Economics · Foundations
What is a recession? How the NBER dates US recessions
How the NBER’s Business Cycle Dating Committee decides when US recessions start and end, why “two negative quarters” is not the rule, and what 2001 and 2022 show.
On this page
Ask ten people what a recession is and most will say "two quarters in a row of falling GDP." That rule of thumb is easy to remember and often right, but it is not how US recessions are officially dated. The dates come from a committee of economists at the National Bureau of Economic Research (NBER), who look at a range of monthly data and decide, often a year or more later, in which month the economy peaked and in which month it bottomed.
The difference matters. The 2001 recession never had two consecutive negative quarters in today's GDP data. In 2022 the first published GDP figures did show two, and no recession was declared. This page explains the NBER's definition and method, how its dates compare with the GDP shortcut, why the announcements come so late, and what all of this means for reading economic headlines.
The official definition
The NBER describes a recession as "a significant decline in economic activity that is spread across the economy and lasts more than a few months." Three words in that sentence carry the method:
- Depth. How far activity falls.
- Diffusion. How widely the decline spreads across industries and regions.
- Duration. How long it lasts.
The committee treats the three as partly interchangeable. A decline that is extremely deep can qualify even if it is short, which is how the February to April 2020 contraction, just two months long, became a recession. A long, broad but shallow decline can qualify too.
The Business Cycle Dating Committee is a group of economists convened by the NBER, a private, non-profit research organization. Its dates have no legal standing; its dates are used by the government, economists and the press simply because they are the accepted record.
What the committee looks at
There is no formula. The committee reviews a set of monthly measures of real (inflation-adjusted) economy-wide activity:
| Indicator | Source | Why it matters |
|---|---|---|
| Real personal income less transfer payments | BEA | Income people earn, excluding government aid |
| Nonfarm payroll employment | BLS | Jobs counted from employers' payrolls |
| Household survey employment | BLS | Jobs counted from a survey of households |
| Real personal consumption expenditures | BEA | What consumers spend, adjusted for prices |
| Real manufacturing and trade sales | BEA | Sales by factories, wholesalers and retailers |
| Industrial production | Fed | Output of factories, mines and utilities |
It also considers quarterly real GDP explained: how gross domestic product is measured and read and gross domestic income (GDI), which measure the same economy from the spending and income sides. But because those are quarterly, they cannot pin down the month of a peak or trough. In recent decades the committee has put the most weight on the first two rows: real personal income less transfers and nonfarm payrolls. The NBER does not use fixed weights.
The committee publishes both a month and a calendar quarter for each turning point, and they do not always line up. The 2020 peak is dated February 2020 monthly but 2019 Q4 quarterly, so a quarterly series and a monthly one can point to different start dates for the same recession.
How dated recessions line up with GDP
The chart compares the NBER's dates with year-over-year real GDP growth. Most shaded periods dip below the zero line, and when growth is negative outside the shading it is usually in the quarter or two right after a trough, while output is still below where it was a year before. But not every recession reaches zero: in 2001 year-over-year growth bottomed at +0.2% (2001 Q4), and peak to trough real GDP fell only 0.4%. That is the case for, and against, a GDP shortcut.
Testing the two-quarter rule
Using today's quarterly GDP data (the annualized quarter-on-quarter growth rates BEA publishes), here is how the rule compares with the NBER record since 1947:
| Result | Episodes |
|---|---|
| NBER recession with two or more negative quarters in a row | 1948–49, 1953–54, 1957–58, 1969–70, 1973–75, 1980, 1981–82, 1990–91, 2007–09, 2020 |
| NBER recession without two negative quarters in a row | 1960–61 (−2.1% in Q2 and −5.0% in Q4 1960, with growth between), 2001 (−1.3% in Q1, +2.5% in Q2, −1.6% in Q3) |
| Two negative quarters in a row without a recession | 1947 Q2–Q3 (−1.0%, −0.8%) |
So the rule caught 10 of the 12 post-war recessions in current data and gave one false signal. That is a respectable record, but the misses show why the NBER does not rely on it. In 2001 the decline was too mild in GDP to produce two consecutive negative quarters, yet payroll employment and other monthly measures fell broadly enough for the committee to date a recession from March to November.
The 2022 test case
In July 2022 the Bureau of Economic Analysis published its advance estimate for the second quarter: real GDP had fallen at a 0.9% annual rate, after falling 1.6% in the first quarter. By the shortcut, the US was in a recession.
The rest of the data disagreed. Nonfarm payrolls grew by an average of 377,000 jobs a month in 2022, and the unemployment rate stayed between 3.5% and 4.0% all year. A recession is a broad decline in activity; in 2022 employment, the indicator the committee weights most, was rising fast. The NBER did not declare a recession.
Revisions then removed the shortcut's signal as well. In today's data, real GDP fell at a 1.0% annual rate in 2022 Q1 and rose at a 0.6% rate in Q2. Year over year, real GDP grew in every quarter of 2022, by 2.3% in Q2 and 1.2% in Q4. The episode is the clearest recent reminder that a first estimate is a first draft (see Seasonal adjustment and data revisions: why the numbers change).
Why the dates come so late
The committee is retrospective: it waits until enough data are in, and revised, that its dates are unlikely to need changing. That reliability is bought with delay.
| Turning point | Announced | Delay |
|---|---|---|
| Peak, Mar 2001 | November 26, 2001 | 8 months |
| Trough, Nov 2001 | July 17, 2003 | 20 months |
| Peak, Dec 2007 | December 1, 2008 | 12 months |
| Trough, Jun 2009 | September 20, 2010 | 15 months |
| Peak, Feb 2020 | June 8, 2020 | 4 months |
| Trough, Apr 2020 | July 19, 2021 | 15 months |
In each of these three recessions the trough took longer to call than the peak, by 3 to 12 months.
