Replay each recession
What the two forecasts showed before each NBER peak since 1980.
The alarm line is 25%. The target window runs from 11 months before the peak through the peak; the recession begins the following month. The 12-month-before snapshot is context outside that window. Lead time is measured to onset. An already-high reading is not a new crossing. Missing early history is shown as missing.
Peak 1980-01 · onset 1980-02
yield-curve benchmark
First warning in the target window: 1980-01; prior month unavailable. 1 months before onset (0 before peak). Coverage: 1 of 12 eligible months.
| Lead to peak | Month | Forecast | Context |
|---|---|---|---|
| 12 months | 1979-01 | — | |
| 6 months | 1979-07 | — | |
| 3 months | 1979-10 | — | |
| 1 months | 1979-12 | — |
12-block model
No warning observed in the available target window. Coverage: 1 of 12 eligible months.
| Lead to peak | Month | Forecast | Context |
|---|---|---|---|
| 12 months | 1979-01 | — | |
| 6 months | 1979-07 | — | |
| 3 months | 1979-10 | — | |
| 1 months | 1979-12 | — |
Peak 1981-07 · onset 1981-08
yield-curve benchmark
First warning in the target window: 1980-10; crossing. 10 months before onset (9 before peak). Coverage: 12 of 12 eligible months.
| Lead to peak | Month | Forecast | Context |
|---|---|---|---|
| 12 months | 1980-07 | 0.0% | Already in recession |
| 6 months | 1981-01 | 100.0% | |
| 3 months | 1981-04 | 100.0% | |
| 1 months | 1981-06 | 100.0% |
12-block model
First warning in the target window: 1980-09; crossing. 11 months before onset (10 before peak). Coverage: 12 of 12 eligible months.
| Lead to peak | Month | Forecast | Context |
|---|---|---|---|
| 12 months | 1980-07 | 2.8% | Already in recession |
| 6 months | 1981-01 | 23.5% | |
| 3 months | 1981-04 | 9.1% | |
| 1 months | 1981-06 | 39.4% |
Peak 1990-07 · onset 1990-08
yield-curve benchmark
First warning in the target window: 1989-08; already above. 12 months before onset (11 before peak). Coverage: 12 of 12 eligible months.
| Lead to peak | Month | Forecast | Context |
|---|---|---|---|
| 12 months | 1989-07 | 98.8% | |
| 6 months | 1990-01 | 82.9% | |
| 3 months | 1990-04 | 47.0% | |
| 1 months | 1990-06 | 69.4% |
12-block model
First warning in the target window: 1990-02; crossing. 6 months before onset (5 before peak). Coverage: 12 of 12 eligible months.
| Lead to peak | Month | Forecast | Context |
|---|---|---|---|
| 12 months | 1989-07 | 25.7% | |
| 6 months | 1990-01 | 19.0% | |
| 3 months | 1990-04 | 14.1% | |
| 1 months | 1990-06 | 7.3% |
Peak 2001-03 · onset 2001-04
yield-curve benchmark
First warning in the target window: 2000-04; already above. 12 months before onset (11 before peak). Coverage: 12 of 12 eligible months.
| Lead to peak | Month | Forecast | Context |
|---|---|---|---|
| 12 months | 2000-03 | 31.7% | |
| 6 months | 2000-09 | 96.6% | |
| 3 months | 2000-12 | 99.7% | |
| 1 months | 2001-02 | 99.5% |
12-block model
First warning in the target window: 2000-12; crossing. 4 months before onset (3 before peak). Coverage: 12 of 12 eligible months.
| Lead to peak | Month | Forecast | Context |
|---|---|---|---|
| 12 months | 2000-03 | 3.8% | |
| 6 months | 2000-09 | 11.9% | |
| 3 months | 2000-12 | 33.7% | |
| 1 months | 2001-02 | 49.9% |
Peak 2007-12 · onset 2008-01
yield-curve benchmark
First warning in the target window: 2007-01; already above. 12 months before onset (11 before peak). Coverage: 12 of 12 eligible months.
| Lead to peak | Month | Forecast | Context |
|---|---|---|---|
| 12 months | 2006-12 | 58.8% | |
| 6 months | 2007-06 | 16.4% | |
| 3 months | 2007-09 | 10.4% | |
| 1 months | 2007-11 | 20.2% |
12-block model
First warning in the target window: 2007-02; crossing. 11 months before onset (10 before peak). Coverage: 12 of 12 eligible months.
