The Inverted Yield Curve: 7-of-7 Recession Track Record

The inverted yield curve is the most reliable single recession indicator in the modern US business-cycle toolkit. When the 10-year Treasury yield falls below the 2-year Treasury yield (or below the 3-month Treasury bill yield in the alternative version), the curve is inverted. On the New York Fed's monthly 10Y-3M series an inversion came before every one of the seven recessions that began after 1969, and on the 10Y-2Y series (daily data from 1976) the final inversion began about 14 months before the recession on average. But inversions have not always been followed by recession: the 10Y-2Y inverted briefly in mid-1998 with no recession for almost three years, and both spreads inverted from 2022 into 2024, the longest stretch on record, with no recession following.

As of 28 October 2026, the 10Y-2Y spread is positive at +38 basis points, having re-steepened from an inversion that ran from July 2022 to August 2024, the longest in the series, without producing a recession within the historical lead-time window. In the four recessions since 1990 the spread turned positive 2 to 6 months before the recession began; this time it turned positive at the end of August 2024 and more than two years have passed without one, so the historical timing has already been missed. The Federal Reserve Bank of New York's yield-curve recession-probability model, in its October 2026 release, puts the 12-month-ahead probability at 12.76 percent.

What the Yield Curve Measures

The Treasury yield curve plots yields on US Treasury securities against their time to maturity, from 1-month and 3-month bills at the short end through 2-year and 10-year notes to 30-year bonds at the long end. Under normal conditions, the curve slopes upward: longer-dated securities offer higher yields than shorter-dated ones, because investors demand additional compensation for the longer time horizon. The premium reflects the increased uncertainty about future inflation (a 10-year inflation forecast is less reliable than a 3-month forecast), the increased uncertainty about future Fed policy (which directly drives short-end yields), the reinvestment-risk associated with reinvesting cash flows over a longer period, and the credit-risk associated with longer-dated US government obligations (very low for Treasuries but not zero).

When the curve inverts, short-term yields exceed long-term yields. The 10-year Treasury yields less than the 2-year. This is unusual because it implies that bond-market participants are willing to lock in a longer-dated lower yield rather than continuously roll over higher-yielding short-dated holdings. The behaviour is rational only if those participants expect short-term yields to fall in the future, which historically reflects expectations that the Federal Reserve will cut the federal funds rate (the primary driver of 2-year yields) in response to weakening economic conditions or recession.

The Two Variants: 10Y-2Y and 10Y-3M

Two specific yield-spread variants are most commonly tracked. The 10-year minus 2-year (10Y-2Y) spread, published as FRED series T10Y2Y, is the most widely cited in financial-media coverage. The 10-year minus 3-month (10Y-3M) spread, published as FRED series T10Y3M, is preferred in academic research on recession prediction, particularly the work of Arturo Estrella and Frederic Mishkin at the Federal Reserve Bank of New York through the 1990s.

The 10Y-3M variant is generally considered the cleaner technical measure because the 3-month bill more directly reflects the current federal funds rate, while the 2-year Treasury reflects an average of expected Fed policy over the next 24 months. The 10Y-3M spread therefore captures the gap between the long-end (growth and inflation expectations) and the immediate Fed policy stance more directly. The 10Y-2Y variant has more historical data available (the 2-year Treasury series begins in 1976; the bond-equivalent 3-month bill series goes back earlier) and is more familiar to retail investors and financial media.

Both variants have similar historical track records: both inverted before every postwar US recession since 1969, with similar lead times. The 10Y-3M variant tends to invert slightly later in the cycle (because the 3-month bill responds more sluggishly to Fed expectations than the 2-year), and the 10Y-2Y variant tends to invert slightly earlier. Either variant captures the underlying signal.

The Seven-of-Seven Track Record

The historical record of yield-curve inversions preceding US recessions is exceptionally consistent. Dates below are the month the New York Fed's 10Y-3M monthly average went negative, with the 10Y-2Y date added where the two differ. The 1969-70 recession was preceded by an inversion that began in December 1968. The 1973-75 recession was preceded by an inversion beginning in June 1973. The 1980 recession was preceded by an inversion beginning in November 1978 (August 1978 on the 10Y-2Y daily series). The 1981-82 recession was preceded by an inversion beginning in October 1980 (and was preceded by the inversion that had already preceded the 1980 recession, with the curve remaining inverted across the brief 1980-81 expansion).

The 1990-91 recession was preceded by an inversion beginning in June 1989 (December 1988 on the 10Y-2Y, about 19 months before the recession peak). The 2001 recession was preceded by an inversion beginning in July 2000 (February 2000 on the 10Y-2Y). The 2007-09 Great Recession was preceded by an inversion beginning in August 2006 (a brief 10Y-2Y inversion came as early as January 2006). The 2020 COVID recession was preceded by an inversion beginning in June 2019 (three days in late August 2019 on the 10Y-2Y) (though the COVID recession was driven by a public-health shock rather than by the economic conditions the inversion typically signals; nonetheless, the inversion was present).

There have been two clear false signals since 1969. The first was a brief 1998 inversion around the Russian-LTCM financial crisis: the 10Y-2Y spread dipped below zero in June and July 1998, and the Fed then cut rates aggressively (the funds rate fell from 5.50 percent to 4.75 percent in three rapid cuts that autumn). The economy kept expanding for almost three more years before the 2001 recession began, well beyond the usual lead-time window. The second is far larger: the 2022-24 inversion, covered below, which ran for more than two years without a recession following.

