BigCapital beta

Article · Markets

Bubbles: what they look like in the data

No price tells you that a boom has become a bubble; what the data can show is the shape. By one rule — a price that at least doubled within five years and then gave back at least half of that rise — the monthly prices of US stocks, gold, silver, oil, bitcoin and ether hold 19 booms and busts since 1881. In the median one the price rose ×4.9 into its peak and then fell 67.0%.

Updated with Oct 2026 data8 min read

Booms and busts since 1881
19

in 6 markets, found by one rule: 4 in US stocks, 2 in gold, 3 in silver, 4 in oil, 4 in bitcoin and 2 in ether

Median rise into the peak
×4.9

from the lowest month of the five years before; after inflation ×4.2

Median fall after the peak
−67.0%

within the five years after; after inflation −67.5%

The deepest fall
−90.1%

Ether 2018: from January 2018 to December 2018, monthly averages

Median time back to the peak
8.8 years

2 of 19 had not regained their peak by October 2026

What counts as a bubble here

Economists disagree on what a bubble is, and a price alone cannot tell a mania from good news that later turned. What prices can show is the shape: a fast climb, then a fall that takes much of it back. Here a month counts as the peak of a boom and bust when:

  1. its price was the highest of the five years up to it;
  2. it was at least twice the lowest monthly price of those five years — the price had at least doubled;
  3. within the next five years, before going above the peak again, the price fell far enough to give back at least half of that rise.

We applied the rule to the monthly prices of six markets: US stocks (the S&P 500 and the indexes it continues, since 1871), gold, silver and Brent oil (since 1960), bitcoin (since 2010) and ether (since 2015). It finds 19 booms and busts: 4 in US stocks, 2 in gold, 3 in silver, 4 in oil, 4 in bitcoin and 2 in ether.

The rule describes a shape, not a cause: some of these were speculative manias, others were shocks — oil jumped in 1990 when Iraq invaded Kuwait. Famous bubbles from before our data — the Dutch tulip mania of 1636–37, the Mississippi and South Sea bubbles of 1720, Britain's railway mania of the 1840s — and Japan's stock and land bubble of the late 1980s have no price series on this site, so they are named here without numbers.

US home prices, at the heart of the 2000s housing bubble, move far more slowly. From July 2001 to the peak in July 2006 they rose ×1.6 — short of the rule's doubling — and then fell 27.4% to February 2012 (35.1% after inflation), the deepest fall in the monthly home price data since 1953. Prices were back at the peak in January 2017. The charts show them beside the episodes, for comparison.

The anatomy of a bubble

Every episode aligned at its peak month: the price as a percentage of the peak, from five years before to five years after, on a log scale (equal steps are equal multiples). The bold line is the median of the 19 paths. Hover or tap a line to see which episode it is; the buttons pick out one kind of market.

On the median path the price stood at 30% of the peak five years before and 55% a year before: the last year alone added +81%. A year after the peak it was down to 62% of it, two years after at 59%. Up by the stairs, down by the elevator: on the way up the median price gained 2.1% a month on average; in the year after the peak it lost 4.0% a month — 2 times as fast. Prices as quoted; the switch shows the same months after US inflation.

S&P 500GoldSilverOilBitcoinEtherUS homes, for comparisonMedian path
Monthly averages, % of each episode's peak month, log scale. Lines below 1% of the peak or above twice the peak are cut at the edge: bitcoin and ether rose from far below. Data: Robert J. Shiller (stocks, US CPI, home prices) · World Bank (metals, oil) · Coin Metrics (bitcoin, ether)

Boom, then bust

Each episode's rise into the peak (left, from the lowest month of the five years before) against its fall after it (right, to the lowest month of the next five years). The biggest rise was Bitcoin 2014: ×13.2K from September 2010; the smallest, S&P 500 1881: ×2.4. The deepest fall was Ether 2018: −90.1%.

In 1 of 19 episodes the fall took back the whole rise or more: S&P 500 1929. Ether 2021: the five years after the peak are not over, so the fall may not be either.

