Every yield curve inversion since 1980: how long, how deep and what followed

A yield curve inversion is a stretch of time when a short-term Treasury yield sits above the 10-year yield, and it is widely watched as a recession warning. This study lists every inversion of the two spreads most often used to measure it, the 10-year minus 2-year and the 10-year minus 3-month, from daily data published by the Federal Reserve Bank of St. Louis through October 30, 2026, and checks each episode against recession dates set by the National Bureau of Economic Research (NBER). The inverted yield curve explained covers how to read the curve and why an inversion is taken as a warning; this page is the episode-by-episode record. Related guides and studies are in the Treasuries hub.

Key findings

• Usually followed, not always. The 10-year minus 2-year spread inverted 10 times from 1980 to October 2026, and 7 of those episodes were followed by a recession within two years. The 10-year minus 3-month spread inverted 9 times from 1982, with 6 followed.
• No recession without an inversion first. Each of the six recessions since 1980 came after a 10-year minus 2-year inversion, with the first warning 6 to 24 months before the NBER peak. Each of the four since the 3-month data begin came after a 10-year minus 3-month inversion, 11 to 23 months ahead.
• Six episodes had no recession within two years. They were a February 1982 dip that came during a recession already under way, two short and shallow inversions in 1998, and the three episodes that began in 2022.
• Two 2022 episodes set the records for length. The 10-year minus 3-month inversion lasted 1,095 days, from October 18, 2022 to October 16, 2025, with 621 inverted trading days; the 10-year minus 2-year inversion lasted 793 days, from July 6, 2022 to September 5, 2024.
• Depth was no guide. The deepest 10-year minus 3-month close, −1.89 points on May 4, 2023, had no recession within two years, while a dip to −0.04 points in the 10-year minus 2-year spread in August 2019 came six months before the February 2020 peak.
• Both spreads were positive when the data end. On October 30, 2026 the 10-year minus 2-year spread stood at 0.38 percentage points and the 10-year minus 3-month at 0.78.

How were inversion episodes defined?

Each episode runs from the first daily close below zero to the last negative close before the spread goes more than 90 days without another.

Spreads rarely cross zero cleanly. They hover near it for weeks, slipping below and climbing back, and treating every crossing as a new inversion would break one long episode into many short ones. A 90-day gap keeps on-and-off negative days together while still separating inversions that belong to different phases of a cycle.

Length and days inverted measure different things. Length counts calendar days from the first inverted day to the last, weekends and holidays included, while days inverted counts only trading days that closed below zero, so subtracting one from the other does not give the days the curve was positive. Read together, they show how steadily a spread stayed negative: the 10-year minus 2-year episode of 2005 to 2007 lasted 526 days but had only 238 inverted days, because the spread kept returning to zero or above.

Recession dates come from NBER's Business Cycle Dating Committee, which identifies the peak and trough of each business cycle. NBER says a recession starts in the month after a peak and ends in the month of the trough, and FRED's USREC indicator follows that convention by flagging recession months from the month after each peak. This study names each recession by its peak month, the last month of expansion. A row reading “Began Jan 1980” therefore points to a contraction whose first month was February 1980, and every lead time would be one month longer if measured to that first month.

The daily 10-year minus 2-year series begins on June 1, 1976 and the 10-year minus 3-month series in January 1982. The 10-year minus 2-year episode that began in August 1978 is included because it was still running in 1980.

How many times has the yield curve inverted since 1980?

Nineteen times across the two spreads: 10 episodes of the 10-year minus 2-year spread since 1980 and 9 of the 10-year minus 3-month since 1982.

The episodes bunch into seven periods: 1978 to 1982, 1988 to 1990, 1998, 2000 to 2001, 2005 to 2007, 2019 to 2020 and 2022 to 2025. Five of those periods were followed by a recession inside the two-year window, and the first of them by two, after the January 1980 and July 1981 peaks. The two periods without one were 1998 and 2022 to 2025.

The two spreads usually inverted in the same period but rarely on the same day. The 10-year minus 2-year went first in 1988, 1998, 2000, 2005 and 2022, sometimes by months: it first inverted on December 13, 1988, more than three months before the 10-year minus 3-month did on March 27, 1989. In 2019 the order flipped. The 10-year minus 3-month spread inverted on March 22, more than five months before the 10-year minus 2-year spread's three-day dip from August 27.

The days-inverted column shows how firmly each spread stayed below zero. The episodes of 1978 to 1980 and 1980 to 1982 had 423 and 400 inverted days, and those that began in 2022 had 539 for the 10-year minus 2-year spread and 621 for the 10-year minus 3-month. At the other end are short returns below zero just before a recession, such as March 1990, with 16 inverted days four months before the July 1990 peak, and the 10-year minus 3-month episode from January 31, 2020, with 13 inverted days a month before the February 2020 peak.

