What Is an Inverted Yield Curve?

The basics

A red flag from the bond market - short-term interest rates ending up higher than long-term ones.

Normally, lending money for longer earns a higher interest rate. A two-year loan should pay more than a three-month loan, because more can go wrong in two years than in three months. When that gets flipped - when shorter-term Treasury rates are higher than longer-term ones - it's called an inverted yield curve. It's unusual, and historically it's been one of the most reliable warning lights in finance.

Why do investors do this? They aren't being irrational. If you believe the Federal Reserve is going to slash interest rates over the next year or two, then long-term bonds suddenly look like a deal - locking in today's yield before the Fed cuts is attractive. Investors crowd into long bonds, push their yields down, and the curve flips. The Fed usually only slashes rates when the economy is in trouble, so the inversion is essentially a market vote that trouble is coming.

The most-watched version is the 10-year Treasury yield minus the 2-year Treasury yield. When that number goes negative - meaning the 2-year is paying more than the 10-year - economists and traders call it an inversion. Every U.S. recession in modern history (going back to 1970) was preceded by this kind of inversion. The track record isn't perfect, and the gap between inversion and recession has varied from six months to almost two years, but the signal has been consistent enough that markets pay close attention.

The practical effect is real, not just symbolic. Banks borrow short-term (deposits) and lend long-term (mortgages, business loans). When the curve inverts, the spread they earn between those two activities collapses or even goes negative, so they pull back on lending. Less lending means slower business investment, fewer mortgages, less hiring. The inversion both predicts and contributes to the slowdown.

Going deeper

The 10Y−2Y Treasury spread turning negative; has preceded every U.S. recession in modern history.

Inversion is the most-watched recession indicator in finance. Two specific spreads dominate the conversation. The 10-year minus 2-year (FRED ticker T10Y2Y) is what bond traders watch in real time, because the 2-year yield prices in roughly two years of expected Fed policy and is highly liquid. The 10-year minus 3-month (T10Y3M) is what Federal Reserve staff and academic economists prefer, because the 3-month yield is very close to actual current Fed policy and historically has the cleanest recession-predictive record.

Mechanically, inversion squeezes bank net interest margins. Banks raise short-dated funding (deposits, fed funds, repo) and deploy it into longer-dated loans (mortgages, commercial real estate, auto loans, business term loans). When the funding rate exceeds the lending rate, that business stops making sense at the margin. Banks tighten credit standards, raise the bar on new lending, and the resulting credit contraction slows economic activity.

The lead time from inversion to recession has historically ranged from about 6 to 23 months. The 1989 inversion was followed by recession 18 months later. The 2000 inversion preceded the dot-com recession by about 13 months. The 2006 inversion preceded the Great Recession by roughly 22 months. The 2019 inversion preceded COVID by about 6 months - though COVID was an exogenous shock, the underlying credit picture had already been deteriorating. The 2022–24 inversion is the longest in the modern data and the deepest since the early-1980s Volcker era (the 1980 T10Y2Y inversion was deeper still, at roughly 240 bps below zero), and as of this writing, no NBER-dated recession has followed.

Depth and persistence both matter. A brief, shallow inversion (a few basis points for a few days) is statistically weaker than a deep, sustained one (50+ basis points for many months). The 2022–24 inversion was both deep (briefly over 100 bps) and very persistent (more than two years), which is why the conversation about whether it was a false positive is more nuanced than the headline suggests.

It's a signal, not a clock. Saying "the curve inverted last week, so we're going into recession" is wrong on two counts: it conflates a probability shift with a deterministic prediction, and it ignores that the Fed, the NBER, and the data themselves all lag the actual turning point. Bond traders and economists treat inversion as one important input into a broader recession dashboard that also includes credit spreads, the Conference Board Leading Economic Index, the SLOOS bank lending survey, ISM Manufacturing PMI, and unemployment-rate momentum (the Sahm rule).

Advanced detail

10Y−2Y vs 10Y−3M; depth + persistence matter; signal weaker post-QE.

