Yield Curve De-Inversion: From Below or From Above, and What Came Next

An inverted yield curve turns positive in one of two ways: short rates fall faster than long rates, or long rates rise faster than short ones. Since 1978, both routes have preceded or accompanied a recession, until 2024.

The September 2024 de-inversion ended the longest inversion on record. Earlier cases help place it, provided nobody asks them for a forecast.

Six exits since 1978

The gap between the 10-year and 2-year Treasury yields (series T10Y2Y) turns negative when short rates exceed long rates. Since 1976 it has gone through six inversions lasting more than 150 sessions, counting breaks of under two months as part of the same episode. Shorter episodes, such as those of 1998 or August 2019, lasted days or weeks.

The latest was also the longest. The spread stayed below zero for 537 consecutive sessions, from July 6, 2022 to August 26, 2024, then for two more sessions through September 5. Its low of −1.08 points on July 3, 2023 did not beat March 1980 (−2.41), but no episode had lasted as long, on an Eco3min calculation from daily Treasury data.

A widely held reading, drawn from the cycles of 1989 to 2007, ties recessions to the moment the curve steepens back rather than to the inversion itself. The five earlier cases partly bear that out. The 1989 exit came 8 months before the business-cycle peak dated by the National Bureau of Economic Research (July 1990), the 2000 exit 3 months before (March 2001), the 2007 exit 6 months before (December 2007). In 1980 and 1982 the curve only turned positive once the recession was under way.

Un-inverting is not the all-clear.

From below or from above

How the spread gets back to zero can be measured by splitting the move, from the deepest point of the inversion to the exit, between the two maturities. The data show two routes.

• From below: short rates fall far more than long rates. From March to May 1980 the 2-year dropped 4.33 points and the 10-year 2.05. In 1989 the figures were 1.82 against 1.41, in 2000 1.20 against 0.73, and from July 2023 to September 2024 1.19 against 0.13.
• From above: long rates rise relative to short rates. From December 1980 to July 1982 the 10-year gained 0.42 point while the 2-year lost 1.27. From November 2006 to June 2007 both rose, by 0.19 point for the 2-year and 0.37 for the 10-year.

An exit from below reflects expected cuts in the policy rate: the market prices in easing, often because it sees activity slowing. An exit from above reflects a higher required return on long bonds, whether from expected inflation, the term premium or the supply of debt. How the whole curve has shifted year by year can be followed in the Treasury curve through time.

Before 2024, both routes ended in the same place. The 2007 exit, from above, came six months before the December 2007 peak, just as the 1989 and 2000 exits, from below, had preceded the 1990 and 2001 peaks. The shape of the exit describes a mechanism; it does not say whether the economy will follow.

The market moves before the Fed

In three of the four episodes since 1989, the curve steepened back before the Federal Reserve cut its policy rate. The December 28, 2000 exit came six days ahead of the January 3, 2001 cut. The June 5, 2007 exit came more than three months before the September 18 cut. The September 5, 2024 exit preceded by about two weeks the September 18 decision, effective the next day. Only in 1989 had the Fed started lowering its target, in June, five months before the exit.

The 2-year yield embeds the decisions expected over two years. When the market anticipates a run of cuts, it prices them in before the first one, and the spread with the 10-year closes. A de-inversion then marks the moment the bond market treats easing as a given, which makes it a signal about expectations more than about the economy. The bank-credit channel through which the inversion itself weighs on activity is set out in how an inverted curve works through bank lending.

2024: an exit with no recession so far

As of September 30, 2026, no business-cycle peak has been dated since the September 2024 exit, more than two years on (recession indicator USREC, based on NBER dates). That lag is already well beyond the 3-to-8-month window of earlier episodes. The NBER committee dates peaks months after the fact, so a peak occurring in 2026 would not yet show up in the series.

What followed took a different shape from the exit itself. In the next six months the 2-year gained 0.21 point and the 10-year 0.56: the curve steepened from above. The spread between the 10-year and the 3-month bill even inverted again for 86 sessions, from February to October 2025, again with no dated recession. On October 7, 2026 the 10-year minus 2-year spread stood at +0.51 point, with the 10-year at 5.27%.

1998 shares some features with it: a brief inversion, Fed easing (three cuts from September to November 1998), then three years of expansion before the 2001 peak. That episode is examined in the 1998 inversion and the monthly signal. It is not enough to make 2024 a lasting exception, for lack of hindsight.

What weakens the comparisons

Six episodes are a very small sample, and their settings differ: an abrupt disinflation in the early 1980s, the technology bubble in 2000, the housing crisis in 2007, an inflation shock followed by a record rate-hiking cycle in 2022. A lag window drawn from three cases is not a law, and the curve-based model described in the New York Fed recession probability model itself yields a probability, not a date.

The make-up of long rates has also changed. The term premium is the compensation for holding a long bond rather than rolling short-term paper. It was negative in 44 of 48 months from 2016 to 2019 and in 42 of 48 from 2020 to 2023, before turning positive again: 0.76 point in August 2026 in the Adrian-Crump-Moench model of the New York Fed. An exit from above can therefore reflect a premium being rebuilt rather than rising expected inflation, which changes how it reads.

Two measures track what comes next without settling it. Splitting each change in the slope between short and long rates shows whether the market expects cuts or demands a higher long-term return. The lag between the exit and any eventual peak, once the NBER rules, will show whether 2024 joins 1998 among inversions with no recession to follow.

Last updated — 8 October 2026