How long before each recession did the curve invert?

Between 6 and 24 months, counting from the first 10-year minus 2-year inversion before each of the six recessions since 1980 to that recession's NBER peak month.

A warning that can arrive anywhere from half a year to two years ahead says little about timing. The 10-year minus 3-month range, 11 to 23 months, is narrower, but it rests on four recessions. Counting every followed episode, including later ones such as January 2020, the leads across both spreads run from 1 to 24 months.

Several recessions drew more than one episode. The 10-year minus 2-year spread inverted again in a 22-day episode in March 1990, four months before the July 1990 peak. The 10-year minus 3-month spread inverted twice before December 2007 and twice before February 2020, the second time on January 31, 2020, one month before the peak. That is why the key numbers count episodes rather than recessions: the 7 followed 10-year minus 2-year episodes point to six recessions, and the 6 followed 10-year minus 3-month episodes point to four.

None of these peaks was known while the inversions were under way. NBER names turning points well after the fact: it announced January 1980 as a peak on June 3, 1980, December 2007 on December 1, 2008 and February 2020 on June 8, 2020. The 2020 recession also had an unusual trigger. When the Fed cut its target range to 0 to 1/4 percent on March 15, 2020, its statement said the coronavirus outbreak had harmed communities and disrupted economic activity in many countries, including the United States. The 2019 inversions came months before that statement, so the record scores them as followed by a recession without showing that the curve foresaw what caused it.

Which inversions had no recession within two years?

Six of the 19 episodes: a February 1982 dip in the 10-year minus 3-month spread, both 1998 inversions and the three episodes that began in 2022.

The 1982 case comes from where the data begin. FRED's daily 10-year minus 3-month series starts in January 1982, in the middle of the recession NBER dates from a July 1981 peak to a November 1982 trough. The February 1982 inversion happened while that contraction was under way, and because the rule looks only for the next peak, the nearest one it finds is July 1990, more than eight years later.

Both 1998 inversions were shallow and brief. The 10-year minus 2-year spread closed below zero on 27 trading days between May 26 and July 27, never by more than 0.07 points, and the 10-year minus 3-month spread on 5 days between September 10 and October 5, at most 0.13 points. The second came during market turmoil. In August 1998 Russia devalued its currency and stopped payments on its debt, and investors moved toward safer and more liquid assets, according to Federal Reserve History. On September 23, 14 banks and brokerage firms put $3.6 billion into the hedge fund Long-Term Capital Management in a rescue the Federal Reserve Bank of New York brought together without lending its own money. The Fed then cut its federal funds rate target on September 29, October 15 and November 17, 75 basis points in all; the first statement cited weakness in foreign economies and less accommodative financial conditions at home. The next NBER peak came in March 2001, close to three years after the first 1998 inversion and outside the two-year window. By then the 10-year minus 2-year spread had inverted again, from February 2, 2000, and that episode does count.

The episodes that began in 2022 are the misses that stand out, because two of them were the longest in the record and the 10-year minus 3-month one was the deepest in its series. The Fed had begun raising its target range on March 16, 2022 and took it to 5-1/4 to 5-1/2 percent on July 26, 2023. Its first cut, of half a percentage point to 4-3/4 to 5 percent, came on September 18, 2024, 13 days after the 10-year minus 2-year spread's last inverted day on September 5. NBER has dated no business-cycle peak since February 2020: its latest announcement, on July 19, 2021, set the April 2020 trough, and FRED's USREC indicator still read zero for September 2026. The two-year windows for the three episodes closed in April, July and October 2024.

Taken together, this is the record of a signal with false positives. Counting episodes, 3 of the 10 for the 10-year minus 2-year spread and 3 of the 9 for the 10-year minus 3-month had no recession within two years. Counting periods instead, and setting aside the 1982 quirk and the 93-day split in 2022, the false alarms come down to two, 1998 and 2022, against five periods that did come before recessions. Either way the sample is small. There have been six recessions since 1980, so a single extra hit or miss would move every share sharply.

How did the 2022 inversions compare with earlier episodes?

The July 2022 and October 2022 episodes ran longer than any earlier one, and the 10-year minus 3-month spread went deeper than at any point since 1982.

In the monthly averages, the 10-year minus 2-year spread stayed below zero for 26 months in a row, from July 2022 to August 2024; the next-longest run since 1980 was 14 months, from September 1980 to October 1981. The 10-year minus 3-month average was negative for 25 months, from November 2022 to November 2024, and sat at exactly zero in December 2024.