The Estrella & Mishkin (1996) paper is the foundational reference. They demonstrated that the 10Y-3M spread had statistically significant predictive power for recessions at horizons of 4–6 quarters ahead, using a probit framework. Their work has been extended repeatedly by Federal Reserve Board and regional Fed staff - Wright (2006, Board), Bauer & Mertens (2018, San Francisco Fed), Engstrom & Sharpe (2018, Board), Benzoni, Chyruk & Kelley (2018, Chicago Fed). The literature consensus is that the slope spread retains predictive power, but its magnitude and timing have become harder to interpret in the post-QE regime.

The 10Y-2Y vs 10Y-3M choice matters in real-time interpretation. The 2-year yield embeds roughly two years of expected Fed funds path plus term premium. The 3-month yield is essentially current Fed funds plus a small premium. So the 10Y-2Y is more forward-looking on the front leg, while the 10Y-3M is anchored to actual current policy. Both have inverted before every modern recession; the lead-lag relationship between the two is informative - typically 10Y-3M inverts after 10Y-2Y by several months, because it requires actual Fed tightening to take the 3M up to the 10Y level. That held in 2000, 2006, and 2022, but the order reversed in 2019: the 10Y-3M inverted in May while the 10Y-2Y did not follow until August.

Depth and persistence: empirical work (Engstrom-Sharpe and subsequent) shows that deeper, more persistent inversions correlate with higher recession probabilities. A rule-of-thumb threshold of 10 consecutive trading days with the spread below zero is sometimes used as a confirmation filter. The 2022–24 inversion was roughly 550 trading days (793 calendar days) below zero on T10Y2Y, the longest in the modern data.

Why the signal has gotten harder to read post-2008: QE compressed long-end term premium, sometimes to negative. With a structurally lower term premium baked into long yields, the 10Y can sit below the 2Y just because the term premium is depressed - not because the market expects aggressive Fed cuts. Adjusting the slope for the ACM term premium gives a cleaner expectations-only spread (sometimes called the "risk-neutral" slope). The risk-neutral 10Y-2Y inverted later and less deeply than the nominal one in 2022–24.

Corroborating indicators practitioners use alongside the slope: (1) Senior Loan Officer Opinion Survey (SLOOS, Fed) net-tightening percentage above ~25% historically corroborates the recession signal; (2) Conference Board Leading Economic Index six-month annualized decline below -4%; (3) high-yield credit spreads (BofA Master II OAS) breaking above ~500 bps; (4) Sahm rule trigger (3-month moving average of unemployment rate up 0.5 pp from its 12-month low); (5) ISM Manufacturing PMI sub-45 for multiple months.

Trading implications: bond curve traders express directional views on the slope using 2s10s futures spreads, 2s10s swap-spread trades, or curve flatteners/steepeners constructed from cash bonds with DV01-matched notional. The 2022 deep-inversion trade was the bear-flattener - short the 2-year, long the 10-year - which paid as the Fed hiked. The 2024-25 trade is the steepener - short the long end, long the short end - which pays as the Fed cuts. Sizing matters because slope moves can dwarf level moves in P&L terms during cycle turning points.

The equity-market relationship is real but lagged. S&P 500 tops have historically come anywhere from a couple of months (2000) to nearly two years (2007) after the initial inversion, with the peak typically occurring during the Fed's pivot from hiking to cutting. The 2022 inversion was followed by an equity rally in 2023–24 even as recession-watchers stayed bearish - a reminder that the curve signals slowdown, not immediate market top.

Expert notes

T10Y2Y / T10Y3M; 2022-24 inversion longest on record without confirmed NBER recession.

The 2019 inversion preceded COVID - an exogenous shock - but credit conditions were already tightening before the pandemic. SLOOS net-tightening had crossed positive in Q3 2019; ISM Manufacturing had broken below 50 in Q3 2019 (August reading of 49.1); high-yield spreads had a widening scare in August 2019 before re-tightening into year-end. The pandemic accelerated and dramatically deepened a downturn that was probably already starting. Treating the 2019 inversion as a "clean kill" therefore requires care.

