One hundred and seventy years of American cycles, as the committee dates them

There is exactly one chronology of American business cycles, it is free, and almost nobody reads it. It runs from a peak in December 1854 to a trough in April 2020, gives thirty-four contractions with their durations in months, and publishes averages by era that make the modern cycle look nothing like the nineteenth-century one. This page reads that table, the criteria the committee applies, and the six monthly indicators it actually consults — and then measures how far industrial production fell in each of the twelve recessions since 1948, because that is a number the table does not contain.

What a cycle is, in the words of the body that dates them

The NBER page carries a one-sentence definition of what its dates mean, and it is more precise than any paraphrase: Recessions — contractions in economic activity — start in the month after a peak in the business cycle, and end in the month of the trough. Its longer statement is that a recession involves a significant decline in economic activity that is spread across the economy and lasts more than a few months, and that the committee treats the three criteria — depth, diffusion and duration — as somewhat interchangeable, so that extreme conditions revealed by one criterion may partially offset weaker indications from another.

That last clause is what made February 2020 datable. The committee writes that it concluded the subsequent drop in activity had been so great and so widely diffused throughout the economy that, even if it proved to be quite brief, the downturn should be classified as a recession. It then dated the trough two months later. A two-month recession exists in this chronology only because the duration criterion was allowed to be overwhelmed by the other two.

The committee also names what it looks at, and the list is short: real personal income less transfers, nonfarm payroll employment, employment as measured by the household survey, real personal consumption expenditures, manufacturing and trade sales adjusted for price changes, and industrial production. It adds that there is no fixed rule about what measures contribute information to the process or how they are weighted, and that in recent decades the two measures we have put the most weight on are real personal income less transfers and nonfarm payroll employment. Gross domestic product is not on the monthly list at all; it enters only in the separate determination of the peak or trough quarter, alongside gross domestic income.

The averages, and why one number for “the cycle” is useless

The chronology ends with four rows of averages, in months, computed by the NBER itself over its own table. They are the clearest evidence that the American business cycle has changed shape.

Contractions have halved, from 21.6 months before 1919 to 10.3 months after 1945. Expansions have more than doubled, from 26.6 months to 64.2. The full cycle has grown from about four years to about six and a quarter. Any sentence of the form “the business cycle lasts about X years” is therefore true only of the era it is taken from, and citing the 1854–2020 average of 58.4 months means averaging a nineteenth-century agricultural economy with a twenty-first-century one.

The dispersion inside the modern era is wider still than the averages suggest. The shortest post-war expansion in the table is 12 months, between the July 1980 trough and the July 1981 peak; the longest is 128 months, from June 2009 to February 2020. The shortest contraction is 2 months in 2020 and the longest 18 months, in 2007–2009. There is no typical cycle to be found in a range of that width.

The twelve post-war recessions, with a number the table does not contain

The chronology gives durations but no measure of severity, and duration is a poor stand-in for it: the shortest contraction in the modern era is also one of the two deepest. The last column below is not in the NBER table. It is industrial production, one of the six indicators the committee names, measured from the level in the peak month to its lowest level between the peak and the trough — computed here from the published monthly index.

Length and depth are almost unrelated in that table. The 2020 contraction lasted 2 months and took industrial production down 16.6%. The 1990 and 2001 contractions lasted 8 months each and took it down 3.8% and 3.5%. The 2007–2009 contraction, the longest in the modern era at 18 months, produced the largest fall at 17.2%— barely more than the two-month one. A reader who knows only how long a recession lasted knows very little about it.

The expansion column carries its own warning. The 128-month expansion that ended in February 2020 is the longest in the entire chronology, longer than any of the thirty-three that preceded it, and it ended anyway. Expansions in this table do not die of old age on any schedule that the numbers reveal: the longest was followed by the shortest contraction, and the 12-month expansion of 1980–81 was followed by one of the longest.

Four things the chronology explicitly refuses to do

It does not accept the two-quarter rule.Asked directly about the financial press’s definition, the committee answers that most of the recessions identified by our procedures do consist of two or more consecutive quarters of declining real GDP, but not all of them, and names 2001 as a case that did not. It then gives four reasons for rejecting the rule: activity is not identified solely with real GDP; the decline must be significant, so small falls in two quarters need not qualify; the main focus is monthly rather than quarterly; and gross domestic income is given equal weight to gross domestic product, with the statistical discrepancybetween them singled out as important in 2001 and 2007–2009.

It does not recognise a “double-dip recession”. The FAQ states that the NBER does not define a special category called a double-dip recession and that two periods of contraction will be either two separate recessions or parts of the same recession, decided by the duration and strength of the upturn after the initial trough. It gives its own example: 1980 and 1981 were treated as separate recessions partly because major indicators bounced back in late 1980 and early 1981.

It does not identify depressions. The committee writes that the NBER does not separately identify depressions in its business cycle chronology— there are only contractions and expansions — while noting that the period generally regarded as a depression in the United States is the 1930s, for which its own dates are a peak in August 1929, a trough in March 1933, a second peak in May 1937 and a second trough in June 1938.

