Key Takeaways
- Median 12-month result after the start of a recession (a National Bureau of Economic Research business cycle peak): +18.8%, against +12.2% for all months, a gap of 6.6 percentage points above the all-months median.
- Higher in the 12 months after 10 of 13 episodes (77%), compared with 78% of all months.
- The range was wide: -25.7% after December 2007 to +37.6% after April 1960.
- Median over 3 months +1.5% (all months +3.2%); over 36 months +43.1% (all months +35.8%).
- Episodes overlap with recessions, policy changes and valuation swings, so the record shows what followed, not what the trigger caused.
What the Historical Record Shows
An episode is counted in the month the month the National Bureau of Economic Research dates as a business cycle peak. After one is counted, the next is counted only once 12 months have passed without the condition, so a long spell is one episode. The data run from June 1926 to September 2026, and 13 episodes have a full 12-month window.
The National Bureau of Economic Research dates a recession from its business cycle peak month, and it announces the date months after the fact.
| After the trigger | Episodes | Median | Average | Higher | All months: median | All months: higher |
|---|---|---|---|---|---|---|
| 3 months | 13 | +1.5% | +1.2% | 10 of 13 | +3.2% | 71% |
| 12 months | 13 | +18.8% | +12.8% | 10 of 13 | +12.2% | 78% |
| 36 months | 13 | +43.1% | +43.3% | 12 of 13 | +35.8% | 87% |
The comparison column uses every month in the record, whether or not a trigger had occurred. Dropping any one episode moves the 12-month median between +17.6% and +19.4%.
Every Episode
| Month | What changed | 3 months later | 12 months later | 36 months later |
|---|---|---|---|---|
| February 2020 | peak; trough April 2020 | +1.5% | +16.4% | +18.6% |
| December 2007 | peak; trough June 2009 | -8.0% | -25.7% | +2.8% |
| March 2001 | peak; trough November 2001 | +1.4% | -1.4% | -2.6% |
| July 1990 | peak; trough March 1991 | -5.9% | +18.8% | +43.1% |
| July 1981 | peak; trough November 1982 | +1.2% | +3.1% | +51.1% |
| January 1980 | peak; trough July 1980 | -1.6% | +20.1% | +83.1% |
| November 1973 | peak; trough March 1975 | +3.0% | -17.6% | +25.3% |
| December 1969 | peak; trough November 1970 | +1.6% | +5.4% | +39.0% |
| April 1960 | peak; trough February 1961 | +7.2% | +37.6% | +45.6% |
| August 1957 | peak; trough April 1958 | +0.9% | +28.2% | +108.8% |
| July 1953 | peak; trough May 1954 | +2.1% | +22.9% | +67.1% |
| November 1948 | peak; trough October 1949 | +4.6% | +23.2% | +62.6% |
| February 1945 | peak; trough October 1945 | +7.5% | +35.6% | +18.6% |
All 13 episodes, newest first. "Not yet" means the window has not finished.
Why the Result Differs From Episode to Episode
Defensive stocks are represented by an equal-weighted average of four industry groups from the Kenneth French Data Library (consumer nondurables, health care, utilities and telecommunications), rebalanced monthly, with dividends included, from July 1926. Other definitions of defensive would give different numbers.
Triggers and the S&P 500 series use monthly averages of daily closes, so results are read only at horizons of three months or longer, and the series before 1913 splice earlier price indexes and are less precise. Results are nominal, before inflation, taxes and fees, and use the reference month of each statistic: inflation and unemployment figures were published after the month ended, so an investor reacting to the release would have started a few weeks later.
Triggers rarely occur alone. An inflation rise, a rate rise and a valuation fall can all begin within the same year, and a recession can start before or after any of them, so each episode mixes several influences and the average hides how different the individual cases were.
What to Watch
The indicators that define or accompany this trigger:
- Unemployment Rate: Latest Reading, Historical Chart, Release Date & Investor Guide
- Initial Jobless Claims: Latest Reading, Historical Chart, Release Date & Investor Guide
- Nonfarm Payrolls: Latest Reading, Historical Chart, Release Date & Investor Guide
- GDP: Latest Reading, Historical Chart, Release Date & Investor Guide
- ISM Manufacturing PMI: Latest Reading, Historical Chart, Release Date & Investor Guide
Frequently Asked Questions
What happened after the most recent episode?
The most recent episode with a full 12-month window began in February 2020 (peak; trough April 2020). Over the next 12 months, U.S. defensive stocks (an equal-weighted blend of four defensive industry groups) returned +16.4%, against a median of +12.2% across all months.
How often did defensive stocks end higher in the 12 months after the start of a recession (a National Bureau of Economic Research business cycle peak)?
In 10 of 13 episodes (77%). Across all months since 1926, 78% of 12-month windows ended higher.
Is this a reliable signal for investing?
The record does not support treating it as one. The 12-month results ranged from -25.7% to +37.6%. Dropping any one episode moves the 12-month median between +17.6% and +19.4%. The episodes also overlap with other events that move markets. It describes what followed in the past, not what will follow.
How is the result measured?
An episode is counted in the month the month the National Bureau of Economic Research dates as a business cycle peak; the next is counted only after 12 months without the condition. Forward results are measured from that month over 3, 12 and 36 months using U.S. defensive stocks (an equal-weighted blend of four defensive industry groups). Triggers and the S&P 500 series use monthly averages of daily closes, so results are read only at horizons of three months or longer, and the series before 1913 splice earlier price indexes and are less precise. Results are nominal, before inflation, taxes and fees, and use the reference month of each statistic: inflation and unemployment figures were published after the month ended, so an investor reacting to the release would have started a few weeks later.