Yield curve inversion
Last updated:
Direct answer
A yield-curve inversion happens when short-term Treasury yields exceed long-term Treasury yields — short money pays MORE than long money, the opposite of the usual term premium. The 2y-10y and 3m-10y spreads are the two most-watched: the 3m-10y has a near-perfect historical record of preceding US recessions by 6-18 months (Estrella & Mishkin 1996). The 2022-2024 inversion was the deepest since the 1980s; whether it "broke" the pattern depends on how you define recession.
Why this question matters
Under normal conditions, longer-maturity bonds pay higher yields than shorter-maturity bonds — investors demand compensation for tying up money longer (the term premium). When the curve inverts, it usually means the bond market expects the Fed to cut rates significantly in the future, which means the market expects a recession or sharp slowdown. The historical record is strong: every US recession since 1955 was preceded by a 3m-10y inversion, with the lone exception being the 1966 inversion that preceded a slowdown but not a formal NBER-dated recession. The lag from inversion to recession start has been 6 to 18 months in modern history. The 2022-2024 inversion challenged the pattern: the 3m-10y inverted in October 2022 and stayed inverted through mid-2024 — by any traditional reading, a recession "should have" started by mid-2024. As of early 2026, no formal NBER recession has been declared. Whether the indicator is broken, late, or signaling a slowdown that didn't quite hit recession status is the live debate.
Normal vs Inverted — side-by-side
| Dimension | Normal | Inverted | Note |
|---|---|---|---|
| Curve shape | Upward-sloping (long > short) | Downward-sloping (short > long) | Normal reflects term premium; inversion reflects expected Fed cuts |
| Most-watched spread (1) | 10y - 2y > 0 | 10y - 2y < 0 | Trading-desk default; 2y is the Fed-policy-sensitive end |
| Most-watched spread (2) | 10y - 3m > 0 | 10y - 3m < 0 | Estrella & Mishkin 1996 — academic recession-predictor spread |
| Implied Fed expectation | Steady or hiking | Cutting (significantly) within ~12-18 months | Long-end pricing reflects average expected short-rate path + term premium |
| NY Fed recession probability model | Low probability of recession in next 12 months | Probability rises sharply — typically 30%+ when inverted >100bps | NY Fed publishes monthly probability based on 3m-10y spread |
| Historical hit rate (1955-2007) | — | ~7/7 inversions preceded NBER recessions | Estrella & Mishkin tracked this through 2007; pattern held |
| Modern era hit rate (2007-2022) | — | 2 of 2 — 2006 inversion preceded 2008 recession; 2019 inversion preceded 2020 recession | COVID-2020 is sometimes excluded as exogenous shock |
| 2022-2024 inversion | — | Deepest since early 1980s — 3m-10y inverted Oct 2022, persisted through mid-2024 | No NBER recession declared as of early 2026 — live debate on whether indicator broke |
| Typical lag from inversion to recession | — | 6 to 18 months historically | Mean ~12 months; varies meaningfully by cycle |
| Asset-class implications | Risk-on regime — equities + credit historically outperform | Risk-off regime over 12-18mo horizon — defensive sectors + duration + cash historically outperform | But timing is hard — equities often peak well AFTER inversion onset |
| Counter-arguments to "broken indicator" (2022-2024 case) | — | Possible explanations: massive fiscal expansion masked slowdown; QT distorting term premium; immigration + labor-supply shifts; "long and variable lags" version 2.0 | Active research debate; no consensus as of 2026 |
Numbers worth remembering
- Every US recession since 1955 was preceded by a 3m-10y Treasury yield curve inversion, with the lone exception being the 1966 inversion (Estrella & Mishkin 1996; NY Fed updates).
- The historical lag from 3m-10y inversion to NBER-dated recession onset has been 6 to 18 months, with a mean of approximately 12 months.
- The 2022-2024 inversion (Oct 2022 - mid-2024) was the deepest 3m-10y inversion since the early 1980s.
- The NY Fed publishes a monthly recession probability model based on the 3m-10y spread; probability typically exceeds 30% when the spread is below -100bps.
- The 2019 inversion lasted ~5 months and preceded the COVID-2020 recession by approximately 6 months — sometimes excluded from "clean" sample as exogenous shock.
