US Recession Probability
Multi-model system · 25 indicators · Walk-forward · Updated daily
As of September 2026, MacroRadar's model estimates a 0% probability of a US recession within the next 6 months — very low risk. The estimate combines 25 economic indicators and is walk-forward — refit at each historical step on the data available at that date — over a history reaching back 40 years, published with its misses.
Recession probability · next six months
0%
Very Low· 25 indicators across 6 economic dimensions
When the official answer arrives
The NBER has dated 6 business-cycle peaks since 1980 and announced each between 4 and 12 complete months after the peak month; the middle of the 6 was 6. The longest was the 2007-12 peak, announced 2008-12-01.
| Peak month | Announced | Months |
|---|---|---|
| 2007-12 | 2008-12-01 | 12 |
| 1990-07 | 1991-04-25 | 9 |
| 2001-03 | 2001-11-26 | 8 |
| 1981-07 | 1982-01-06 | 6 |
| 1980-01 | 1980-06-03 | 5 |
| 2020-02 | 2020-06-08 | 4 |
Dates from the NBER Business Cycle Dating Committee. Counting, ranking and dating six past announcements says nothing about a seventh.
What this means
Recession risk is low. The combination of labor market, financial, and real economic indicators is not consistent with conditions that have historically preceded recessions. This does not mean a recession is impossible — it means the current data does not support that conclusion.
The Sahm Rule indicator is at -0.03 (triggers at 0.50). Well below the trigger threshold. The labor market remains resilient.
MacroRadar Recession Probability — Model History (1985–Present)
6-month recession probability from MacroRadar's walk-forward model: each point comes from a model refit on observations up to that date, on an expanding window. It is walk-forward in estimation, not point-in-time in data — the input series and the NBER recession dating are as later revised, and vintage-level revisions are not reconstructed in this history. A point-in-time backtest exists in the code and is not what this series runs; measured and written down on 2026-08-28. Shaded peaks correspond to actual NBER recessions. 141observations. The model’s record against the two recessions in this history — the 2020 call and the 2008 miss — is published on the homepage track record.
Contributing Indicators
Yield Curve (10Y-2Y)
Inversion has preceded every recession since 1970
0.4
2026-09-02
Sahm Rule
Triggers when unemployment rises 0.5pp above 12-month low
-0.03
2026-07-01
Initial Jobless Claims
Rising claims signal labor market deterioration
203,000
2026-08-22
Consumer Sentiment
Sharp declines often precede consumer spending pullbacks
55.2
2026-07-01
Industrial Production
Sustained declines coincide with every recession
102.99
2026-07-01
High Yield Spread
Widening spreads signal rising credit stress
2.65
2026-09-01
How this model works
MacroRadar's recession model combines multiple machine learning techniques to estimate the probability of a US recession within the next 6 months.
- Monitors 25 economic indicators across employment, credit, production, consumer confidence, and financial markets.
- Detects regime structure — whether the economy is expanding, slowing, or contracting — rather than relying on any single indicator.
- Walk-forward validated on 40 years of data: refit at each step on observations up to that date. Retrained daily with the latest observations.
- Probability-calibrated: when the model says 30%, roughly 30% of similar historical readings were followed by a recession.
Most Similar Historical Periods
| Period | Similarity | Recession followed? |
|---|---|---|
| 2025-02-01 | 80% | No (within 24 months) |
| 2007-07-01 | 72% | Yes — in 6 months |
Similarity based on multi-dimensional macro state distance. Higher % = more similar conditions to today.
How recession risk affects portfolios
At current low-risk levels, broad equity exposure has historically performed well. The regime-optimized portfolio maintains a growth-oriented tilt with standard diversification.
Cite this page
MacroRadar, "US Recession Probability," https://www.macroradar.io/recession-probability (as of September 2026).
Frequently Asked Questions
What is the probability of a US recession in 2026?
MacroRadar's recession model currently estimates a 0% probability of recession within the next 6 months. The model is walk-forward — refit at each historical step on observations up to that date, with input series as later revised — and its full history is published, hits and misses alike: it was early on 2020 and it missed 2008.
How accurate is this recession model?
The model is walk-forward validated — refit at each historical step on observations up to that date, so it is always tested on periods it has never seen; its input series are as later revised rather than reconstructed to the vintages of the day. It combines multiple machine learning techniques and 25 economic indicators to produce probability-calibrated estimates, and is retrained daily. Accuracy is best judged from the published history rather than a summary: it was elevated a quarter before the 2020 downturn and did not signal 2008.
Which indicators signal recession risk?
The model monitors indicators across five channels: yield curve shape, labor market conditions (unemployment, jobless claims), consumer confidence, industrial activity, and credit stress. No single indicator is reliable on its own — the model's strength is in combining signals across dimensions.
Is the US in a recession right now?
Recessions are officially declared by the NBER, often months after they begin. MacroRadar provides a real-time probability estimate based on leading and coincident indicators, giving an earlier signal than waiting for the official determination.