| Lead to peak | Month | Forecast | Context |
|---|---|---|---|
| 12 months | 2006-12 | 24.4% | |
| 6 months | 2007-06 | 76.4% | |
| 3 months | 2007-09 | 56.2% | |
| 1 months | 2007-11 | 62.3% |
Peak 2020-02 · onset 2020-03
2020: exogenous pandemic shock. Included here; excluded from the primary scorecard.
yield-curve benchmark
First warning in the target window: 2019-03; already above. 12 months before onset (11 before peak). Coverage: 12 of 12 eligible months.
| Lead to peak | Month | Forecast | Context |
|---|---|---|---|
| 12 months | 2019-02 | 50.2% | |
| 6 months | 2019-08 | 75.4% | |
| 3 months | 2019-11 | 34.3% | |
| 1 months | 2020-01 | 32.5% |
12-block model
First warning in the target window: 2019-06; crossing. 9 months before onset (8 before peak). Coverage: 12 of 12 eligible months.
| Lead to peak | Month | Forecast | Context |
|---|---|---|---|
| 12 months | 2019-02 | 19.3% | |
| 6 months | 2019-08 | 35.5% | |
| 3 months | 2019-11 | 15.9% | |
| 1 months | 2020-01 | 15.4% |
False alarms and unresolved warnings
Every consecutive run at or above the alarm line outside a recession, including single months, with no recession onset within its target window. Unknown outcomes appear as unresolved warnings. Missing months, successful warnings, and recession months break a run. This strict monthly definition differs from the main page’s grouped episodes and post-recession exclusions. The 2022–2024 period is included.
| Episode | Months above line | Highest forecast | Outcome |
|---|---|---|---|
| 1982-12 to 1982-12 | 1 | 26.9% | false alarm |
| 1983-02 to 1983-03 | 2 | 85.9% | false alarm |
| 1986-04 to 1986-04 | 1 | 48.0% | false alarm |
| 1986-06 to 1986-07 | 2 | 69.0% | false alarm |
| 1986-10 to 1987-03 | 6 | 70.6% | false alarm |
| 1988-11 to 1989-07 | 9 | 98.8% | false alarm |
| 1995-06 to 1995-07 | 2 | 31.3% | false alarm |
| 1995-10 to 1996-02 | 5 | 40.6% | false alarm |
| 1998-01 to 1998-03 | 3 | 34.3% | false alarm |
| 1998-05 to 1999-03 | 11 | 61.0% | false alarm |
| 2000-03 to 2000-03 | 1 | 31.7% | false alarm |
| 2002-10 to 2002-10 | 1 | 30.2% | false alarm |
| 2006-01 to 2006-03 | 3 | 33.0% | false alarm |
| 2006-07 to 2006-12 | 6 | 66.1% | false alarm |
| 2019-01 to 2019-02 | 2 | 50.2% | false alarm |
| 2020-06 to 2020-09 | 4 | 80.8% | false alarm |
| 2022-08 to 2024-04 | 21 | 99.7% | false alarm |
| 2024-05 to 2024-12 | 8 | 98.1% | unresolved |
| 2025-02 to 2025-11 | 10 | 53.8% | unresolved |
| Episode | Months above line | Highest forecast | Outcome |
|---|---|---|---|
| 1988-11 to 1989-07 | 9 | 58.0% | false alarm |
| 1995-11 to 1996-03 | 5 | 35.5% | false alarm |
| 1996-06 to 1996-06 | 1 | 34.7% | false alarm |
| 1996-08 to 1996-09 | 2 | 33.8% | false alarm |
| 2006-11 to 2006-11 | 1 | 26.8% | false alarm |
| 2011-07 to 2011-10 | 4 | 67.2% | false alarm |
| 2020-05 to 2020-07 | 3 | 100.0% | false alarm |
| 2022-05 to 2024-04 | 24 | 83.0% | false alarm |
| 2024-05 to 2024-06 | 2 | 37.0% | unresolved |
| 2024-08 to 2024-08 | 1 | 31.3% | unresolved |
| 2024-11 to 2024-12 | 2 | 35.7% | unresolved |
As of Oct 7, 2026. Computed from the published monthly back-test probabilities, rounded to four decimals. The pre-2010 record uses many later-revised inputs and is an optimistic ceiling. EBP back-history predates its 2012 publication and favors the benchmark. Overlapping 12-month targets are not independent observations.