Normally the US government pays more to borrow for ten years than for three months, much as a five-year CD usually pays more than a savings account. When that order flips and short-term Treasuries yield more than long-term ones, the yield curve is said to be inverted. Inversions draw attention because most US recessions of the past 60 years were preceded by one, but the most recent and longest inversion was not. This page explains what an inversion is, how to read the two standard spreads, what the record since 1959 shows and does not show, and what followed the 2022 to 2024 episode, with the curve's current shape shown live. The Treasury yield curve page covers the curve's other shapes, and the US Treasuries hub covers the securities.
What is an inverted yield curve?
An inverted yield curve is one where short-term Treasuries yield more than long-term ones, the reverse of the usual pattern of more pay for longer loans. The inversion can cover the whole curve or only part of it, such as 2-year notes yielding more than 10-year notes while 30-year bonds still yield the most.
The cause is usually expectations. A 10-year yield reflects the average short-term rate investors expect over the next decade plus a term premium, as what moves Treasury yields explains. When the Fed has pushed short rates high to fight inflation and markets expect it to cut later, that expected average falls below today's short rate, and the long yield can drop below the bill yield. A low or negative term premium deepens the effect.
Worked example: Why a 10-year note yielded less than a 3-month bill in June 2023
- On June 28, 2023, the 3-month par yield was 5.44% and the 10-year par yield was 3.71%, an inversion of 1.73 points.
- The New York Fed's ACM model put the average short-term rate expected over the next ten years at 4.63%, below the bill's 5.44%, because investors expected the Fed to cut eventually.
- The same model put the 10-year term premium at -0.97 point, meaning investors accepted less than the expected path of rates to hold long Treasuries.
- Expected rate cuts account for about half of the 1.73-point gap and the negative term premium for the rest. The model works with zero-coupon yields, so its parts do not match par yields to the hundredth.
How do you read the 10-year minus 2-year and 10-year minus 3-month spreads?
Subtract the shorter yield from the 10-year yield: a positive result means the curve slopes up over that stretch, and a negative result means it is inverted. Spreads are quoted in percentage points or in basis points, where 100 basis points equal one point, so -0.50 and -50 basis points mean the same thing.
The 10-year minus 2-year spread is the one most quoted in markets, because the 2-year captures expected Fed policy over the next two years. The 10-year minus 3-month spread is the one the New York Fed uses in its recession model, which takes monthly averages and converts the 3-month bill's discount rate to a bond-equivalent yield so the two maturities compare fairly. The 3-month version reacts to the Fed's current rate, while the 2-year version reacts to the policy expected next.
- Oct 30, 2026
- Oct 30, 2025
| Maturity | Yield | A month earlier | A year earlier |
|---|---|---|---|
| 1 month | 4.06% | 4.02% | 4.05% |
| 6 weeks | 4.09% | 4.13% | 4.02% |
| 2 months | 4.16% | 4.16% | 4.04% |
| 3 months | 4.16% | 4.20% | 3.92% |
| 4 months | 4.18% | 4.29% | 3.87% |
| 6 months | 4.25% | 4.33% | 3.79% |
| 1 year | 4.36% | 4.54% | 3.70% |
| 2 years | 4.56% | 4.88% | 3.61% |
| 3 years | 4.65% | 5.00% | 3.61% |
| 5 years | 4.71% | 5.09% | 3.72% |
| 7 years | 4.80% | 5.19% | 3.90% |
| 10 years | 4.94% | 5.29% | 4.11% |
| 20 years | 5.34% | 5.68% | 4.64% |
| 30 years | 5.31% | 5.64% | 4.65% |
| 10-year minus 2-year | +0.38 pts | +0.41 pts | +0.50 pts |
| 10-year minus 3-month | +0.78 pts | +1.09 pts | +0.19 pts |
| Maturity | Real yield (TIPS) | Nominal yield | Implied inflation (nominal minus real) |
|---|---|---|---|
| 5 years | 2.25% | 4.71% | 2.46% |
| 7 years | 2.42% | 4.80% | 2.38% |
| 10 years | 2.60% | 4.94% | 2.34% |
| 20 years | 2.92% | 5.34% | 2.42% |
| 30 years | 3.08% | 5.31% | 2.23% |
In the live table, the spread rows show both gauges for today, a month ago and a year ago; a negative sign marks an inversion. The chart's dashed line for a year ago shows whether the curve has steepened or flattened since. Daily readings flicker around zero when the curve is flat, which is why the New York Fed works with monthly averages.
The two gauges can disagree for months. In the latest cycle, the 10-year minus 2-year spread turned negative for good on July 6, 2022, while the 10-year minus 3-month spread did not follow until October 25, 2022. On the way out, the 2-year version ended its run on August 26, 2024, apart from two single days in early September, but the 3-month version stayed negative until December 12, 2024, because bill yields fell only as fast as the Fed actually cut.
How reliably has an inversion predicted recessions?
