Subtract the 2-year Treasury yield from the 10-year and you normally get a positive number. Longer bonds carry more risk, so investors want to be paid more to lend for a decade than for two years. When that spread flips negative, when short rates climb above long rates, the yield curve is inverted. And every single time that's happened since 1970, a recession showed up afterward.
The spread that keeps getting it right
The record is hard to argue with. The curve inverted ahead of the recessions in 1973-75,1980,1981-82,1990-91,2001,2007-09, and 2020. The lead time swings a lot, anywhere from 6 to 24 months between the first inversion and the official start of the recession as dated by the NBER. There's one arguable false positive, a brief shallow inversion in the mid-1990s that never turned into a recession, though growth did slow noticeably around then.
Why it actually works
The mechanism isn't mysterious, but a few forces are working at once. Short-term yields track the Fed's policy rate pretty closely. When the Fed hikes to fight inflation or cool things down, the short end rises. Long-term yields are a different animal. They reflect what the market expects for growth and inflation over the next decade, so when investors think the economy is heading for trouble, they'll accept lower long yields because they figure the Fed will have to cut eventually.
So an inversion is really the market saying policy is tight enough to do damage, and the Fed will have to walk it back. The bond market is pricing in a policy mistake, or at least a setup where tightening now buys you weakness later.
There's also a direct channel through the banks. Banks borrow short through deposits and short-term funding, and lend long through mortgages and business loans. Invert the curve and that spread goes negative, which makes the core banking model unprofitable at the margin. Banks react by tightening lending standards, credit dries up, and investment and spending slow. So the inversion doesn't just predict a recession. It helps cause one.
Which spread to watch
People argue about the best spread to track. The 10-year minus 2-year is the one everyone quotes, partly for the long clean data history and partly because it picks up both signals at once, the policy read through the 2-year and the growth read through the 10-year.
Some economists prefer the 10-year minus 3-month, which is what the New York Fed uses in its recession probability model. That one hugs current Fed policy more tightly since the 3-month bill tracks the funds rate closely. Work by Arturo Estrella and others at the New York Fed found the 10Y-3M gives slightly cleaner recession forecasts on lead time and signal clarity.
In practice both spreads invert around the same window and have similar records. The 10Y-2Y usually goes first, since the 2-year is partly forward-looking and starts pricing in future cuts before the 3-month rate rolls over. Watching both gives you the fuller picture.
The timing problem
Knowing a recession is coming is useful. Knowing when is a lot more useful, and that's where the signal gets slippery. The lag from inversion to recession has run anywhere from about 6 months to over 2 years. The 2006 inversion led the Great Recession by roughly 18 months. The 2019 inversion came before the 2020 recession, though a pandemic triggered that one, which muddies the causal story.
That variable lag is a real problem if you trade on it. Sell equities the moment the curve inverts and you might sit in cash for 12 to 18 months while the market keeps ripping. Some of the best returns in a cycle land after the curve has already inverted but before the recession shows up. The S&P 500 gained over 20% between the 2006 inversion and the October 2007 peak.
So I treat inversion as a risk-management signal, not a sell button. When the curve inverts, the odds shift meaningfully toward bad outcomes. That's the moment to stress test the portfolio, pull back on leverage, add some defensive weight, and make sure you've got enough liquidity that you're never a forced seller into a downturn. It's the same discipline I lean on running scenarios across assets on Blockcircle, where the point isn't to call the top but to know how the book behaves if things break.
The re-steepening signal
There's a second signal that gets less airtime and often times better. When the curve un-inverts, going from negative back to positive, the recession is usually close. The reason is that the curve re-steepens because the Fed starts cutting into deteriorating data, which drags the short end down faster than the long end falls. By the time you get that re-steepening out of an inverted state, the damage is already in motion.
That pattern led the 2001 recession by about 4 months and the 2007-09 recession by roughly 6. It's tighter and more actionable than the first inversion, but it asks for patience, because you're waiting on the second phase of a two-phase indicator.
Where it breaks down
None of this is infallible. Structural shifts in the bond market, like heavy central bank asset buying under QE, can distort the term premium and squash long yields for reasons that have nothing to do with growth. The Fed's balance sheet since 2008 has arguably made the long end less informative than it used to be.
It's also very much a US indicator. It reads the US economy well and travels poorly to other countries, especially ones with different policy frameworks or thinner government bond markets. And even where its recession record is strong, it tells you nothing about how deep or how long the downturn runs.
Caveats aside, the curve is still one of the more reliable macro tells you've got. When it inverts, it's flagging something most other indicators won't confirm for months. The real question isn't whether to watch it, it's how you act on a signal that nails the destination and stays vague about the arrival time.