Bear Markets Since 1950: Depth, Duration, and When You Actually Recovered

Every guide to bear markets quotes the same reassuring pair of numbers: the average drop is about a third, and the market always comes back. Both are true and neither is the number you need. The question that decides whether a plan survives is how long you spend waiting — and the honest answer depends entirely on whether you measure the index price, the price with dividends reinvested, or what your money could actually buy when you got it back.

Those three measurements disagree violently. On price, the S&P 500 has had eleven bear markets since 1950 and the median one returned to its previous peak in 24.5 months. Count dividends and the wait gets shorter. Count inflation as well and two of those episodes took twelve years or more — and the 2000 and 2008 crashes stop being two events and become a single drawdown lasting nearly thirteen years.

The eleven bear markets since 1950

A bear market here means a fall of 20% or more from a previous all-time high, measured on daily closing prices of the S&P 500, with the episode ending on the first close back at that high. Intraday prices would add one or two near-misses; closing prices are used throughout so that every row is measured the same way.

Median: 33.5% deep, 14.6 months down, 15.2 months back up, 24.5 months peak to peak. Mean: 34.7% deep and 38.1 months peak to peak. Eleven episodes in 76.8 years is roughly one every seven years.

The median is two years. The plan has to survive the mean.

The gap between the median (24.5 months) and the mean (38.1 months) is the whole problem. Eight of the eleven episodes were back at their old high inside 40 months. The other three took 65.6, 86.2 and 90.2 months — between five and a half and seven and a half years each. A distribution this skewed cannot be summarised by its centre, and a plan built on “about two years” is a plan that has quietly assumed the two worst cases away.

The last column explains part of the asymmetry. A 25.4% fall needs 34.1% to undo. A 56.8% fall needs 131.3%. Losses and the gains that reverse them are not symmetrical, and the gap widens brutally as the loss deepens — the arithmetic behind every risk management rule. The 2007 episode was 8.6 percentage points deeper than 1973, but it needed 38.2 points more to get back.

Dividends shorten the wait. Inflation more than undoes the favour.

The table above tracks the index price, which is what “the S&P 500” means in a headline and not what an investor holds. A real holding pays dividends, and the money that comes back has to buy something. Those two adjustments pull in opposite directions, and they are not close to equal. Using Robert Shiller’s monthly series from 1950 — same source, same frequency, four ways of measuring the same index:

Read the third row against the first: reinvested dividends cut the time spent 20% or more underwater from 13.2% of months to 8.5%, roughly a third less. Now read the fourth against the third: inflation pushes it back up to 24.3%, nearly three times the nominal figure. Over this period inflation cost the investor more waiting time than dividends saved.

The same ranking shows up in the compound returns. From January 1950 to September 2026 the index compounded at 8.30% a year on price, 11.65% with dividends reinvested, and 7.85% once inflation is removed from that total. Dividends were worth about 3.4 points a year; inflation took about 3.8 back. This is the same mechanism that erodes cash savings, applied to an asset that outruns it — but only just, and never on schedule.

2000 and 2008 were one drawdown, not two

Apply the same 20% threshold to total return after inflation and the eleven price bear markets collapse into six real ones. Five of the eleven drop out, for two different reasons. Three stayed under the 20% threshold once dividends are counted — 1956 (−16.0%), 1966 (−18.1%) and 2020 (−18.9% on this monthly series, though daily data puts it at −33.6%). Two disappear because they began while the index was still below an earlier real peak and so never started a new drawdown: 1980, which fell inside the episode that began in 1973, and 2007, which is absorbed into the one that began in 2000. The six that remain look very different:

The fifth row is the one worth sitting with. In price terms, an investor who bought at the March 2000 top was whole again in May 2007, lost it all over again, and recovered a second time in March 2013: two bear markets with a brief peak between them. In total return after inflation there was no peak between them. The August 2000 high was not exceeded until May 2013 — one continuous drawdown of 153 months, with its low point in March 2009, eight and a half years after it began.

