The Yield Curve Un-Inverted 25 Months Ago. Here's What Happened After Every Prior Un-Inversion Since 1978

Correction October 31, 2026: The current episode’s permanent un-inversion date is corrected from August 27, 2024 to September 6, 2024 — after the first crossing on August 27 the spread dipped back to -0.04 on September 3 and -0.02 on September 5, 2024, and only then stayed non-negative — so the 2022-24 inverted stretch runs about 26 months, not 25. The 1978-80 low is corrected from -2.28 to -2.41 points, reached on March 20, 1980. Source: Federal Reserve Bank of St. Louis (FRED), series T10Y2Y, daily.

The gap between the 10-year and 2-year Treasury yield has closed at zero or above on every trading day for nearly 25 months now, its last negative close having been September 5, 2024. That "un-inversion" gets far less attention than the inversion that preceded it, but a look at the four comparable episodes since 1978 shows the un-inversion date, not the inversion itself, is the marker analysts have historically watched most closely. This piece lays out when each prior un-inversion happened, how many months later a recession officially started, and where the current 25-month stretch sits against that history.

Key takeaways

The 10-year minus 2-year Treasury spread has closed at zero or above on every trading day since September 6, 2024 — nearly 25 months — with no NBER recession declared.
In the three most recent prior episodes the recession began 3 months (March 2001), 6 months (December 2007) and 5 to 6 months (February 2020) after the curve un-inverted.
The 1978-80 episode ran the other way: the recession started in January 1980 while the curve was still inverted, as low as -2.41 points on March 20, 1980, and the spread did not turn positive until May 2, 1980.
The spread closed at +0.31 on September 24, 2026, against a 2026 high of about +0.74; the NBER’s most recently dated peak is still February 2020.

Figures checked: September 25, 2026.

What "un-inversion" means, and why it gets less attention than it should

The 10-year minus 2-year Treasury spread (ticker T10Y2Y) is one of the most-cited recession indicators because an inverted curve — short-term yields above long-term yields — has preceded every U.S. recession since the 1970s. But the curve doesn't stay inverted forever, and the historical record shows the moment it flips back to positive, not the moment it first goes negative, tends to sit much closer to the actual start of a recession. That is what this piece tracks: the date each inversion episode permanently ended, and how that lined up with the next NBER-dated recession.

According to daily data from the Federal Reserve Bank of St. Louis (FRED, series T10Y2Y), the spread closed at +0.31 percentage points on September 24, 2026, roughly in the middle of its 2026 range (it ran from about +0.74 in early February down to +0.20 on September 21 before ticking back up). It has not closed negative even briefly since September 5, 2024, when it printed -0.02. The first crossing back to zero came on August 27, 2024, but short dips followed on September 3 (-0.04) and September 5, so the unbroken non-negative run dates from September 6, 2024.

Five episodes since 1978, lined up against NBER recession dates

The table below combines two primary sources: the daily T10Y2Y spread from FRED, and the official U.S. business-cycle peak dates published by the National Bureau of Economic Research (NBER), last updated by the NBER's Business Cycle Dating Committee on March 14, 2023. "Permanent un-inversion" is defined here as the last date the spread crossed from negative to zero or positive and then stayed non-negative for an extended run, based on the daily closing series.

Inversion episode	Permanent un-inversion date	Next NBER recession start (peak month)	Gap
~Aug 1978 – May 1980 (roughly 21 months negative)	May 2, 1980	January 1980	Recession had already started about 4 months before the curve un-inverted
~Feb 2000 – Dec 2000	December 29, 2000	March 2001	Recession started about 3 months after un-inversion
~Jan 2006 – Jun 2007	June 6, 2007	December 2007	Recession started about 6 months after un-inversion
~May 2019 – Aug 2019	August 30, 2019	February 2020	Recession started about 5-6 months after un-inversion
~Jul 2022 – Sep 2024 (roughly 26 months negative, the longest stretch in this series)	September 6, 2024	None declared as of September 25, 2026	Nearly 25 months and counting with no NBER recession call

Every date in the "permanent un-inversion" column comes directly from the FRED daily series, checked for the last sustained negative-to-positive crossing before each subsequent recession (or, for the current episode, before today). Every recession-start date comes directly from the NBER's own published chronology, which classifies the peak month as the final month of the preceding expansion.

The two cleanest cases: 2001 and 2007-09

The 2000 and 2006-07 episodes are the ones most often cited as the "textbook" pattern. The curve un-inverted on December 29, 2000, and the NBER later dated the next recession's start to March 2001 — a three-month gap. The curve un-inverted again on June 6, 2007, and the NBER dated that recession's start to December 2007 — a six-month gap. In both cases, the un-inversion arrived after the curve had already spent more than a year inverted, and the recession followed within two quarters. This is the basis for the popular shorthand that a steepening curve, not a flat or inverted one, is the more immediate warning.

The 2019 episode is similar in timing — the curve un-inverted for good on August 30, 2019, and the NBER later dated the following recession's start to February 2020, a five-to-six month gap — but it comes with an obvious caveat: that recession was triggered by the COVID-19 pandemic, an exogenous shock, not by the kind of credit or investment slowdown the yield curve is usually thought to be forecasting. Treating it as confirmation of the pattern requires some caution.

The exception: 1978-1980 broke the pattern

The oldest episode in this data (FRED's T10Y2Y series begins in June 1976) runs the timing backward. The curve inverted around August 1978 and, unlike the other four episodes, stayed inverted almost continuously — reaching as low as -2.41 points on March 20, 1980 — for about 21 months. The recession itself started in January 1980, according to the NBER, while the curve was still deeply inverted. The spread did not turn positive until May 2, 1980, by which point the recession was already four months old. In this one case, the un-inversion was a lagging signal of a recession already under way, not a leading one. A five-observation sample cannot support a firm rule, and this episode is the reason for treating the more recent three-to-six-month pattern as a tendency rather than a law.