For markets, the delay means the official verdict is almost never news. By the time the NBER announced the December 2007 peak, the S&P 500 had already fallen sharply; by the time it announced the June 2009 and April 2020 troughs, stocks had long since rallied (the hub page, How the economy affects stocks: growth, inflation, jobs and rates, has the numbers). Investors who want a timelier read use leading indicators and real-time rules such as the The Sahm rule: a recession signal from the unemployment rate, accepting that they produce false alarms.
The US record since 1945
According to the NBER's table, the 13 contractions from 1945 to 2020 averaged 10.3 months, and expansions averaged 64.2 months. The longest post-war recession ran 18 months (December 2007 to June 2009); the shortest, two months (February to April 2020). In GDP terms the 2007–09 decline was the deepest of the post-1948 recessions before 2020: real GDP fell about 4.0% from its 2008 Q2 high to its 2009 Q2 low. The 2020 drop was deeper still, 9.1% from 2019 Q4 to 2020 Q2, and the 2001 one the mildest, 0.4%. The The business cycle: expansions, peaks, recessions and troughs page lists every post-war cycle with its length.
As of the NBER's published dates, no recession has started since April 2020.
Pitfalls when reading recession talk
- "Technical recession" is a headline phrase, not a status. It usually means two negative GDP quarters. The NBER does not use the term.
- First estimates move. GDP for a quarter is re-estimated in the following months and again in annual updates. A recession called on the advance estimate can vanish, as in 2022.
- Recession is not the same as a bad economy. Prices can be rising fast and real wages falling without activity contracting. High inflation with weak growth has its own name: Stagflation: high inflation, weak growth and the 1970s lesson.
- Other countries, other conventions. Outside the US the two-quarter shortcut is widely used in commentary, so a "recession" abroad may not mean what the NBER means. Comparisons across countries need the same definition.
- Market drawdowns are not recessions. A 20% fall in stocks, a bear market, has happened both with and without recessions (see The 2022 bear market: inflation, Fed hikes and falling bonds).
Ways to call a recession, compared
| Approach | Based on | Timeliness | Weakness |
|---|---|---|---|
| NBER dating committee | Many monthly indicators, judgment | Months to over a year late | Too late to act on |
| Two negative GDP quarters | Quarterly real GDP | About a month after the second quarter | Misses mild recessions; revised |
| The Sahm rule: a recession signal from the unemployment rate | Rise in the unemployment rate from its low | Monthly | Can trigger without a recession |
| Yield curve inversion: what it signals and its recession record | 10-year minus short-term Treasury yields | Daily; lead time varies widely | Long, variable lead; false signals |
| Leading indexes | Claims, permits, orders, spreads, stocks | Monthly | Many false alarms |
How Finabase does it
Finabase does not date recessions or publish GDP and employment data, and there is no economic calendar in the app. The GDP chart on this page is ours, drawn from public BEA data with the NBER's published dates.
What the app does show is how markets react while economists argue. The market overview collects index levels, futures and the day's movers; the Treasury table lists yields from 1 month to 30 years, the inputs to the yield-curve signal; and the economy topic in Discover gathers news coverage of GDP and jobs releases.
Frequently asked questions
Who decides when the US is in a recession?
The Business Cycle Dating Committee of the National Bureau of Economic Research, a private, non-profit research organization. Its dates are not legally binding, but they are the record used by government agencies, economists and the media.
Is two quarters of negative GDP a recession?
Not by the official US definition. It is a common rule of thumb that matches most recessions, but the NBER looks at many monthly indicators. The shortcut missed the 2001 recession, and in 2022 the first estimates showed two negative quarters with no recession declared.
How long after a recession starts is it announced?
Usually several months to a year. The NBER announced the March 2001 peak 8 months later, the December 2007 peak 12 months later and the February 2020 peak 4 months later. End dates take longer: 15 months for both the June 2009 and April 2020 troughs.
Was there a recession in 2022?
No. Real GDP was first reported falling in the first two quarters of 2022, but employment grew strongly all year and the NBER dated no recession. Revised data now show growth in the second quarter.
Summary
A US recession is whatever the NBER's Business Cycle Dating Committee says it is: a significant, broad decline in activity lasting more than a few months, judged on depth, diffusion and duration across monthly measures led by real income excluding transfers and payroll employment. Two consecutive negative GDP quarters is a useful shortcut that matches 10 of the 12 post-war recessions in today's data, but it missed 1960–61 and 2001, fired falsely in 1947 and, on first estimates, in 2022. Because the committee waits for complete data, its dates arrive months or more than a year late, long after markets have moved.
Sources and further reading
- NBER — Business Cycle Dating (definition, criteria and indicators)
- NBER — US business cycle expansions and contractions (monthly and quarterly dates)
- NBER — Business Cycle Dating Committee announcements
- BEA — Gross Domestic Product, Second Quarter 2022 (Advance Estimate), July 28, 2022
- BEA — Real GDP, percent change from preceding period, SAAR (FRED series A191RL1Q225SBEA), downloaded 2026-10-30
- BEA — Real gross domestic product (FRED series GDPC1), downloaded 2026-10-30
- BLS — All employees, total nonfarm (FRED series PAYEMS) and unemployment rate (UNRATE), downloaded 2026-10-30