This is a strong claim, so it is worth checking against an entirely separate dataset. Using Yahoo’s S&P 500 Total Return index on daily data deflated by CPI-U, the same drawdown runs from 24 March 2000 to 7 May 2013 — a peak 58.2% deep and 13.1 years long. Two independent sources, two frequencies, the same month of recovery. The daily data also picks up two brief real drawdowns the monthly series smooths away (a 21.1% fall in 1990 lasting eight months, and 33.6% in 2020 lasting six), so the table above should be read as the long episodes, not an exhaustive list.

The nominal figures from that same daily series show how much the adjustment matters: without the inflation adjustment, 2000 recovered in October 2006 after 6.1 years and 2007 recovered in April 2012 after 4.5 years. Two ordinary bad decades become one lost one.

Depth predicts duration — but much of that is arithmetic

Across the eleven episodes, depth and the time spent climbing back are strongly related: the correlation between the size of the fall and the months from trough to old high is −0.86, meaning deeper falls took longer to undo. The correlation between how long the fall took and how long the recovery took is weaker, +0.63.

Both deserve a caveat that is usually left out. Part of the first relationship is not information but definition: a deeper fall mechanically requires a larger gain, so at any given rate of recovery it must take longer. And eleven observations is a small sample in which a single episode moves the number materially. Treat these as a description of what happened, not as a rule for forecasting the next one — 1980 fell 27% and was back at its high 2.7 months after the low, while 1987 fell 33% and took 19.7 months.

The recovery arrives before it feels safe

The practical question during a bear market is never “will it recover” but “can I wait until the coast is clear”. The record says the coast clears after the fact. Measured from each of the eleven lows:

• The following twelve months were positive in all eleven cases, with a median gain of +33.7%.
• The first three months alone delivered a median +17.2%, and the first six a median +22.1%.
• In the median episode, 78% of the entire climb back to the old high happened within twelve months of the low.

Nobody rings a bell at the bottom, and none of this is a case for trying to catch it. It is a case against the more common manoeuvre: selling during the fall with the intention of buying back once things stabilise. By the time stability is visible, the median episode has already handed back more than three quarters of its ground. That is the same arithmetic that makes missing a handful of the best days so expensive, and the reason the measured cost of investor behaviour concentrates in exactly these windows.

What this changes about a plan

Match the horizon to the worst case, not the median. Money needed inside three years has no business in an asset whose real recovery took twelve years twice in seventy-six. That is the honest basis for a risk profile: not how a 30% fall feels, but whether the money can wait out a decade of it.

Size positions so waiting stays possible. The index recovered in every one of the eleven cases. Individual positions carry no such guarantee, which is why position size, not the forecast, decides whether a drawdown is survivable.

Have the buying rule written before the fall. A pre-committed schedule — a rebalancing rule or a fixed contribution — mechanically buys more at lower prices, without requiring a view on the bottom. The lump sum versus phasing-in evidence is the same argument about a different decision.

Expect to spend most of your time below a previous high. On daily closes since 1950 the S&P 500 has been at an all-time high on just 8.1% of sessions, 10% or more below one on 35.2%, and 20% or more below one on 16.0%. Being underwater is the normal state of the asset, not the exception.

What this does not say

This is one index, in one country, over one period that happened to be extraordinarily good for it — the same exercise on the Nikkei from 1989, or on a portfolio concentrated in a sector, produces a far worse table. The 20% threshold is a convention with nothing behind it; a 19.4% fall in 1976–78 is absent from the table for that reason alone. Monthly total return data smooths the extremes, so the depths in the real table are understated relative to daily prices. And eleven episodes cannot establish a distribution: the next bear market is free to be deeper, longer or both, and nothing here forecasts when it starts. Sector positioning — which exposures have historically held up better in a fall — is a separate question, covered in crisis-proof investments.

Frequently asked questions

How long does a bear market last on average?
Measured on S&P 500 closing prices since 1950, the median bear market took 14.6 months to reach its low and 15.2 months more to regain the previous high — 24.5 months peak to peak. The mean is much longer at 38.1 months, because two episodes (1973 and 2000) took over seven years each.

