US recession by end of 2027?

Logged 14 Oct 2026, 16:31 UTC

• The base rate for a recession onset in any given ~15-month window is about 18%, derived from 12 NBER recessions across 75 years; the resolution rule's alternative two-consecutive-negative-GDP trigger broadens this only slightly, since such episodes nearly always coincide with NBER recessions.
• Signal: the window is partly burned with no recession — Q1 2026 grew 2.5% and Q2 2026 grew 2.2%, so 2025 and H1 2026 are essentially resolved NO, and the effective remaining exposure is Q3 2026 through Q4 2027.
• Signal: current growth momentum is solidly positive; for YES, the earliest possible path now requires negative prints starting roughly Q4 2026, and sustained expansions growing at 2%+ rarely roll into contraction within one quarter absent a shock — this cuts the 18% base meaningfully.
• Narrative to discount: none of the retrieved facts document financial stress, yield-curve inversion, or labor deterioration, so there is no case-specific reason to price above base rate; equally, no facts justify pricing near zero, since recessions often arrive with little advance warning inside a 14-month window.
• The Q3 2026 advance estimate on Oct 29, 2026 is the next hard information point; a negative print would roughly double or triple this probability, a solid positive print would shave it further.

What would change it: A negative Q3 or Q4 2026 GDP print (starting with the Oct 29 advance estimate) or an abrupt financial/geopolitical shock would push the estimate well above 25%; conversely, continued >2% growth through mid-2027 would compress it toward 3–5%.

-10.3 probability points from the market at generation · med confidence · Polymarket

Base rate: NBER records 12 recessions in 75 years (1945–2020), roughly one onset every 6.25 years, which over the effective remaining window of ~5–6 quarters (Q3 2026–Q4 2027) implies a Poisson probability of roughly 1 − e^(−1.25/6.25) ≈ 0.18.

Chosen by hand.

Formed blind at 13.0%, without sight of the price; published at 23.3%, the market’s 33.5% moved 50% of the way toward it by fixed rule. None of the cited facts could be shown to be one the market had not already priced. That check is recorded, not enforced.