10Y-3M Yield Curve vs S&P 500

The 10Y-3M spread inverted on October 25, 2022 and remained negative through December 13, 2024, a 780-day stretch that ranks as the longest sustained inversion in the FRED record going back to 1981. The trough of -190 bps printed in March 2023, materially deeper than the 2s10s trough of -110 bps over the same window and consistent with the NY Fed's ranking of the 10Y-3M as the more reliable recession signal of the two. Mechanics during the inversion: the 3-month T-bill yield touched 5.50% in October 2023 (Fed funds at 5.25-5.50% upper bound), while the 10-year Treasury yield ranged between 3.30% and 5.00% before stabilizing around 4.40% through 2024. The uninversion event on December 13, 2024 occurred when the 10-year yield rose to 4.396% as the 3-month yield declined to 4.32%, both the 'bull steepener' (short rates falling faster than long rates) variant typical of Fed-pivot phases. Since uninverting, the spread has oscillated, briefly re-inverting in February 2026 before returning to positive territory at +74 bps by late April 2026.

Across the 8 prior US recessions back to 1968, the 10Y-3M spread inverted before each one with lead times ranging from 7 to 24 months, averaging roughly 12 months from peak inversion depth to recession onset. The 2022-2024 inversion is at month 37 from initial inversion (October 2022) and month 16 from uninversion (December 2024) without an NBER-dated recession, the longest such delay in the post-WWII record. Three structural arguments are circulating among macro desks. First, the AI capex cycle ($180B Alphabet 2026 capex, $135B Meta) has produced a productivity tailwind that offsets restrictive monetary policy. Second, household balance sheets entered the inversion period with locked-in 30-year mortgages near 3% from 2020-2021 refinancing, which insulated consumer cash flow from the rate-hike cycle. Third, the Cleveland Fed's text-mining model on Beige Book sentiment puts current recession probability at 24% as of November 2025, well below the 50% threshold the model historically associates with recessions within 12 months. The pattern of 'inversion without recession' is not unprecedented: 1966-1967 produced a similar non-recession inversion, but no post-war US inversion has run this long without delivering a recession or recanting via clear false-signal classification.