- The term-premium component of the 10-year Treasury yield was estimated near zero (or negative) through much of the 2022-2024 inversion per NY Fed ACM (Adrian-Crump-Moench) decomposition — one explanation offered for why the inversion may not have predicted a recession in the traditional sense.
How to pull this data
Computing the curve inversion signal is straightforward once the Treasury yields are in hand — subtract the short-rate from the long-rate. When the Tidore macro vertical ships, both yields are available via one API key.
# Tidore (in development, M3+ release) — pull 3m + 10y from Fed H.15
curl -H "Authorization: Bearer $TIDORE_API_KEY" \
"https://api.tidore.co/v1/macro/treasury-yield?tenor=3m&from=2000-01-01"
curl -H "Authorization: Bearer $TIDORE_API_KEY" \
"https://api.tidore.co/v1/macro/treasury-yield?tenor=10y&from=2000-01-01"
# Inversion signal: 10y_yield - 3m_yield < 0
# Today — primary source: Fed H.15 release
# https://www.federalreserve.gov/releases/h15/Authoritative sources
- NY Fed — Yield Curve as a Leading Indicator (Estrella & Mishkin foundational paper + monthly probability updates)
- Federal Reserve — H.15 Selected Interest Rates (daily 10-year, 2-year, 3-month Treasury yields)
- NBER — US Business Cycle Expansions and Contractions (official recession dating)
- NY Fed — Adrian-Crump-Moench term-premium decomposition
- Estrella & Mishkin 1996 — The Yield Curve as a Predictor of US Recessions
Frequently asked questions
Is the yield curve currently inverted?
As of mid-2025 the 3m-10y had normalized after the deepest sustained inversion (Oct 2022 - mid-2024) since the early 1980s. Check the latest values at Fed H.15 release directly — the answer changes monthly. The 2y-10y normalized earlier than the 3m-10y in 2024.
Does an inverted yield curve always predict a recession?
Historically yes for the 3m-10y spread: every US recession since 1955 was preceded by a 3m-10y inversion, with the 1966 inversion as the lone "false signal" (slowdown but no NBER-dated recession). The 2022-2024 inversion is the live debate — no NBER recession has been declared as of early 2026, which would be either a second "false signal" or a sign the lag has stretched beyond historical norms. Active research debate; no consensus.
What is the difference between 2y-10y and 3m-10y inversion?
The 3m-10y is the academic standard (Estrella & Mishkin 1996) and what the NY Fed uses in its monthly recession probability model. The 2y-10y is the trading-desk default — it's more sensitive to Fed-policy expectations (the 2y is the Fed-policy-sensitive end of the curve) but has a less clean historical record. In recent cycles the 2y-10y has tended to invert FIRST and the 3m-10y to invert LATER as the cycle matures.
How long does it take from inversion to recession?
Historically 6 to 18 months from the start of 3m-10y inversion to the start of NBER-dated recession, with a mean of about 12 months. The 2022-2024 inversion is currently testing the upper bound — Oct 2022 inversion would have implied recession start by ~April 2024 at the long end of historical lag, which has not happened.
What causes a yield curve to invert?
Two structural mechanisms. (1) Short-end pressure — when the Fed hikes the federal funds rate aggressively, short-maturity yields rise quickly to track Fed policy. (2) Long-end anchoring — long-maturity yields reflect average expected future short-rates plus a term premium; if markets expect significant Fed cuts in the future (typical pre-recession), long yields don't rise as fast as short yields, and the curve flattens or inverts. The 2022-2024 inversion was primarily mechanism #1 (Fed hiked aggressively from near-zero to 5.25%+) with mechanism #2 secondary (markets priced in eventual cuts).
Should I trade on yield curve inversion signals?
The historical record is strong directionally (recession follows) but weak on timing precision (6-18 month lag, with wide error bars). Asset-class historical patterns: equities often peak 3-6 months AFTER inversion onset, then decline through and out of the eventual recession; long-duration Treasuries often outperform during the inversion-to-recession transition; cyclical sectors underperform defensive sectors. None of this is investment advice — it's historical pattern observation, and the 2022-2024 episode is a reminder that "every previous inversion preceded a recession" does NOT guarantee the next one will.