Inverted Yield Curve: What Does It Really Say About an Upcoming Recession?

Between July 2022 and September 2024, the spread between US 10-year and 2-year yields stayed negative for 539 sessions, the longest inversion since the series began in 1976, with a trough at −1.08 points. Two years after it turned positive again, no recession has been dated by the NBER, the benchmark authority in the United States (data through the end of August 2026). Did the most trusted signal in macroeconomics get it wrong? We went back to the official series of the Federal Reserve Bank of St. Louis to measure what inversions have actually foretold over fifty years.

What the 10-year minus 2-year spread measures

In normal times, a government pays more to borrow for 10 years than for 2: lenders demand a premium for tying up their money longer. The curve inverts when the 2-year yield rises above the 10-year yield, which happens when the central bank has raised its policy rates sharply and markets expect it will have to cut them, in other words a slowdown. Two measures dominate: the 10-year minus 2-year spread, the most widely quoted, and the 10-year minus 3-month spread, used by the New York Fed in its recession-probability model. On September 22, 2026, the former stood at +0.25 points.

Six recessions, as many prior inversions

Since 1976, the NBER has dated six US business-cycle peaks: January 1980, July 1981, July 1990, March 2001, December 2007 and February 2020. Each was preceded by an inversion of the 10-year minus 2-year spread. For the five "classic" cycles, the lead was 17, 10, 19, 13 and 24 months, about 17 months on average. In 2020, the inversion had lasted only three sessions in August 2019, and the recession was triggered by an external shock, the pandemic: crediting the signal would be a stretch.

The record also includes two false alarms. In 1998, the spread dipped below zero for 27 sessions with no recession in the following 30 months: the 2001 recession only came after a new inversion in 2000. And the 2022-2024 episode, the longest in the series, has not led to any dated recession so far. Above all, a lead of anywhere between 10 and 24 months is far too variable to time an investment decision.

Why the signal worked, and why it may be wearing out

The mechanism is well known: when the central bank raises rates to curb inflation, credit becomes more expensive, the margins of banks, which borrow short and lend long, get squeezed, and credit supply tightens. Several explanations have been put forward for the recent cycle, none of them settled: savings accumulated during the pandemic, households and companies shielded by fixed-rate loans taken out before the hikes, a term premium kept compressed after years of central-bank bond purchases, and large public deficits that supported activity.

The real warning often comes when the curve re-steepens

One detail in the data deserves attention. In the 1989, 2000 and 2006-2007 cycles, the recession began 8, 3 and 6 months after the curve turned positive again, just as the central bank was cutting short rates in a hurry. In 1980 and 1981, by contrast, it started while the curve was still inverted. For the recent episode, the curve turned positive in September 2024: 24 months later, no recession has been dated, a clear departure from the historical pattern.

Stocks often rise between the inversion and the recession

Selling stocks at the first inversion would often have been costly. Between the first day of inversion and the start of the recession, the Nasdaq Composite gained 10.0% in 1978-1980, 13.2% in 1980-1981, 23.9% in 1988-1990, 18.4% in 2005-2007 and 18.5% in 2019-2020. The only exception is 2000: −46.4%, as the dot-com bubble started bursting before the recession itself.

An inversion therefore signals rising macroeconomic risk, not an exit point. Between the signal and the recession, markets most often kept rising for more than a year.

What it means for asset allocation

Used on its own, the yield curve provides a timeline far too vague to justify drastic moves. It is best combined with other indicators: employment (the Sahm rule, based on rising unemployment), corporate credit spreads, and leading activity indicators. Rather than selling everything, an investor can gradually reduce the risk of a portfolio that has drifted toward more equities, tighten their rebalancing discipline and check that potential losses remain bearable. The simulation tools available on /outils let you test the effect of such adjustments on a portfolio, for educational purposes.

The 10-year minus 3-month spread, the other gauge watched by the New York Fed

The 10-year minus 3-month spread tells the same story, more strongly. It stayed inverted from October 2022 to October 2025, 621 sessions, with a trough at −1.89 points, the deepest in a series that starts in 1982. In the past, its inversions preceded recessions by 16 months in 1989, 11 months in 2000, 17 months in 2006 and 11 months in 2019. It too has not been followed by any dated recession so far, and it too is back in positive territory.

In short, the inverted yield curve has been an excellent risk indicator but a very poor timing tool, and its latest episode remains, for now, a false alarm. Method: T10Y2Y, T10Y3M, USREC and NASDAQCOM series from the Federal Reserve Bank of St. Louis (FRED), daily data through September 22, 2026; recessions dated by their NBER peak; an inversion episode groups negative sessions less than 90 days apart. These historical analyses do not predict the future and do not constitute investment advice.