Yield curve inversion: what it signals and its recession record

An inverted yield curve is one where short-term Treasury securities yield more than long-term ones. Normally investors are paid extra for lending longer, so the curve slopes up. When it slopes down instead, the bond market is saying something unusual: that today's short rates are higher than the average short rate investors expect over the coming years. Historically, that has usually meant investors expected the Federal Reserve to cut, and the Fed usually cuts because the economy weakens.

That logic gave the inversion its reputation as the most reliable recession warning in economics. Every US recession since the late 1960s was preceded by an inversion of at least one standard measure. But the record also includes a false alarm in 1966, signals that arrived as much as two years early, and the longest inversion in the modern data, from 2022 to 2024, which as of October 2026 has not been followed by a recession.

This page explains how inversion is measured, walks through the full track record month by month, compares the two most quoted spreads, and lists the reasons the signal can fail. For the shapes and mechanics of the curve itself, start with The yield curve: shapes, slopes and what each end tells you.

Measuring an inversion

An inversion is not a vague shape; it is a negative number. Pick a long maturity and a short one and subtract:

Three choices change the answer, so state them whenever you quote a result.

• Which pair. 10y–3m compares a long yield with a rate the Fed controls almost directly. 10y–2y compares it with the market's view of policy two years out.
• Daily or monthly. A spread can dip below zero for two days without the monthly average ever going negative. In August 2019 the daily 10y–2y spread inverted for three trading days, while the monthly average stayed at +0.06 points.
• Which bill series. Long-history research often uses the 3-month bill on a discount basis (FRED's TB3MS), while daily charts use the constant-maturity 3-month yield. In 2025 the daily constant-maturity version dipped below zero on and off between late February and mid-September, with a low of −0.27 points on 4 April; the monthly discount-basis version stayed positive all year.

Why an inverted curve can point to recession

There are two complementary explanations, and they reinforce each other.

The expectations story. A long yield is roughly the average of expected short rates plus a term premium. If the 10-year yields less than the 3-month bill, then even before counting any premium, investors expect short rates to be lower on average over the decade than they are now. In the past, the Fed has cut rates sharply mainly when the economy was heading into or already in a recession. So an inversion is the market pricing those cuts.

The credit story. Banks fund themselves mostly short term and lend longer. When the curve inverts, the margin on new lending narrows or disappears, and lenders tend to tighten standards. Less credit means less spending and investment. On this view, an inversion doesn't just forecast a slowdown; it contributes to one.

Both stories describe the late stage of a Fed tightening cycle: short rates have been pushed high to slow inflation, and investors believe the Fed will eventually have to reverse. Whether a recession follows depends on how hard the economy is hit before it does.

The track record, 1966 to 2026

The table pairs each run of negative monthly spreads with the next recession peak as dated by the National Bureau of Economic Research (see What is a recession? How the NBER dates US recessions). "Lead" is months from the first negative month to the NBER peak month. Short gaps of positive months inside a run are merged.

Reading down the table gives an honest scorecard.

• Hits. Every one of the eight NBER recessions that began between 1969 and 2020 was preceded by a negative monthly reading on at least one of the two spreads.
• Misses by one measure. The 10y–3m monthly spread never went negative before the 1990 recession; its lowest reading in 1989 was +0.13 points in June. The 10y–2y caught it, 18 months early. In 2019 the reverse happened.
• False signals. The 1966 inversion was followed by a slowdown but not an NBER recession. The 10y–2y dipped negative for one month in 1998 with no recession after.
• The 2020 caveat. The 2019 inversion preceded the February 2020 peak, but that recession was triggered by the pandemic, which no bond market forecast. Counting it as a success flatters the indicator.
• Long and variable leads. Five to 22 months is a wide window, and the signal says nothing about how deep a recession will be.

The 2022–2024 inversion

The most recent episode is the longest in the daily data and the most instructive.

Here is what the daily data record.

• The Fed began raising its target range in March 2022. The 10y–2y spread turned negative for two days in early April 2022 and then continuously from 6 July 2022 to 26 August 2024, 537 trading days in a row.
• At its deepest, on 3 July 2023, the spread was −1.08 points. In monthly terms, July 2023 averaged −0.93 points, the most negative monthly reading since September 1981 (−1.14).
• The 10y–3m spread on a constant-maturity basis was negative from 25 October 2022 to 12 December 2024. Its daily low was −1.89 points on 4 May 2023.
• The 10y–2y spread stopped being negative on 27 August 2024 (it was 0.00 that day) and stayed at or above zero apart from two single days in early September. The Fed's first cut of the cycle took effect on 19 September 2024.
• By 28 October 2026 the spreads were +0.38 (10y–2y) and +0.87 (10y–3m).

What happened to the economy? The NBER's chronology, as of the data used here, lists no recession after the February–April 2020 contraction. Quarterly real GDP growth, at a seasonally adjusted annual rate, was low in the first and fourth quarters of 2025 (0.1% and 0.2%) but never negative in two consecutive quarters, and was 2.5%, 2.2% and 3.4% in the first three quarters of 2026. The unemployment rate was 4.2% in September 2026. On those numbers, the most famous inversion in decades has not, so far, been followed by a recession.