It does not claim the trough is a return to health. The committee is explicit that recessions and expansions refer to the direction of change in economic activity, not its level, and gives two of its own measurements: after the June 2009 trough, real personal income less transfers did not exceed the level of the previous peak until July 2011, and nonfarm payroll employment did not exceed the level of the previous peak until May 2014— fifty-nine months after the recession officially ended.

The quarter and the month can disagree

The chronology carries two dates for each turning point, a month and a quarter, and the NBER marks in red the cases where the quarter does not contain the month. The most recent such case is the February 2020 peak, which the table dates to 2019Q4— a quarter that ended six weeks before the peak month. The committee explains that it makes a separate determination of the calendar quarter of a peak or trough, using quarterly measures including real GDP and real GDI, so the two answers come from different data and need not agree.

This is not a technicality for anyone comparing cycles across countries or joining monthly and quarterly series. Dating the last expansion’s end to the fourth quarter of 2019 rather than to February 2020 moves it across a year boundary, and any annual figure built on the quarterly date will disagree with any built on the monthly one.

What these sources do not say

• The chronology stops at April 2020 and is dated. The NBER page records 14 March 2023 as the last update of the business cycle data. Nothing in it describes 2021 to 2026.
• It contains no measure of severity. Duration is the only quantity in the table. The industrial production column on this page was computed here and is not part of the NBER publication.
• It says nothing about causes. No row of the chronology names a reason for a contraction, and none is supplied here.
• It does not date cycles outside the United States. No euro area, British or Japanese chronology was read for this page.
• It does not tell you when the next peak will be. The committee dates turning points retrospectively and announced its last six peaks between four and eleven months after the fact.
• Nothing here connects a cycle to an asset price. No market series was read for this page and no such statement appears above.

How to verify this yourself

1. Open the NBER expansions and contractions table and count the rows with both a peak and a trough. There should be thirty-four.
2. Read the four summary rows at the bottom of that table and check the contraction column: 17.0, 21.6, 18.2 and 10.3 months.
3. Find the note above the table recording the date the business cycle data were last updated.
4. Open the FAQ and search for the phrase “two consecutive quarters”. Read both the question and the following one, which lists four reasons.
5. Download fredgraph.csv?id=INDPRO, take the value for December 2007 and the lowest value up to June 2009, and compute the fall. You should get about minus 17.2%.
6. Look at the last row of the chronology and note that the peak month is February 2020 while the peak quarter is 2019Q4.

Where to go next on this site

• Recession indicators — four signals tested against these same dates, with the announcement lag measured.
• What GDP is — the quarterly measure, its three estimates and how much they move afterwards.
• Leading and lagging indicators — which series turn first, and two that stopped being published.
• All macroeconomics guides — the area index, one click away.

Sources opened for this page

Each source below was opened and read on 13 August 2026. Verified on 13 August 2026.

• National Bureau of Economic Research — US Business Cycle Expansions and ContractionsUsed for: all thirty-four peak and trough pairs with their contraction, expansion and cycle durations, the four era averages of 17.0, 21.6, 18.2 and 10.3 months for contractions and 41.4, 26.6, 35.0 and 64.2 for expansions, the note that the data were last updated on 14 March 2023, and the sentence defining when a recession starts and ends. Read 13 August 2026.
• National Bureau of Economic Research — Business Cycle DatingUsed for: the definition emphasising depth, diffusion and duration as somewhat interchangeable, the reasoning behind the February 2020 peak, the list of six monthly indicators, the statement that real personal income less transfers and payroll employment carry the most weight, and the separate quarterly determination using GDP and GDI. Read 13 August 2026.
• National Bureau of Economic Research — Business cycle dating procedure, frequently asked questionsUsed for: the four reasons for rejecting the two-quarter rule and the naming of 2001 as an exception, the refusal to define a double-dip recession and the 1980–81 example, the refusal to identify depressions with the 1929, 1933, 1937 and 1938 dates, the July 2011 and May 2014 recovery examples, and the duration convention. Read 13 August 2026.
• Federal Reserve Bank of St. Louis — FRED series INDPRO, Industrial Production Total IndexUsed for: the monthly index from January 1919 to June 2026, from which the peak-to-low fall in each of the twelve post-war recessions was computed, ranging from minus 3.5% in 2001 to minus 17.2% in 2007–2009. Read 13 August 2026.

Change log

• 13 August 2026— page rewritten from scratch against the NBER chronology, the Business Cycle Dating page and the dating procedure FAQ, plus the full industrial production index, all opened that day. Removed from the previous version: a four-phase cycle model with named stages that appears in no source; an average cycle length quoted as a single number across 1854 to 2020; the two-quarter definition of a recession presented as standard, which the committee rejects in writing with four reasons; a claim that expansions end after a typical number of years; and a description of recovery that ignored the committee's own statement that a trough is a change of direction, not a return to the previous level.

Vextor Capital is not authorised under MiFID II as an investment firm, is not registered with the SEC, FINRA, the CFTC or the NFA, is not an investment adviser, holds no position in any instrument named on this page, and receives no payment from any bank, broker, exchange, index provider or data vendor. Every outbound link on this page points to a statistical agency, a central bank or a public data series.