Market Cycles Explained: Bull and Bear Markets, Recessions, and What Past Crashes Show

Start hereIn the seven months to late August 1987, the Dow gained 44%. On 19 October 1987 the Dow fell 22.6% in a single trading session. Those losses were not followed by a recession. So what does a crash tell you about the economy?

Markets and economies rise and fall unevenly, and people have names for the swings: bull and bear, expansion and recession. The names come with rules of thumb, not a clock.

Past patterns are not predictions. Everything below describes what has already happened. It is here to help you read the news and the data, not to say what anyone should do with money.

011. A simple example

In the 1920s the Dow Jones Industrial Average rose six-fold, from 63 in August 1921 to 381 in September 1929.

By July 1932 the Dow had fallen 89% from its peak. It did not get back to its 1929 high until November 1954. Our reading: that is one full turn, with a long rise, a steep fall and a slow climb back.

022. Bull and bear markets

A bull market is a time when stock prices are rising and market sentiment is optimistic, in the words of the SEC’s investor website. Generally, it means a rise of 20% or more in a broad market index over at least two months.

A bear market is the opposite: falling prices and pessimistic sentiment. Generally, it means a fall of 20% or more in a broad market index over at least two months.

Partly A bear market means a 20% fall. The SEC’s site uses that rule of thumb, with a time condition and the word “Generally”. Our reading: 20% is a convention for labelling a move, not an official line.

For a side-by-side comparison, see Bull vs Bear Market: What’s the Difference?.

033. How US recessions are dated

The NBER’s Business Cycle Dating Committee keeps a chronology of US business cycles. The NBER is the National Bureau of Economic Research. Its definition says a recession involves a significant decline in economic activity that is spread across the economy and lasts more than a few months.

The committee treats depth, spread and length as somewhat interchangeable. So it counted the 2020 downturn as a recession even though it was brief. That recession lasted two months, from a February 2020 peak to an April 2020 trough.

It looks at several monthly measures, and there is no fixed rule for how they are weighted. In recent decades it has put most weight on real personal income less transfers and on payroll jobs.

Partly A recession means two quarters in a row of falling GDP. The NBER says most of the recessions it has dated include two or more quarters in a row of falling real GDP, but not all. The 2001 recession did not include two quarters in a row. The committee considers a range of measures, not GDP alone.

Dates arrive late

The committee waits until the data are clear, so it usually names a peak months after it happened. The December 2007 peak was announced on 1 December 2008, and the June 2009 trough on 20 September 2010.

How long they last

From 1945 to 2020, the average US recession lasted 10.3 months and the average expansion 64.2 months, by the NBER’s count. The expansion before the February 2020 peak lasted 128 months, the longest in the table.

The NBER calls expansion the normal state of the economy and says most recessions are brief. After the June 2009 trough, payroll jobs did not pass their old peak until May 2014.

044. Stocks and the economy move separately

Our reading: a falling market and a shrinking economy are two different events.

Doesn't hold up A stock market crash always means a recession. On 19 October 1987 the Dow fell 22.6% in a single trading session. Those losses were not followed by a recession or a banking crisis. US stocks passed their pre-crash highs less than two years later.

In 2022 the S&P 500 fell about 25% from 3 January to 12 October (our calculation from FRED daily closes). As of October 2026, the NBER’s most recent trough is still April 2020, so no recession has been dated since.

055. Past downturns at a glance

Episode	Index	Peak to low	Fall	Back above the old peak	NBER recession, peak to trough
1929–32	Dow	3 Sep 1929 to 8 Jul 1932	89%	23 Nov 1954	Aug 1929 to Mar 1933
1973–74	Nasdaq Composite	11 Jan 1973 to 3 Oct 1974	about 60%	7 Sep 1978	Nov 1973 to Mar 1975
1987	Dow	one day only, 19 Oct 1987	22.6%	US stocks: less than two years later	none
2000–02	Nasdaq Composite	10 Mar 2000 to 9 Oct 2002	about 78%	23 Apr 2015	Mar 2001 to Nov 2001
2007–09	Nasdaq Composite	31 Oct 2007 to 9 Mar 2009	about 56%	27 Apr 2011	Dec 2007 to Jun 2009
2020	S&P 500	19 Feb to 23 Mar 2020	about 34%	18 Aug 2020	Feb 2020 to Apr 2020
2022	S&P 500	3 Jan to 12 Oct 2022	about 25%	19 Jan 2024	none (as of Oct 2026)
1989–2009	Nikkei 225 (Japan)	29 Dec 1989 to 10 Mar 2009	about 82%	22 Feb 2024	not a US index

