventual downturn. It sets the macro regime context, not the entry.
The 2022–24 inversion — and its lesson
The inversion that began in late 2022 became the deepest and longest 10y−3m inversion in the modern record — and the recession it “predicted” did not arrive on schedule. Candidate explanations: post-pandemic excess savings and fiscal flows blunted rate sensitivity; the long end was distorted by QE/QT balance-sheet effects; and much of the tightening bit sectors (housing, banks) without breaking employment. The honest takeaways cut both ways: the sample of modern recessions is small (n ≈ 8), so “never a false positive” was always a fragile claim — and at the same time, long-and-variable lags mean a signal isn't refuted just because the clock ran long. Treat the spread as a regime input with a strong but imperfect record, not a prophecy.
Caveats
- Which spread you use matters. 10y−2y inverts earlier and more often (more false signals); 10y−3m is the better-studied recession input and inverts closer to the event.
- Central-bank balance sheets distort the long end — term premium has been compressed for much of the QE era, making inversion mechanically easier than in the 1970s–90s samples.
- Pair it with faster confirmations: credit spreads and volatility term structure react in weeks, not quarters.