For much of 2022 through 2024, one of the market's favorite recession alarms was flashing red. The 10-year minus 3-month Treasury spread eventually reached -1.89 percentage points on May 4, 2023. If you followed markets at the time, the message often sounded much simpler than the underlying model: inverted curve, recession coming. (analysis snapshot, research source)
Then the focal forecast month arrived. May 2024 is not an NBER recession month in the chronology available through September 12, 2026. That sounds like a failure, and at that particular forecast horizon it was a miss. But the more useful question is what the miss actually tells us about one of macroeconomics' strongest historical signals. (research source, research source, analysis snapshot, research source)
First, use the right yield curve
The New York Fed model in this analysis does not use the popular 10-year minus 2-year spread. It uses the 10-year Treasury rate minus the 3-month Treasury rate and asks a specific question: what is the probability that the economy will be in recession twelve months later? (research source)
An inversion means the short rate is above the long rate, so the spread is negative. That simple shape can reflect expectations about future short rates, monetary conditions, term premia, growth, inflation, and other forces. It is useful information, not a magic causal switch. (research source, research source)
The signal earned its reputation
It is easy to mock a forecast after it misses. That would skip the interesting part. New York Fed research comparing many proposed recession indicators found that the Treasury term spread had the highest predictive power at horizons of four to six quarters, with lagged versions also helping at shorter horizons. (research source)
So the right starting point is not that the yield curve is folklore. It has serious historical evidence behind it. The mistake is upgrading a historically useful probability signal into a deterministic countdown clock. (research source, analysis snapshot, research source, research source)
The 2022-24 warning became extreme
The recent inversion was not a tiny technical dip below zero. The official FRED 10-year minus 3-month series reached -1.89 percentage points on May 4, 2023. Across the 22 available daily observations retained for May 2023, the average spread was -1.7345 percentage points. (analysis snapshot)
That gives us a clean focal month for asking what the New York Fed's maintained model would imply. It also keeps the analysis tied to the exact spread the model uses rather than substituting a more familiar yield-curve chart. (analysis snapshot, research source)
Reconstructing the May 2023 warning
Using the New York Fed's maintained September 2026 probit coefficients, the retained May 2023 average spread maps to a 71.4% recession probability for May 2024. This is a Grizzly Bulls retrospective reconstruction. It is not a claim that the New York Fed published exactly 71.4% in May 2023. (analysis snapshot)
The target matters. The model is framed around recession status twelve months ahead. It does not mean a recession is guaranteed to begin sometime during the following year. From May 2023, the focal target month is May 2024. (research source, analysis snapshot)
May 2024 arrived without an NBER recession
As of September 12, 2026, NBER's public chronology still lists February 2020 as the most recent peak and April 2020 as the most recent trough. May 2024 is therefore an expansion month in the chronology available for this research package. (research source)
At the focal twelve-month horizon, the high-probability signal missed. That is worth saying plainly. What does not follow is that the signal suddenly has no predictive information. (research source, research source, analysis snapshot, research source)
A 71% probability is not a promise
A 71.4% probability still leaves roughly 28.6% on the other side. A no-recession outcome can happen even when a probability model is useful. One realized outcome also cannot tell us whether a model is well calibrated; calibration is something you evaluate across many comparable forecasts. (analysis snapshot, research source, research source, research source)
The New York Fed has emphasized another reason for humility: recession probabilities from this kind of model carry meaningful statistical uncertainty, and recessions are rare enough that estimating the model precisely is difficult. In other words, even the probability number has uncertainty around it. (research source)
This is the part I find more interesting than declaring the curve either right or wrong. A useful signal can be noisy. A strong historical relationship can miss. Good risk management starts when we stop demanding certainty from tools that were built to measure probability. (research source, analysis snapshot, research source, research source)
Why the curve can contain information
One plausible channel runs through credit. New York Fed research argues that monetary tightening and a flatter term spread can compress intermediary net interest margins, reduce the profitability of lending, and contribute to tighter credit supply. (research source)
That does not prove that every inversion causes a recession. It gives us one reason the curve can carry economic information. The spread sits at the intersection of short-rate policy, longer-term expectations, and the incentives of financial intermediaries. (research source, research source, analysis snapshot, research source, research source)
The curve has normalized. The lesson has not.
The New York Fed's September 6, 2026 chart shows the August 2026 monthly-average 10-year minus 3-month spread back at +0.87253 percentage points, with a 13.8825% modeled recession probability for August 2027. (research source)
That current reading will age. The more durable takeaway is simpler: the yield curve can be informative without being prophetic. Treating higher recession risk as a scheduled recession was the category error. (research source, research source, analysis snapshot, research source, research source)
Sources and methodology
This article is compiled from the reviewed Grizzly Bulls research package that supports the claims above.
The measured figures are a historical snapshot as of September 12, 2026; they should not be read as a claim that the underlying coverage is unchanged today.
- 2022–24 10-year minus 3-month Treasury inversion and twelve-month recession-probability reconstruction — Grizzly Bulls analysis, as of September 12, 2026.
- The Yield Curve as a Leading Indicator — Federal Reserve Bank of New York.
- Probability of US Recession Predicted by Treasury Spread — Federal Reserve Bank of New York, as of September 6, 2026.
- 10-Year Treasury Constant Maturity Minus 3-Month Treasury Constant Maturity — Federal Reserve Bank of St. Louis / FRED.
- Business Cycle Dating — National Bureau of Economic Research.
- What Predicts U.S. Recessions? — Federal Reserve Bank of New York, as of September 1, 2014.
- How Uncertain Is the Estimated Probability of a Future Recession? — Federal Reserve Bank of New York — Liberty Street Economics, as of May 1, 2025.
- Monetary Cycles, Financial Cycles, and the Business Cycle — Federal Reserve Bank of New York, as of January 1, 2010.