The Un-Inversion Paradox

An empirical pattern that practitioners have long noticed and that academic research has subsequently formalised: recessions typically begin not when the yield curve inverts but when it un-inverts (re-steepens). In every postwar US recession that the yield curve preceded, the curve had inverted some months earlier, then began re-steepening as bond markets priced in Fed easing in response to weakening conditions, and the recession officially began soon after. On the 10Y-2Y series the spread turned positive two to six months before each of the four recessions since 1990; in 1980 and 1981 the curve was still inverted when the recession began.

The mechanism is straightforward. Inversion occurs when the Fed has tightened monetary policy and the bond market expects future easing in response to deteriorating conditions. Re-steepening occurs when the deteriorating conditions become evident enough that the Fed actually begins cutting (causing short-end yields to fall faster than long-end yields). The actual Fed cuts and the deteriorating conditions that prompted them are then closely associated with the official recession start that NBER subsequently dates. The 2-year Treasury yield, which closely tracks expected Fed policy, falls sharply as the cuts arrive. The 10-year Treasury yield falls less sharply because long-run growth and inflation expectations adjust more slowly. The result is a fast re-steepening that historically maps onto the recession start.

The pattern means that the recession-monitoring focus should arguably be on the un-inversion rather than the inversion itself. By the time the curve is re-steepening, the economic deterioration that the inversion forecast is becoming evident. The few months between re-steepening and the official recession start are the period in which leading indicators (Conference Board LEI, ISM PMI), coincident indicators (Sahm rule), and confirming GDP / employment data all begin pointing in the same direction.

The 2022-25 Inversion: Longest Without Recession

The 2022-25 inversion was the longest on record without producing a recession within the historical lead-time window. The 10Y-2Y inversion began in July 2022 as the Federal Reserve's aggressive 2022-23 tightening cycle pushed short-end yields up faster than long-end yields, and reached its deepest point of negative 108 basis points on 3 July 2023. The 10Y-3M followed in late 2022 and went deeper, its monthly average reaching negative 171 basis points in May 2023. The 10Y-2Y turned positive at the end of August 2024 as markets priced the Fed's first cuts; the 10Y-3M monthly average stayed negative until December 2024 and dipped below zero again briefly in March and April 2025.

The duration was unprecedented: about 26 months on the 10Y-2Y and 25 consecutive months on the 10Y-3M monthly average, against previous longest runs of about 20 months (10Y-2Y, August 1978 to May 1980) and 18 months (10Y-3M, November 1978 to April 1980). Yet NBER has dated no recession since the two-month COVID recession of 2020. The indicators NBER weighs (employment, real personal income, spending, industrial production) did not deteriorate enough to support one.

Several explanations have been offered for the unusual lag. The COVID-era fiscal and monetary stimulus had built up substantial household and corporate cash balances that buffered economies against tighter monetary conditions for longer than typical. The labour market remained tight throughout the inversion period (unemployment averaged under 4 percent for most of the inversion), suggesting the underlying economy was less interest-rate-sensitive than in prior cycles. The inversion itself was driven heavily by the Fed's tightening response to inflation rather than by deteriorating long-run growth expectations, which may have produced a less reliable recession signal than inversions driven by long-end yield declines. Whichever explanation is most accurate, the 2022-25 inversion is now an important data point in any future application of yield-curve recession indicators.

The 2026 Re-Steepening Question

The re-steepening came in mid-2024 as markets priced, and the Federal Reserve then delivered, cuts to the federal funds rate in response to the rising Sahm rule reading and softening labour-market data. The 2-year Treasury yield fell faster than the 10-year Treasury yield (the standard pattern), and the 10Y-2Y spread crossed back into positive territory at the end of August 2024. As of 19 October 2026, the spread sits at +38 basis points. It had narrowed to a +20 basis point low on 21 September after the Fed's 16 September rate increase to 3.75-4.00% pushed short-end yields up, then recovered to +51 on 7 October.

The historical pattern would have put a recession within about six months of the re-steepening, that is, by early 2025, and none came. The New York Fed recession-probability model puts the 12-month-ahead probability at 12.76 percent. Read against its own published file that is an ordinary month, not a low one: the series has averaged 15.3 percent since 1960 with a median of 9.1 percent. Nor is there a reliable pre-recession band to compare it against, because the maximum reading in the year before each postwar recession ranges from 8 percent to 94 percent, with a median near 32 percent. Notably, the other primary signals (Sahm rule 0.00, ISM PMI 54.5, jobless claims near 199,000) all sit in healthy territory, which weakens the case that this re-steepening is the prelude to a near-term recession.

With more than two years gone since the 10Y-2Y turned positive and no recession, the 2022-25 inversion already stands as the most significant false signal in the modern record, far larger than the brief 1998 dip. A recession starting in late 2026 or 2027 would not rescue it: that would be a lead time well beyond anything in the historical record. The other indicators (Sahm rule, ISM PMI, jobless claims, LEI, consumer confidence) are the better guide to near-term risk.

For the full indicator dashboard, see indicators. For the related real-time indicator, see the Sahm rule. For the probability models that incorporate yield curve data, see recession probability models.