How to read it: the 1929 and 1987 rows come from Federal Reserve History. The 1987 row shows the one-day fall, not the full fall from the 1987 high. The rest are our calculations from daily closing levels on FRED, the St. Louis Fed’s data site. Our reading: different indexes hold different companies, so their falls are not directly comparable.

For the same 2007–09 slump, the S&P 500 fell 57% from its October 2007 peak to its March 2009 low. The 1973–74 fall overlapped with an oil shock that nearly quadrupled the price of oil, from $2.90 a barrel to $11.65 in January 1974.

Doesn't hold up Markets always bounce back quickly. Our reading of the table: recovery times ran from about six months (S&P 500, 2020) to about 34 years (Nikkei after 1989). The Nasdaq took about 15 years after 2000, and the Dow about 25 years after 1929. Those are index levels only; the FRED S&P 500 series, for one, leaves out dividends.

066. Cycle theories and what the evidence says

Here are five well-known cycle ideas, with the evidence for and against.

The Dow theory

A 1998 study called its best-known practitioner, William Peter Hamilton, “Wall Street’s most famous chartist”. Hamilton set out his market calls in editorials.

Still argued The Dow theory can call the market’s turns. In 1933 Alfred Cowles published a study of forecasts by 24 financial publications from 1928 to 1932, and found “no evidence of skill in forecasting”.

His 1930s test of the Dow theory seemed to give strong evidence against Hamilton. In 1998 three researchers re-read Cowles’s evidence and found it supports the opposite: Hamilton’s calls from 1902 to 1929 earned positive risk-adjusted returns.

Our reading: both sides looked at the same record of one writer’s calls from 1902 to 1929, and they read it differently. With one writer and one period to go on, the question is still open.

Kondratiev’s long waves

In 1925 the economist Nikolai Kondratiev described long waves of 45 to 60 years. He worked mainly from commodity price indexes for Britain, France and the United States, from the late 1700s to 1920.

Still argued The economy moves in half-century waves. A 1981 test on real growth in Britain, France, Germany and the US came out “clearly negative” for a long wave.

A 1984 study found growth differences that fit Kondratiev’s timing in several industrial countries, at least since the 1890s. It found no support for the waves in British, French or world production data before 1890. The 1984 authors say earlier negative results leaned on the same four countries.

Its authors added that the swings might have one-off causes and “need not necessarily be repeated in the future”. They also said there are so few long periods to compare that numbers alone cannot prove long cycles for now.

Minsky’s financial instability hypothesis

The economist Hyman Minsky set out the hypothesis in a 1992 working paper. He sorted borrowers into three kinds: hedge, speculative and Ponzi finance.

• Hedge borrowers can meet all their payments from their cash flows.
• Speculative borrowers can pay the interest but must keep rolling over the debt.
• Ponzi borrowers cannot cover even the interest from their operations, so they sell assets or borrow more.

His second theorem says that over long periods of prosperity, an economy shifts from financing that makes it stable to financing that makes it unstable.

In April 2009 Janet Yellen, then head of the San Francisco Fed, called the crisis “a classic case of the kind of systemic breakdown that he—and relatively few others—envisioned”. She described his view that borrowers, lenders and regulators are “lulled into complacency as asset prices rise”.

The economist Alessandro Vercelli later offered “a constructive criticism” of the three categories. He also argued that a fuller theory must explain the calm periods between crises, not only the crises. Our reading: Minsky describes how risk can build up in good times. He does not give a date for when it breaks.