The 2022–2024 inversion ran roughly 550 trading days (793 calendar days) below zero on T10Y2Y - the longest in the modern data series. As of mid-2026 the NBER had not confirmed a recession during the inversion window, prompting genuine debate about whether the signal still holds under balance-sheet-expanded Fed regimes (compressed term premium, abundant reserves, an effective floor under risk assets from the Fed put). Two camps: (1) the structural-break camp argues the post-2008 regime fundamentally changed transmission and the slope signal no longer cleanly maps to recession; (2) the patience camp argues NBER dating typically lags by 12+ months and a 2024–25 recession could still be backdated.

The Engstrom-Sharpe near-term forward spread (NTFS) is the leading academic alternative. Defined as the implied 6-quarter-ahead 3-month rate minus the current 3-month rate, NTFS captures whether the market is pricing imminent Fed cuts. Engstrom & Sharpe (2018, FEDS Working Paper 2018-055; see also their June 2018 FEDS Note "(Don't Fear) The Yield Curve") argue NTFS has cleaner predictive power than the 10Y-3M slope. NTFS inverted in late 2022 alongside the traditional slope measures. Fed staff use NTFS as part of their internal recession-probability dashboard.

Cleveland Fed publishes a probit-based recession probability model using the slope (https://www.clevelandfed.org/indicators-and-data/yield-curve-and-predicted-gdp-growth). The model spiked to ~60–70% recession probability during 2023, which would historically have been a near-certain call. The model has since been criticized for not adjusting for term-premium regime change.

NBER methodology: the NBER Business Cycle Dating Committee does not use a mechanical rule (two negative GDP quarters is not the NBER definition - that's a separate informal rule). NBER looks at real personal income less transfers, nonfarm payrolls, real personal consumption expenditures, real wholesale-retail sales, industrial production, and household-survey employment, weighing them holistically. Average lag from peak to announcement is about 6–18 months. The committee has occasionally refused to date short downturns even when GDP fell briefly.

The Sahm Rule (Claudia Sahm, Federal Reserve staff, 2019) - 3-month moving average of the unemployment rate rising 0.5 pp from its prior 12-month low - has been a more timely real-time recession indicator than the curve in recent cycles. Sahm-Rule triggered in mid-2024; the Federal Reserve subsequently cut rates 100 bps over the back half of 2024. Whether the Sahm trigger was a true recession signal or a sample-period false positive (driven by labor-supply normalization rather than demand weakness) is itself an active debate.

Foreign curves matter for global cycles. UK gilt curve, German Bund curve, Japanese JGB curve all have their own inversion histories. The synchronized global inversion of 2022–23 was the broadest in the post-1980 data, lending weight to global-recession concerns that didn't fully materialize. Cross-country yield-curve research (Estrella & Mishkin 1997; Chinn & Kucko 2015; BIS Quarterly Reviews for ongoing coverage) is the relevant literature.

For practitioners building real-time recession probability dashboards, the standard ingredient list is: (1) one or more slope measures (10Y-3M, 10Y-2Y, NTFS), (2) SLOOS net-tightening, (3) credit spreads (BofA HY OAS, IG OAS), (4) ISM Manufacturing and Services PMIs, (5) Conference Board LEI, (6) Sahm rule. No single indicator is reliable enough to ignore the others. The Conference Board's LEI itself includes the slope as one component, so be careful about double-counting.

Data provenance for this site: T10Y2Y and T10Y3M are both pulled from FRED daily. They are computed by FRED as DGS10 - DGS2 and DGS10 - DGS3MO respectively from the underlying CMT series. The published values can differ slightly from the spread you'd get from individual bonds because CMT yields are read off a fitted par curve (built from on-the-run quotes) rather than taken from any single security - see the on-the-run discussion in the Yield Curve explainer.

NBER recession bars on charts: the standard reference for shading historical recession bars is the FRED USREC monthly series. The most recent recession on file is the brief February–April 2020 COVID recession. Any bars beyond that point on charts would be premature.