The stock market has crashed many times, so there is no single date. In the United States, the most cited crashes began on October 24, 1929 (Black Thursday), October 19, 1987 (Black Monday), and March 16, 2020 (the COVID-19 panic). Each qualifies because a major index fell at least 20% from its peak, usually far faster than an ordinary decline.
Crashes are better defined by speed and depth than by a single day. A bear market is a 20% fall from a high; a crash is that fall compressed into days or weeks. Knowing the exact mechanism matters because it changes what an investor should do, and when.
What actually turns a decline into a crash
A crash is a liquidity and leverage event, not merely a change in sentiment. Prices are set at the margin by the last trade, so a modest imbalance of sellers over buyers can move an index sharply when order books thin out. Three mechanisms do most of the work.
First, leverage forces selling. When investors borrow against securities, a price drop triggers margin calls, and lenders demand cash or liquidation. Forced sales push prices lower, which triggers more margin calls. This feedback loop is why crashes cluster in hours rather than months.
Second, volatility and indexing amplify moves. Volatility-targeting funds and risk-parity strategies automatically cut exposure when volatility spikes, so they sell into falling markets. Market makers widen spreads and reduce inventory exactly when volume surges, removing the liquidity buyers need.
Third, information arrives all at once. A single shock, such as a bank failure or a pandemic, reprices the whole earnings outlook. Traders revalue every future cash flow in the same session, so the adjustment happens in one repricing rather than a gradual drift.
Plain language: crashes happen when forced sellers meet vanishing buyers, and both feed on each other.
The major US crashes, dated and quantified
The table below compares the sharpest episodes using peak-to-trough closing values on the S&P 500 and its predecessor indexes. Dates mark the start of the steepest phase, not the full decline.
| Episode | Key start date | Peak-to-trough decline | Approximate duration |
|---|---|---|---|
| Wall Street Crash | Oct 24, 1929 | About 86% (1929–1932) | Nearly 3 years |
| Black Monday | Oct 19, 1987 | About 34% in days | Weeks |
| Dot-com bust | Mar 2000 | About 49% | About 2.5 years |
| Global financial crisis | Sep 2008 | About 57% | About 17 months |
| COVID-19 crash | Mar 16, 2020 | About 34% | About 1 month |
On Black Monday, the Dow fell 22.6% in a single session, still the largest one-day percentage loss in its history. The 1987 crash was not driven by a recession; it was a valuation and program-trading event that reversed within two years. The 1929 crash, by contrast, coincided with bank failures and a policy contraction, which is why it lasted years.
The COVID-19 crash is the clearest modern example of speed. The S&P 500 fell about 34% from its February 19 peak to its March 23 trough, roughly 33 calendar days, then recovered its high by August 2020. Speed and depth are separate variables, and both matter for how a crash is defined.
Edge cases that complicate the label
Crashes do not always look like crashes. Some declines grind lower over many months and never produce a single catastrophic day. The 2000–2002 dot-com decline and the 1973–1974 bear market were slow bleeds, yet they erased more wealth than many one-day panics.
Intraday moves can qualify even when closing prices do not. The October 1987 crash included a record intraday drop, and the 2010 flash crash briefly erased nearly USD 1 trillion in market value within minutes before partially recovering (Securities and Exchange Commission and Commodity Futures Trading Commission joint report, 2010). A close-based definition would miss it.
Circuit breakers change the picture. US exchanges halt trading after a 7% decline from the prior close (Level 1), 13% (Level 2), and 20% (Level 3), with the first two halts lasting 15 minutes and the third closing the market for the day (Securities and Exchange Commission rules). These thresholds were triggered four times in March 2020 alone, which shows how quickly a routine decline can become a structured halt.
Professional caveat: crash statistics depend heavily on the index and the window you choose. The Dow, the S&P 500, and the Nasdaq Composite can tell different stories for the same week. Any single number is an artifact of methodology, not a fact about the market.
Misconceptions worth correcting
The most common error is treating a crash as a single day. Most crashes are multi-day processes, and the worst day often comes before the bottom. Buying the first big down day is not the same as buying the low.
Another misconception is that crashes always signal recession. The 1987 crash did not produce a recession, and the 2020 crash preceded a brief but severe contraction that reversed quickly. Markets and the real economy move on different clocks, and the gap can be months.
A third error is assuming crashes are unpredictable in principle. While timing is unreliable, risk builds predictably: rising leverage, compressed volatility, and stretched valuations all raise the odds. None of them tells you the date, which is the honest limit.
Red-flag warnings: Buying on margin to recover losses after a drop, which compounds forced-selling risk. Assuming a circuit-breaker halt means the bottom has arrived. Treating a single volatile day as a durable trend in either direction. Ignoring concentration risk when one sector drives most of an index's gains.
What the record implies for practical decisions
The historical record supports a few concrete actions. Keep enough cash or short-term Treasuries to cover 6–12 months of expenses so you never sell assets during a forced drawdown. Set a rebalancing band, commonly 5 percentage points from target allocation, and act on it mechanically rather than emotionally.
Define your own thresholds before a decline happens. A common framework treats a 10% fall as a correction (review, do not act), a 20% fall as a bear market (rebalance toward target), and a 30%-plus fall as a crisis (check liquidity, confirm no forced sales). Reconcile these explicitly: a 20% fall is both a bear-market trigger and a rebalancing signal, so decide which rule governs before the event, not during it.
For long-horizon investors, the data favor staying invested. Missing the best few days after a bottom can erase years of returns, and bottoms cannot be timed reliably. The cost of sitting out recovery days is one of the best-documented penalties in market history.
Limits of what the record can tell you
Every crash is a sample of one. The 1929, 1987, 2008, and 2020 episodes differ in cause, policy response, and market structure, so no rule guarantees the next one behaves the same. Statistics drawn from a handful of events carry wide error bars.
Survivorship and index changes also distort long-run comparisons. The S&P 500 has been reconstituted repeatedly, and failed companies drop out, which flatters historical index returns. Pre-1957 data rely on predecessor series with different construction.
Finally, past drawdowns do not set a ceiling on future ones. A 20% bear-market threshold is a labeling convention, not a physical law. Declines can stop at 15% or run past 50%, and the honest position is that depth and duration are uncertain in advance.
This article is general educational information, not an investment recommendation or a forecast. It does not promise any outcome or return. For decisions about your portfolio, consult a qualified financial professional who knows your circumstances.