That does not prove the indicator is broken. It might be early; the 2006 inversion led by 22 months on the 10y–2y measure, and this one ended two years ago. Or the conditions that made inversions predictive in the past may have changed. Both readings are defensible, and the data alone can't settle it.

Un-inversion and timing

One pattern stands out in the record: recessions have usually started not while the curve was inverted but after it turned positive again. In the monthly 10y–3m data, the spread was positive again by May 2007, seven months before the December 2007 peak; it was back above zero in January 2001, two months before the March 2001 peak; and it turned positive in October 2019 before dipping once more in February 2020, the month of the peak.

The mechanism fits the expectations story. As a slowdown becomes visible, the market prices deeper cuts, front-end yields fall fastest, and the curve steepens from the short end, a "bull steepening" in the language of The yield curve: shapes, slopes and what each end tells you. Some analysts therefore treat the un-inversion, not the inversion, as the moment of highest risk. Others note that it is impossible to know in real time whether a re-steepening is the start of a recession or the end of a false alarm. The 2024 un-inversion shows the difficulty: as of October 2026, two years later, no recession had followed.

Why the signal can fail

• Term premium distortions. If the term premium on long bonds is unusually low or negative, for example because central banks or foreign reserve managers are heavy buyers, long yields can sit below short yields without any expectation of big rate cuts. A low term premium mechanically makes inversions easier to produce.
• Policy may cut without a recession. Markets can price cuts that come about because inflation falls, not because the economy collapses. An inversion then reflects expected normalization, not distress.
• Inflation regimes differ. Many of the classic inversions happened when inflation was high and the Fed was raising rates aggressively to suppress it. The relationship estimated on that history may not carry over.
• Few observations. There have been only eight recessions since 1969. Any statistic built on eight events has wide uncertainty, and the false signals matter as much as the hits.
• Self-reference. Once everyone watches a signal, policymakers and markets may react to it, which can change what it predicts.

10y–3m versus 10y–2y

Neither is strictly better. The 10y–3m spread has a longer and slightly cleaner record in the monthly data; the 10y–2y spread is more forward-looking and inverts earlier. Watching both, and looking for agreement, is more informative than trusting one.

Other recession signals work on different evidence. The The Sahm rule: a recession signal from the unemployment rate is based on the unemployment rate, which is a coincident rather than a leading indicator, and it too triggered in August 2024 without an NBER recession following, as of the data. Comparing leading and coincident signals is the subject of Leading, lagging and coincident indicators: what turns first.

How Finabase does it

Finabase doesn't plot spreads or flag inversions; there is no yield-curve or spread tool in the app. You can still check the shape yourself in seconds.

• The Treasury yields page lists live yields for 1, 3 and 6 months and 1, 2, 3, 5, 10, 20 and 30 years. Subtract the 2-year (US2Y=T) or the 3-month (US3M=T) from the 10-year (US10Y=T) to get the two spreads on this page.
• The 2-year and 10-year symbol pages chart each yield over any range, so you can see which end moved.
• For the economic backdrop, the economy news topic gathers headlines.

The spread history and the track-record table on this page were built by Finabase Research from public Federal Reserve data and the NBER chronology.

Frequently asked questions

Does an inverted yield curve always mean a recession is coming?

No. Inversions preceded each of the eight US recessions from 1969 to 2020, but there were also false signals in 1966 and 1998, and the 2022–2024 inversion has not been followed by an NBER recession as of October 2026. It is a warning that raises the odds, not a certainty.

How long after an inversion does a recession start?

In the monthly data since 1969, the gap from the first negative reading to the NBER recession peak ranged from 5 to 22 months, depending on the episode and the spread. That range is too wide to time markets with.

Which yield curve inversion matters most, 2s10s or 3m10y?

Both are widely used. The 10-year minus 3-month spread has the longer and somewhat cleaner research record; the 10-year minus 2-year is the market's favorite and tends to invert earlier. Each has missed one recession in the monthly data, so checking both is safer than relying on one.

What does it mean when the yield curve un-inverts?

It means the long-minus-short spread has turned positive again, usually because short yields fell as markets priced rate cuts. Historically, several recessions began after un-inversion rather than during the inversion, but a re-steepening can also mark the end of a false alarm, as it has so far after 2024.

Was the 2022–2024 inversion the longest ever?

It was the longest continuous 10-year minus 2-year inversion in the daily data used here (537 trading days, July 2022 to August 2024). In the monthly data, the 1978–1982 episode spanned a longer stretch, with brief positive interruptions.

Summary

An inverted yield curve means short Treasury yields exceed long ones, usually because investors expect the Fed to cut rates in the future. Measured by the 10-year minus 3-month or 10-year minus 2-year spread, an inversion came before each of the eight US recessions from 1969 to 2020, with leads of 5 to 22 months, and recessions often began after the curve turned positive again. The signal also produced false alarms in 1966 and 1998, missed one recession on each measure, and, after the deepest and longest inversion in decades from 2022 to 2024, has not been followed by an NBER recession as of October 2026. Treat it as a meaningful but imperfect piece of evidence, and always say which spread and which data you are using.