The Doomboard
Every recession indicator this site holds, what it says today, and — the part nobody prints next to it — how many times it has been wrong.
Right now
0 of 6 indicators are signalling. None of them. That is not a forecast that nothing will happen — the average lead time below is measured in years, and the one indicator with a perfect record needed thirty-six months of patience to earn it.
| Indicator | Now | Right 24mo | Right 36mo | Caught | False alarms | Lead |
|---|---|---|---|---|---|---|
| 10-year minus 3-month spread since 1982 | quiet 0.89 | 63% | 75% | 4 of 4 | 2 of 8 | 17 mo |
| 10-year minus 2-year spread since 1976 | quiet 0.33 | 67% | 89% | 6 of 6 | 1 of 9 | 16 mo |
| Sahm rule since 1959 | quiet -0.07 | 25% | 27% | 3 of 9 | 8 of 12 | 5 mo |
| Chicago Fed financial conditions since 1971 | quiet -0.56 | 40% | 60% | 5 of 7 | 4 of 10 | 17 mo |
| St. Louis Fed financial stress since 1993 | quiet -0.79 | 29% | 43% | 3 of 3 | 8 of 15 | 11 mo |
| Banks tightening lending standards since 1990 | quiet 0 | 43% | 43% | 3 of 4 | 4 of 7 | 6 mo |
“Right” is precision: of the times it fired, how often a recession began inside the window. “Caught” is recall: of the recessions that happened while the series existed, how many it called. They are different questions and the famous claims are always about the second one.
10-year minus 3-month spread 1982-01-04 to 2026-09-11 · 4 recessions in that span
US recession threshold Inverted. Zero is not a tuned threshold -- the spread is inverted or it is not.
The version the Federal Reserve's own research prefers. It starts in 1982, so it is judged on four recessions rather than six -- a shorter history, not a better one.
10-year minus 2-year spread 1976-06-01 to 2026-09-11 · 6 recessions in that span
US recession threshold Inverted. Zero is not a tuned threshold -- the spread is inverted or it is not.
The famous one, and the claim usually made for it holds: it inverted before all six recessions since 1976. The other half of the record is that it has inverted nine times, one of which was followed by nothing at all.
Sahm rule 1959-12-01 to 2026-08-01 · 9 recessions in that span
US recession threshold 0.50 is Claudia Sahm's own threshold, not one chosen here.
Built to identify a recession that has already started, so its lead time is short by design.
Chicago Fed financial conditions 1971-01-08 to 2026-09-04 · 7 recessions in that span
US recession threshold Above zero means tighter than average, which is how the index is constructed.
Financial conditions, not the economy: it tightens in market events that never reach the real economy.
St. Louis Fed financial stress 1993-12-31 to 2026-09-04 · 3 recessions in that span
US recession threshold Above zero means above-average stress, by construction.
Spikes on market stress of any kind, which is why it cries wolf so often.
Banks tightening lending standards 1990-04-01 to 2026-07-01 · 4 recessions in that span
US recession threshold A net share above zero means more banks tightening than loosening.
Quarterly survey. Credit drying up is a mechanism, not just a correlate.
How each indicator is scored
A signal is a stretch where the indicator is past its threshold, with crossings back and forth inside six observations' worth of each other treated as one episode rather than several. It is scored a hit when a recession begins within the window, and a false alarm when one does not. Signals too recent to have resolved are counted separately and never as misses.
Thresholds are each indicator's own, not this site's. Zero for a spread, because a spread is inverted or it is not. Zero for the Chicago Fed and St. Louis Fed indices, because both are built so that zero is average. 0.50 for the Sahm rule, because that is Claudia Sahm's number. None were chosen by trying values and keeping whichever scored best — with eight recessions to fit, that procedure produces a beautiful table and no information.
Why there are two accuracy columns
Because the answer changes, and every dashboard that prints one number has quietly made the choice for you. The 10-year minus 2-year spread is right 67% of the time if you require the recession within two years, and 89% if you allow three.
Two signals move between those columns, and they are worth naming. The inversion that began in December 2005 was followed by the 2008 recession twenty-five months later — one month outside the shorter window, which feels harsh. The inversion of May 1998 was followed by the 2001 recession thirty-five months later, and counting that as a successful call is a considerable stretch. The wider window is not simply more patient; it also hands credit to signals that were nearly three years early.
That is the honest shape of the thing. These indicators say something is coming without saying when, and an instrument you must hold for three years to be vindicated by is not the same instrument as one that pays off next quarter.
Precision and recall are not the same claim
“The yield curve has preceded every recession” is true. It is a statement about recall — the curve inverted before all six recessions since 1976, and this page reproduces that. It says nothing whatever about how often the curve inverts and nothing happens, which is precision, and which is where it looks much more ordinary.
Reading only the first number is how a coin flip becomes a law of nature. Both are in the table, side by side, for exactly that reason.
What is deliberately not here
- No price, no index level, no drawdown. The daily price history behind this site's derived statistics is licensed and cannot be republished, so there is no chart of the market falling. What is here instead is the machinery underneath one — rates, credit, hiring, financial conditions.
- No composite doom score. Averaging six indicators with different frequencies, histories and hit rates into one number would invent precision that none of them have. They disagree; the disagreement is the information.
- No credit spreads. The high-yield spread is one of the better stress indicators and it is ICE's index; the investment-grade series rest on Moody's ratings. Both are vendor data, on the same footing as the price history. The Federal Reserve's own stress indices cover much of the same ground and are public domain.
- No forecast. Nothing here predicts anything. It is a record of what these measures did before, which is the only thing a historical record can be.
Where it comes from
Every series is published by a US federal producer — the Federal Reserve Board, the Chicago and St. Louis Federal Reserve Banks, the Bureau of Labor Statistics, the Department of Labor — and is public domain. Recession dates are the NBER's business-cycle chronology. Related: the yield curve in detail, futures positioning extremes, and short interest.