The daily record gives the same ranking. The 10-year minus 3-month episode's 1,095 days and the 10-year minus 2-year episode's 793 days are the two longest in the table, ahead of the 673-day 10-year minus 2-year inversion of 1980 to 1982, and their 621 and 539 inverted days are the most of any episode.

Depth tells a different story for each spread. The 10-year minus 3-month low of −1.89 points on May 4, 2023 is the deepest in its record, almost twice the −0.96 of February 1982. The 10-year minus 2-year low of −1.08 on July 3, 2023 was shallower than the −2.41 reached on March 20, 1980 and the −1.70 of December 17, 1980. In monthly averages, that spread bottomed at −2.14 in March 1980, against −0.93 in July 2023.

Length and depth did not sort warnings from false alarms. The shallowest 10-year minus 2-year episode, three days in August 2019 with a low of −0.04, came six months before the February 2020 peak, while a nearly identical two-day dip to −0.05 in April 2022 had no recession within two years, and neither did the deepest 10-year minus 3-month episode. A table of 19 episodes cannot say whether a long inversion that ends without a recession reflects a recession still to come, a changed economy or a signal that misfired, and this study does not try to.

What can daily spread data not tell you?

It cannot explain why a recession followed an inversion, how severe the recession was, or whether the pattern will hold.

• Six recessions make a small sample. Each hit or miss moves the shares a lot, and testing two spreads under several possible rules invites picking whichever looks best after the fact.
• The rules shape the counts. A gap shorter than 90 days would split more episodes, and a three-year window would turn 1998 into a hit, since the March 2001 peak came less than three years after both 1998 inversions. Monthly averages would erase the shortest dips: the 10-year minus 2-year spread averaged 0.06 points in August 2019 and 0.21 in April 2022, months that each contain daily inversions.
• Hindsight is built in. An episode's end is known only after 90 days without a negative close, and NBER dates peaks months later, so the table shows what could be known only afterward.
• These are curve readings, not trades. Constant maturity yields come from a par yield curve that Treasury fits to indicative bid-side quotes the New York Fed collects at or near 3:30 p.m. each trading day, not to actual transactions. A close a few hundredths of a point below zero is a reading from that curve, not a price anyone traded at.
• Other sources date inversions differently. Many charts use monthly averages or a different 3-month rate. The inverted yield curve explained works from the New York Fed's monthly series, so its start dates and lows differ slightly from those here.
• What followed is not what caused it. The table records which recessions came after an inversion, not why, and the recessions since 1980 had different triggers, including the coronavirus outbreak in 2020.
• US Treasuries and recessions only. It says nothing about other countries' yield curves, about stock or bond returns after an inversion, or about how deep each recession became.

What does this record mean for someone watching the curve?

It shows a signal that came before every recession since 1980 but also appeared when no recession followed within two years.

In this record no recession arrived without an inversion first, yet 6 of the 19 episodes passed without a recession within two years, and lead times ran from 1 to 24 months. Past performance never predicts future results, and a pattern built on six recessions is no exception. The history says an inversion has been worth noticing; it does not say when, or whether, a recession will follow the next one.

For savers, the practical side of the curve is simpler than the forecasting debate. While the curve is inverted, short-term bills yield more than longer notes; when it turns positive, as it was on October 30, 2026, longer maturities yield more but carry more price risk if rates rise. Short-term vs long-term Treasuries sets out that trade-off, and the study of T-bill yields vs inflation shows how often rolling bills kept up with prices, including through the 2022 to 2024 inversion.

Fed policy, inflation expectations and the term premium all move the two ends of the curve, as what moves Treasury yields explains, and the Treasury yield curve today page shows the latest par yields and spreads. For how stocks behaved once the Fed started cutting, including in cycles with and without a recession, see the S&P 500 after Fed rate cuts study.

How to cite or reuse the inversion data

• The episode table, one row per inversion, downloads as yield-curve-inversions-history.csv. It is licensed under CC BY 4.0, so it can be reused, commercially or not, with credit and a link to this page.
• Quote counts with their method. The 90-day episode rule, the 24-month recession window and the October 30, 2026 end date all change the numbers, and a source that uses monthly averages will count differently.
• For the 2022 episodes, say that no NBER-dated recession followed within two years, as of October 2026, rather than that no recession followed at all.
• Suggested credit: Source: Sonic Wallet analysis of FRED daily Treasury spreads (T10Y2Y and T10Y3M) and NBER business-cycle dates, data through October 30, 2026.
• How sources are chosen and corrections are handled is set out in the editorial policy.