How many bear markets has the S&P 500 had since 1950?
Eleven, using a 20% fall from an all-time high on closing prices: 1956, 1961, 1966, 1968, 1973, 1980, 1987, 2000, 2007, 2020 and 2022. That is roughly one every seven years. Counts differ between sources depending on whether intraday prices are used and how closely spaced episodes are grouped.

What was the worst bear market since 1950?
On price, 2007–2009: a 56.8% fall requiring a 131.3% gain to undo, with the old high regained in March 2013. On total return after inflation the longest was the drawdown that began in August 2000 and did not end until May 2013 — 12.8 years, absorbing both the dot-com crash and the financial crisis as one continuous episode.

Do dividends change how long recovery takes?
Yes, materially. Since 1950 the index spent 13.2% of months more than 20% below its previous peak on price alone, but only 8.5% once dividends are reinvested. Dividends added about 3.4 percentage points a year to the compound return over the period.

How long did it take to recover from the 2008 crash?
In price terms, from the October 2007 peak to the March 2013 close: 65.6 months, or about five and a half years. With dividends reinvested it was roughly four and a half years. After inflation it was not a separate recovery at all — it was part of the longer drawdown that began in 2000 and ended in 2013.

Should I sell and buy back when the market stabilises?
The record makes that expensive. In all eleven episodes the twelve months after the low were positive, with a median gain of 33.7%, and in the median case 78% of the entire climb back happened within that first year — well before conditions felt settled. Selling into the fall converts a temporary drawdown into a permanent one whenever the re-entry point is missed.

Sources

• S&P 500 daily closing prices, 3 January 1950 to 18 September 2026, from Yahoo Finance (ticker ^GSPC, 19,321 sessions), retrieved 20 September 2026. Source of the eleven-episode table, the depth and duration figures, the post-trough rebound figures and the share of sessions spent below a previous high.
• Robert J. Shiller, U.S. Stock Markets 1871–Present and CPI Data (monthly series accompanying Irrational Exuberance), file retrieved 20 September 2026, covering January 1871 to September 2026. Source of the real total return series and of the four-way comparison table. Note that Shiller’s monthly price is an average of daily closes within the month, which smooths peaks and troughs; depths taken from it are therefore conservative relative to daily data. The most recent months of CPI are the author’s estimates.
• S&P 500 Total Return index, Yahoo Finance (ticker ^SP500TR), daily, 4 January 1988 to 18 September 2026. Used as an independent cross-check on the 2000–2013 real drawdown, deflated by CPI-U held constant within each month.
• Consumer Price Index for All Urban Consumers, U.S. city average, all items, not seasonally adjusted (series CUUR0000SA0), U.S. Bureau of Labor Statistics, retrieved via the BLS public API on 20 September 2026. Used to verify the CPI figures in the Shiller file, which matched exactly for the months checked through July 2026 (333.952 in June, 333.918 in July); the final months of the file are the author’s estimates and differ slightly from the published figure.
• Episode definitions, drawdown calculations, correlations and rebound figures are our own, computed from the series above. A bear market is a fall of 20% or more from an all-time closing high, ending at the first close back at that high.

Related reading

• How to manage investment risk — the recovery arithmetic in the last column, and the hierarchy it sits inside.
• Stock market mistakes, ranked by what they cost — what selling during the fall costs, measured.
• Behavioral finance: six biases with a measured cost — why the selling happens at the low and not before.
• Position sizing and the risk of ruin — the index always came back; a position need not.
• How often should you rebalance — the rule that buys the fall without requiring a view.
• Lump sum vs dollar-cost averaging — the same evidence applied to when you put money in.
• Why your savings lose value every year — the inflation adjustment that turned two recoveries into one lost decade.
• Crisis-proof investments — which exposures behaved differently inside these falls.

This article is general information, not personalised investment advice. It does not take into account the financial situation, objectives or risk tolerance of any individual reader, and the same text is distributed to all readers. All figures describe specific historical periods for one index and do not predict future outcomes; past bear markets do not establish the depth, duration or timing of future ones. Capital is at risk and past performance does not indicate future results.