S&P 500GoldSilverOilBitcoinEtherUS homesFall after the peak
Rise (left, in the market's color): the peak month against the lowest month of the five years before it, log scale. Fall (right): the lowest month of the five years after, before any higher month, against the peak. Monthly averages. Source: World Bank, Commodity Price Data (The Pink Sheet), CC BY 4.0
Every episode as a table
EpisodePeakRise into the peakFall after itRise given backBack at the peak
S&P 500 1881Railroad boom; the panic of 1884Jun 1881×2.4from Jun 1877−35.6%to Jan 188561%Dec 1900after 19.5 years
S&P 500 1929The Roaring Twenties and the Wall Street CrashSep 1929×3.4from Oct 1924−84.8%to Jun 1932120%Sep 1954after 25 years
S&P 500 1937Recovery from the Depression; the recession of 1937–38Feb 1937×3.8from Jun 1932−52.2%to Feb 194271%Apr 1946after 9.2 years
S&P 500 2000The dot-com boomAug 2000×2.7from Aug 1995−43.7%to Feb 200370%May 2007after 6.8 years
Gold 1980The inflation of the 1970sJan 1980×6.1from Aug 1976−55.2%to Jan 198566%May 2006after 26.3 years
Gold 2011After the 2008 crisis and the euro debt crisisSep 2011×3.0from Oct 2006−39.3%to Dec 201559%Jul 2020after 8.8 years
Silver 1980The Hunt brothers' attempt to corner silverJan 1980×9.6from Jan 1976−85.7%to Jun 198296%Apr 2011after 31.2 years
Silver 2008The commodity boom and the 2008 crisisMar 2008×4.3from Apr 2003−48.9%to Nov 200864%Sep 2010after 2.5 years
Silver 2011The metals boom after the 2008 crisisApr 2011×4.3from Nov 2008−67.0%to Jan 201687%Sep 2025after 14.4 years
Oil 1990Iraq's invasion of KuwaitOct 1990×3.8from Jul 1986−62.3%to Dec 199384%May 2004after 13.6 years
Oil 2000OPEC output cuts; the 2001 recessionSep 2000×3.3from Dec 1998−43.0%to Dec 200161%Feb 2003after 2.4 years
Oil 2008The commodity boom and the 2008 crisisJul 2008×4.9from Sep 2003−68.9%to Dec 200886%Not yet18.2 years so far
Oil 2022Reopening after the pandemic; the war in UkraineJun 2022×5.1from Apr 2020−47.8%to Dec 202559%Apr 2026after 3.8 years
Bitcoin 2011Bitcoin's first boom; the Mt. Gox hack of June 2011Jun 2011×292from Sep 2010−85.1%to Nov 201185%Feb 2013after 20 months
Bitcoin 2014The 2013 boom; the collapse of Mt. Gox in 2014Jan 2014×13.2Kfrom Sep 2010−71.4%to Sep 201571%Dec 2016after 2.9 years
Bitcoin 2017The 2017 boom of new coins (ICOs)Dec 2017×1,136from Dec 2012−75.9%to Feb 201976%Nov 2020after 2.9 years
Bitcoin 2021The 2021 boomNov 2021×83.7from Nov 2016−72.1%to Dec 202273%Mar 2024after 2.3 years
Ether 2018The 2017 boom of new coins (ICOs)Jan 2018×1,671from Oct 2015−90.1%to Dec 201890%Jan 2021after 3 years
Ether 2021The 2021 boom of DeFi and NFTsNov 2021×567from Dec 2016−72.2%to Dec 202272%Not yet4.9 years so far

Monthly averages. The switch above the anatomy chart shows the same months after US inflation. “Back at the peak”: the first month after the bottom whose average was at or above the peak's.

How long the way back took

From the peak month to the first month when the price stood at the peak again, the median episode took 8.8 years (counting the ones still below their peak as the longest). The fastest was Bitcoin 2011: 20 months; the longest, Silver 1980: 31.2 years, back only April 2011.

2 of 19 had not regained their peak by October 2026: Oil 2008 and Ether 2021. After inflation the median is 14.2 years, and 4 are still below: Silver 1980, Oil 2008, Oil 2022 and Ether 2021.

FallingClimbing backBack at the peak
Falling: from the peak to the lowest month. Climbing back: to the first month at or above the peak again. Open bars: not back yet, to the latest month of the data. Data: Coin Metrics, CC BY-NC 4.0

Were stocks expensive at their peaks?

Stocks have a yardstick that metals and coins lack: earnings. Robert Shiller's CAPE divides the price by the average of the previous ten years' earnings, both after inflation. At the stock peaks the rule finds it stood at 19.0 (Jun 1881), 32.6 (Sep 1929), 22.2 (Feb 1937) and 42.9 (Aug 2000). Its average since 1881 is 17.8 (median 16.6); the highest was 44.2, December 1999.

In October 2026 the CAPE was 40.7: only 1% of all months since 1881 had a higher one, and it was higher still at the peak of August 2000. A high CAPE says that prices are high against past earnings; it does not say when, or whether, they will fall.

CAPEAverage since 1881: 17.8
Price divided by the average of ten years of earnings, both after inflation (CAPE), monthly. Dotted lines: the stock peaks the rule finds. Data: Robert J. Shiller (stocks, US CPI, home prices) · World Bank (metals, oil) · Coin Metrics (bitcoin, ether)
MonthCAPEAgainst the average
Jun 188119.01.1×
Sep 192932.61.8×
Feb 193722.21.2×
Aug 200042.92.4×
Latest, Oct 202640.72.3×

Where the markets stand now

A rule like this names a bubble only after it bursts, so it cannot say whether any market is in one today. It can measure the first half of the shape: how far each price has risen from its lowest month of the last five years. At their latest monthly average the S&P 500, gold, silver and bitcoin stood at least twice their lowest month of the past five years.