Where the current, 25-month-old episode stands

The present un-inversion is already the longest-running of the five by a wide margin. Two years and nearly three weeks have passed since September 6, 2024, with no NBER recession call, compared with three, six, and roughly six months in the 2001, 2007-09, and 2019-20 episodes, respectively. Only the 1978-1980 case ran longer without the "expected" post-un-inversion recession showing up quickly — and in that case, the recession had, in fact, already started before the curve un-inverted at all. Neither historical pattern currently matches what has happened since September 2024: this is not a case of "the recession already started and the curve is catching up" (the way 1980 played out), nor a case of "recession within six months" (the way 2001, 2007-09 and 2019-20 played out). It is simply an unresolved fifth data point, and the honest reading of a five-observation history is that it does not yet fit either precedent.

It's also worth being precise about what "no recession" means here: the NBER's own site states its most recently dated peak is February 2020 and its most recent trough is April 2020, with no newer peak announced as of this writing. That is a data availability fact, not a forecast — the NBER dates recessions after the fact, sometimes with a lag of six months or more once one is clearly under way, so an absence of a declaration today does not rule one out having recently begun.

There is a second way to read the same span of time since the April 2020 trough: as an unusually long expansion in its own right, independent of what the yield curve has done. Roughly 77 months have passed since that trough (through September 2026), with no peak yet dated. Per the NBER's own published chronology, only four U.S. expansions since 1945 have run longer than 77 months trough-to-peak: 106 months (February 1961 to December 1969), 92 months (November 1982 to July 1990), 120 months (March 1991 to March 2001), and 128 months (June 2009 to February 2020). A long expansion does not by itself say anything about the yield curve's forecasting record — the two are separate facts drawn from the same NBER table. But it does mean the current episode is being measured against a business cycle that, on duration alone, already sits among the five longest since the Second World War, which is part of why a single un-inversion date carries less certainty here than it would in a shorter, more typical cycle.

What this indicator does and doesn't tell an investor

Four to five data points is not a statistically reliable sample, and this indicator should be read as descriptive history, not a predictive model. It says nothing about magnitude, duration, or which sectors would be hit hardest in a downturn, and it says nothing about timing more precise than "a matter of months" even in the episodes where the pattern held. The honest way to use it is as one input worth tracking alongside others — credit spreads, initial jobless claims, the unemployment rate's own trend, and corporate earnings guidance — rather than as a standalone signal to act on. The curve's flattening trend through 2026 (from roughly +0.74 in February to +0.31 by late September) is worth watching precisely because a renewed slide back toward zero, not the current positive level itself, is what would put the present episode back on the historical map. This is general market history and education, not a forecast of what the spread or the economy will do next, and it is not a personalized recommendation to buy, sell, or hold any security.

Tip: Indicators like the yield curve are easier to use well once you understand what they measure and how often they mislead — a solid stock market investing book covers that ground. Writing down what each signal made you do, and why, in an investment journal is what lets you grade those calls honestly later. (These are Amazon Associate links — we may earn a small commission on qualifying purchases.)

FAQ

When did the yield curve actually un-invert?

The spread first crossed back to zero on August 27, 2024, but it dipped negative again twice afterwards, to -0.04 on September 3 and -0.02 on September 5, 2024. The unbroken non-negative run — the permanent un-inversion by the definition used here — dates from September 6, 2024.

How long after an un-inversion has a recession usually started?

Three months in the 2001 case (un-inversion December 29, 2000; NBER peak March 2001), six months in 2007-09 (June 6, 2007; peak December 2007), and five to six months in 2019-20 (August 30, 2019; peak February 2020). The 1978-80 episode reversed the order: the January 1980 peak came about four months before the May 2, 1980 un-inversion.

Has a recession been declared for the current cycle?

No. The NBER’s most recently dated peak is February 2020 and its most recent trough April 2020, with the chronology last updated March 14, 2023. The NBER dates recessions after the fact, sometimes six months or more after one is clearly under way, so the absence of a declaration today does not rule out one having recently begun.

Where does the spread stand now?

It closed at +0.31 percentage points on September 24, 2026. Its 2026 range ran from about +0.74 down to +0.20 on September 21, so a renewed slide toward zero, rather than the current level, is what would put this episode back on the historical map.

Can five episodes predict the next recession?

No. Five observations are not a statistically reliable sample, and the record here is descriptive history rather than a model. It says nothing about the magnitude, duration or sector impact of any downturn, and nothing about timing more precise than a matter of months even where the pattern held.

Sources and methodology

Federal Reserve Bank of St. Louis, FRED, "10-Year Treasury Constant Maturity Minus 2-Year Treasury Constant Maturity" (T10Y2Y) — daily closing spread, retrieved September 25, 2026; last value used: +0.31 on September 24, 2026
National Bureau of Economic Research, "US Business Cycle Expansions and Contractions" — official peak/trough chronology, data last updated March 14, 2023
NBER, "Business Cycle Dating" — confirms the most recent peak (February 2020) and trough (April 2020) as of this writing

All spread figures are in percentage points (10-year Treasury constant-maturity yield minus 2-year Treasury constant-maturity yield), sourced directly from the U.S. Treasury Department via FRED, not seasonally adjusted, market close basis. "Permanent un-inversion" dates were identified by the writer directly from the daily FRED series as the last negative-to-positive crossing before an extended non-negative run in each episode; NBER peak dates are taken as published, with no adjustment. This article is for general market education only and is not personalized investment, tax, or legal advice; consult a qualified, licensed professional about your own situation.

Search This Blog

USStockInsights