Often, but not on a schedule: since 1970 every recession followed an inversion of the 10-year minus 3-month spread, while two inversions had no recession within two years. The table lists every monthly inversion in the New York Fed's data since 1959, matched with the next recession in the NBER's dating.
| Inversion began | Months inverted | Deepest monthly spread | Next recession began | Lead, months | Model's peak probability |
|---|---|---|---|---|---|
| January 1966 | 7 | -0.49 | January 1970 | 48 | 41.1% |
| December 1968 | 13 | -0.51 | January 1970 | 13 | 41.8% |
| June 1973 | 16 | -1.59 | December 1973 | 6 | 68.1% |
| November 1978 | 18 | -3.28 | February 1980 | 15 | 93.8% |
| October 1980 | 12 | -3.51 | August 1981 | 10 | 95.4% |
| June 1989 | 5 | -0.16 | August 1990 | 14 | 33.2% |
| July 2000 | 7 | -0.70 | April 2001 | 9 | 46.3% |
| August 2006 | 10 | -0.51 | January 2008 | 17 | 41.7% |
| June 2019 | 5 | -0.36 | March 2020 | 9 | 37.9% |
| November 2022 | 25 | -1.71 | None declared | None | 70.9% |
The spread is the monthly average 10-year yield minus the 3-month bill yield, in percentage points. A recession begins in the month after the NBER's business cycle peak, so the December 1973 start follows the November 1973 peak. The 1966 row combines a one-month dip in January with a six-month inversion from September 1966; the June 2019 row includes one more inverted month in February 2020. A two-month dip of -0.03 in March and April 2025 is left out.
The model behind the last column is a probit regression the New York Fed estimated on data from 1959 to 2009. It turns each month's spread into the probability of a recession twelve months later.
N is the standard normal distribution function. A spread of zero gives about 30%, which is why the model never reads as reassuring when the curve is flat, and the -1.71 spread of May 2023 gives 70.9%. Three patterns stand out in the table: the two deepest inversions, from 1978 to 1981, came during the fight against double-digit inflation; mild inversions, such as 1989's -0.16, still preceded recessions; and the two misses, 1966 and 2022, both had model readings above 40%.
What happened after the 2022 to 2024 inversion?
No recession has been declared: the longest inversion in the New York Fed's data ended in late 2024, and the NBER has not dated a new peak. The timeline shows how the episode built and unwound.
Worked example: The 2022 to 2024 inversion, start to finish
- July 6, 2022: the 10-year minus 2-year spread turned negative, reaching its deepest point of -1.08 on July 3, 2023.
- October 25, 2022: the 10-year minus 3-month spread followed, reaching -1.89 on June 1, 2023.
- May 2023: the monthly spread averaged -1.71, and the New York Fed's model turned it into a 70.9% probability of recession by May 2024.
- September to December 2024: the Fed cut its target by a full point, the 3-month yield fell and the 10-year rose, and the monthly spread turned positive in December 2024.
- 2025: daily readings dipped below zero on and off until October 16, 2025, and the monthly spread was slightly negative in March and April, then positive every month from May.
- September 2026: the monthly spread averaged 0.96, which the model, updated October 4, 2026, converts into a 12.8% probability of recession by September 2027.
Why the signal failed this time is debated, and the data support at least one part of the answer. The June 2023 decomposition above shows that expected rate cuts accounted for only about half the inversion; a term premium below zero did the rest. An inversion driven partly by investors paying up for long Treasuries says less about the economy than one driven by expectations of sharply lower rates. The other part of the inversion's message did come true: the Fed cut rates by a full point in late 2024 and by three more quarter points in late 2025, as the inverted curve had implied.
The record also argues for patience before calling the signal wrong or right. The NBER announced the December 2007 peak on December 1, 2008, almost a year after the recession began, and its most recent announcement, on July 19, 2021, dated the trough of April 2020. A recession that starts in any given month is usually confirmed well after the fact.
Why has inversion tended to come before recessions?
Because inversions usually mark the late stage of Fed tightening: short rates are high enough that markets expect the economy to slow and the Fed to cut. The curve is recording a forecast that the policy rate is unsustainably high, and in most past cycles that forecast came true through a downturn.
Two mechanisms add to the expectations story. Banks fund themselves at short rates and lend at longer ones, so an inverted curve squeezes the margin on new loans and can make credit harder to get. And investors who expect weaker growth buy long Treasuries to lock in yields, which pushes long yields down further. Neither mechanism is automatic: the 1966 inversion was followed by a credit crunch and a slowdown in 1967 that stopped short of a recession.
What does an inverted curve mean for savers?
It means short Treasuries pay more than long ones for now, which favors bills only while rates stay high, because bill yields fall first when the Fed cuts. In June 2023 a 3-month bill paid 5.44% against 3.71% for a 10-year note; by December 17, 2025, after the Fed's cuts, the 3-month par yield was 3.64%, below what the 10-year had locked in two and a half years earlier.
That is reinvestment risk, the trade-off examined in short-term vs long-term Treasuries. An inversion makes staying short look free, and for a while it is. Spreading money across several maturities, as a T-bill ladder does, avoids betting everything on one rate path. Current bill yields by term are on today's T-bill rates, the Treasury bill calculator shows what a given rate pays, and how Fed rate decisions affect T-bill rates shows how quickly bills follow a cut.
What are the risks of over-reading the yield curve?
The biggest is treating it as a timer: lead times have ranged from 6 to 17 months, and some inversions have produced no recession at all.
- Timing. Even when a recession follows, the curve gives no date; the lead has been as short as 6 months and as long as 17.
- Term premium distortions. When long yields are held down by a negative term premium, as in 2023, the curve can invert more deeply than expectations alone justify.
- Shocks the curve cannot see. The 2020 recession followed the 2019 inversion, but it was caused by a pandemic that no bond market forecast.
- Small samples. The table above has eight recessions and two misses; a model fitted to so few events can change its message after one more.
- Gauge choice. The 10-year minus 2-year and 10-year minus 3-month spreads can disagree for months, and daily readings flicker where monthly averages do not.
- Late confirmation. The NBER dates recessions months after they begin, so neither a hit nor a miss is clear in real time.
The curve is best read as a summary of what bond investors expect from the Fed and how much they want to be paid for time, which is useful in itself. Figures on this page are dated history and model output, not forecasts.