The table compares each market's latest monthly average with five years earlier, with its lowest month of the last five years and with its highest month ever, prices as quoted. Over five years the S&P 500 rose ×1.7; its latest month was a record.

MarketAgainst 5 years earlierAgainst the 5-year lowAgainst the record
S&P 500Oct 2026×1.7×2.1from Oct 2022At a record
GoldOct 2026×2.3×2.5from Oct 2022−18.2%record Feb 2026
SilverOct 2026×2.4×3.0from Sep 2022−38.1%record Jan 2026
OilOct 2026×1.5×1.99from Dec 2025−6.6%record Jul 2008
BitcoinOct 2026×1.4×4.8from Dec 2022−28.8%record Jul 2025
EtherOct 2026×0.6×1.99from Dec 2022−44.6%record Nov 2021
US homesJul 2026×1.3×1.3from Jul 2021At a record

Monthly averages, prices as quoted; the latest whole month of each market's data.

Questions people ask

What is a market bubble?

A price that rises far and fast — beyond what later proves lasting — and then falls back. No price level marks it in advance. In the 19 booms and busts here, the median one rose ×4.9 in the run-up and fell 67.0% afterwards; five years before its peak the price had stood at 30% of it.

What was the biggest bubble in history?

In the data on this site, by the rise into the peak: Bitcoin 2014, ×13.2K; by the fall: Ether 2018, −90.1%; by the time back to the peak: Silver 1980, 31.2 years. Older famous ones, such as the tulip mania of 1637 or the South Sea bubble of 1720, have no monthly price series comparable to these.

Is the stock market in a bubble now?

No measure can say that in advance. What the data show for Oct 2026: the S&P 500 had risen ×1.7 in five years (the rule's boom starts at ×2), and Shiller's CAPE was 40.7, against an average of 17.8 since 1881, 32.6 at the 1929 peak and 42.9 at the 2000 peak.

Is bitcoin a bubble?

By this rule bitcoin has gone through 4 booms and busts, with falls of: −85% after 2011, −71% after 2014, −76% after 2017 and −72% after 2021. Each time the price later rose above the old peak. Over the five years to Oct 2026 it rose ×1.4; in October 2026 it stood 28.8% below its record of July 2025 (monthly averages).

How long does it take to recover after a bubble bursts?

In these 19 episodes the median wait from the peak back to the same price was 8.8 years; it ranged from 20 months (Bitcoin 2011) to 31.2 years (Silver 1980), and 2 are still below their peak. After inflation the median is 14.2 years.

Go deeper

How the numbers are calculated

  • The rule: a month is the peak of a boom and bust if its monthly average price was the highest of the five years up to it, at least twice the lowest month of those five years, and if within the next five years, before any higher month, the price fell far enough to give back at least half of the rise from that low. Peaks that share the same bottom count once (the highest). Where a market's data start less than five years before a peak, the rise is measured from the first month of data.
  • Monthly averages: every price here is a monthly average — for bitcoin and ether the average of daily closes, whole months only. Daily highs and lows went further: an average smooths the very top and the very bottom.
  • After inflation: prices divided by the US consumer price index (Shiller's monthly CPI), in dollars of Oct 2026. The episodes are found in prices as quoted; the switch shows the same months after inflation, so a rise, a fall or the way back can differ.
  • US home prices are not one of the rule's markets: they are shown for comparison, from the highest month before their deepest fall in the monthly data since 1953. Shiller's index, not seasonally adjusted (prices peak in the summer).
  • CAPE: Robert Shiller's cyclically adjusted price-earnings ratio — the S&P 500 price divided by the average of the previous ten years of earnings, both after inflation — from Shiller's data file.
  • Stocks, CAPE, US CPI and home prices: Robert J. Shiller, Yale University — S&P Composite data (ie_data.xls) and U.S. Home Price and Related Data (Irrational Exuberance, 3rd ed., Princeton University Press, 2015, as updated by the author).
  • Gold, silver and Brent oil: World Bank Commodity Price Data (The Pink Sheet), monthly averages in US dollars, CC BY 4.0. Source: World Bank.
  • Bitcoin and ether: Data: Coin Metrics Community Network Data (daily reference rate PriceUSD), CC BY-NC 4.0, used here for a non-commercial purpose.
  • S&P 500 data notes (Shiller's file): Oct price is Oct 2nd close; Sept/Oct CPI estimated; Oct GS10 is Oct 2nd value.
  • Bitcoin: Daily close in nominal US dollars (not inflation-adjusted): the Coin Metrics reference rate at 00:00 UTC ending each day, so the row dated D is the price at midnight UTC between D and D+1. The earliest prices (2010-2012) come from very thin markets dominated by one exchange, Mt. Gox, with tiny volumes: single days can be noisy or stale (the price does not change from 2011-06-19 to 2011-06-25, while Mt. Gox was halted after a hack), so returns measured from those years are approximate.
  • Not investment advice: past booms and busts say nothing certain about future prices.