Stock Market Crashes: Every S&P 500 Fall Since 1950, Timed
A market crash is a sudden, severe fall in stock prices, measured in days or weeks rather than the months a slow decline takes. The S&P 500's worst single day since 1950 was 19 October 1987, when it closed 20.47% lower, and the quickest fall of 20% came in 2020.
People say “crash” about any bad week. The daily record of the S&P 500 lets the word be measured instead: how far prices fell, how quickly, and how long the damage took to undo.
How it works
A crash is a fall defined by its speed. No exchange, regulator or index provider publishes a crash threshold. The word is used when prices drop hard enough, and fast enough, that selling feeds on itself: stops trigger, margin calls force sales, and some funds cut risk by rule, all at the same time.
What turns a bad day into a crash is the absence of buyers at the old prices. Prices fall until someone is willing to take the other side, and in a panic that level can sit far below the last trade. Much of the damage happens between one close and the next, which is why a stop order placed below the market does not cap a loss at its price on a day like 19 October 1987.
The same word covers very different events. Some crashes are a single session. Others are a run of sessions stacked together over a few weeks. The site’s bear market page deals with the slow, label-driven version; this page is about the fast one.
Crash versus bear market
Speed is the difference that matters to a trader. A bear market is the size of a fall, usually 20% from a high. A crash is the pace. Nine of the S&P 500’s eleven falls of 20% or more since 1950 took between 111 and 310 trading days to first close 20% below the record. Two did it in weeks.
Those two are the crashes. In 1987 the index needed 38 trading days from its record close of 336.77 on 25 August to a close 20% below it, and most of that fall came in the final few sessions. In 2020 it needed 16 trading days, from 3,386.15 on 19 February to 12 March.
Market-wide trading halts exist for exactly these days. How they work, and at what levels, is on the circuit breaker page.
A worked example
Take a hypothetical $10,000 in a fund that tracked the S&P 500’s price exactly. On Friday 16 October 1987 the index closed at 282.70. On Monday 19 October it closed at 224.84. Divide the second by the first: 224.84 / 282.70 = 0.7953, so the $10,000 was worth about $7,953 one session later.
Now the 2020 version, spread over five weeks. The same $10,000 on the record close of 19 February 2020, 3,386.15, was worth about $6,608 on 23 March at 2,237.40, a fall of 33.9% in 23 trading days.
The climb back is bigger than the fall. From 2,237.40, getting back to 3,386.15 needs a rise of 51.3%, because the gain is measured from the lower base. The index closed at a new record on 18 August 2020, 103 trading days after the low.
None of this includes dividends or fund costs. It is price arithmetic on the index, which is the cleanest way to see how quickly a crash rearranges an account.
The original data
The data: every daily close of the S&P 500 from 3 Jan 1950 to 18 Sep 2026, 19,299 trading days, downloaded from Yahoo Finance (^GSPC), price only. Each fall of 20% or more is measured from a record close to the lowest close before the next record. There are 11. The timings below are counted in trading days from those dates, and the full table is published as a CSV of every fall and its timing.
The fastest falls were not the slowest to repair. The 2020 fall ran 33.9% in 23 trading days and the record came back after 126 trading days in all, about half a year. The 1987 fall ran 33.5% in 71 trading days, and the record came back on 26 July 1989, 485 trading days after the peak.
The long repairs followed long falls. From March 2000 the index fell for 637 trading days, 49.1% in all, and took about 7.2 years to reach a new record. From January 1973 the fall lasted 436 trading days, 48.2% deep, and the record took about 7.5 years. Eleven episodes are too few to call that a rule, and the 1980 to 1982 fall, 430 trading days long but repaired 58 trading days after its low, shows how loose the pattern is.
Single days tell the same story of clustering. Only 30 of the 19,299 sessions closed 5% or more below the day before, and 13 of those fell in 2000 to 2009 alone. Nine closed 7% or more lower; two closed 10% or more lower: 19 October 1987 at −20.47% and 16 March 2020 at −11.98%.
The ten worst days since 1950
Every one of the ten worst closes sits inside four episodes. Two came in October 1987, four between late September and early December 2008, three in March 2020 and one on 27 October 1997. Bad days arrive in groups, which is the reason a single ugly session is usually followed by more volatility rather than calm.
A year later, the index was higher after nine of the ten. Measured 252 trading days on, the close was up after every one except 29 September 2008, when it was still 4.1% lower a year afterward. The biggest rebound followed 16 March 2020: 66.1% higher a year later.
Those ten are not ten separate tests. Three came within five sessions of each other in 2020, and four within ten weeks in 2008, so the nine rebounds are really four episodes counted several times over. Read it as a description of what happened, not a rule about what follows a bad day.
When it fails
The first failure is treating a crash as a buying signal on its own. After 29 September 2008 the index was still lower a year later, and the October 2008 falls were followed by five more months of decline before the March 2009 low. Buying at the close of 29 September 2008 meant sitting through a further fall of 38.9% to the low.
The second is assuming a stop order sets the loss. A crash happens between prices. On 19 October 1987 the index closed 20.47% below the previous close, so an order to sell 5% lower would have filled far below its price.
A third is expecting every crash to heal like 2020. The half-year repair in 2020 is the fastest in the file. After 1987 the wait was under two years, but the slow falls of 1973 and 2000 took seven years and more, and there is no way to know on day one which kind has begun.
A fourth is leverage. A margined or leveraged position can be closed out at the bottom by the broker, which turns a temporary price fall into a permanent loss before any recovery arrives.
And a fifth is judging crashes from the index alone. A single stock or a small account can fall much further than the S&P 500 did, and the index figures here say nothing about one company’s recovery.
Related
The bear market page covers the slower 20% label and its recovery arithmetic, which is the other half of this story. The circuit breaker page explains the trading halts built for crash days. And drawdown shows how to measure a fall like these in your own account, from its peak rather than from what you paid.
Write down, before the next fast fall, what you will do if your account drops a third in a month. The decision is far easier to make calmly on a quiet day than on the third red day in a row.
— Michael Whitman
This page is educational, not financial advice. Test every idea on your own charts before risking money.