“Sell in May and go away”

A 2002 study of 1970 to 1998 reported unusually high returns from November to April in US and other stock markets. It called this the Halloween effect.

Still argued Stocks do better from November to April. A 2004 critique found the US result rested on two outliers: the October 1987 crash and the August 1998 collapse of the hedge fund Long-Term Capital Management.

Allowing for those two, the effect was no longer statistically significant. Its authors also found no usable effect in S&P 500 futures from 1982 to 2003.

Answering the sceptics, a 2012 working paper by Ben Jacobsen and Cherry Zhang used 62,962 observations from stock indexes worldwide. It found returns on average 4% higher from November to April than from May to October. It was published in 2021.

A 2013 out-of-sample test found the gap persisted, at about 10 percentage points across 37 countries. A 2014 study found that a strategy based on it did not significantly outperform.

Our reading: the two sides disagree about how much a few unusual months should count, and about whether the gap holds up under stricter tests.

The presidential cycle

A 1980 study looked at US stock returns over 1961–78, by year of the presidential term. Stocks returned an average 21.7% in the year beginning two years before an election, and 15% in the year before it. They returned 3.6% in the year after the election.

A later paper looked at 1965 to 2003. In general, prices fell in the first half of a presidency, reached a low in the second year and rose in the second half.

A 2008 working paper, published in 2014, found average excess returns 9.8 percent higher in the last two years of a presidential term than in the first two, over 1948–2008. It could not explain the pattern and called it a puzzle.

Can't confirm The third year of a US presidency is reliably strong. Our reading: all three studies describe years that had already passed when they were written. They show a pattern in that past, not that it will keep going.

077. Why past patterns are not predictions

Doesn't hold up Past performance tells you what comes next. US SEC staff write that past performance “cannot predict how an investment strategy will perform in the future”. The same bulletin adds three points.

• SEC staff suggest questioning any performance presentation that does not cover periods with different market conditions, “including both up and down markets”.
• Cherry-picked results show only periods of good returns and leave out the bad ones.
• Back-tested results, where a strategy is applied to past data, are hypothetical and do not reflect actual performance.

A Federal Reserve History essay on 1929 draws a lesson for central banks: “Detecting and deflating financial bubbles is difficult”.

088. What you can do with this

• Read an NBER announcement for its dates. Look for the peak or trough month it names.
• Read a shaded chart with care. FRED shades recessions using the NBER’s turning points. It keeps shading a recession until the NBER sets its end date.
• Read a drawdown figure fully. Ask which index it uses, which peak and which low, and whether dividends are included.

09Where to go next

• NBER Business Cycle Dating: the committee’s definition, its announcements and the full table of US peaks and troughs.
• FRED: the data website of the Federal Reserve Bank of St. Louis, where many economic indicators can be found and downloaded.
• Federal Reserve History: plain-language essays on the 1929 crash, the 1973–74 oil shock, the 1987 crash and the Great Recession.
• Investor.gov: the SEC’s investor site, with its glossary and bulletins.

For more on reading numbers, see Numbers That Mislead and our page on survivorship bias.

10Quick answers

11Use this

12Quick quiz

How we checked this · Last checked October 2026

• We read the NBER, Investor.gov, FRED help and Federal Reserve History pages directly on 19 October 2026.
• Index falls other than 1929, 1987 and the S&P 500's 57% for 2007–09 are our own calculations from FRED daily closing levels. FRED blocks scripts, so we downloaded the data files with a different tool and did the sums ourselves.
• For most research papers we read the abstract. We read the full text of Cowles's 1944 paper, the 1984 IIASA paper and Minsky's 1992 paper. We could not open the 2002 study behind "Sell in May", so we describe it as the 2004 critique and the 2012 paper do.
• This page does not use the term "correction". FINRA's list of market terms calls a move of at least 10% a correction.
• The 1980 study's figure for the second year after an election is printed with a sign we could not read reliably, so the page leaves that year out.
• The line about the NBER's most recent trough was true on 19 October 2026. A new NBER announcement would change it.

How we check every article →

13Sources