Why Stock Markets Crash, and What the Record Actually Shows

The part that cannot be answered, and the part that can

There is no shortage of explanations for why markets fall. They arrive in quantity after each episode, they are fluent, and they are almost never available beforehand. A reason discovered afterwards is not a reason anybody could have acted on.

So this page splits the question in two. The first half is mechanism: what has to be true of a market for a crash to be possible at all, which is a question about structure rather than about any particular fall. The second half is the record — what falls in the Indian market have actually looked like, measured across 27.2 years from 30 June 1999 to 31 August 2026. Only the second half produces numbers, and they turn out to be more useful than any of the narratives.

What a price is, and why that makes falls inevitable

A share price is not a measurement of a company. It is the amount at which the most eager buyer and the most willing seller last agreed, and it is then quoted as though it applied to every share in existence.

That last step is the whole of it. A price is set by the small fraction of holders who traded today, and then applied to everyone who did not. In ordinary conditions this is a harmless convenience. When many holders want to sell at once it stops being harmless, because the buyers who were willing at yesterday's price are quickly used up and the price has to keep falling until it reaches people who will buy.

Three things then make matters worse rather than better, and all three are structural rather than particular to any one episode. Some holders have borrowed against what they own and must sell when its value falls, so the fall itself manufactures more sellers. Some hold on behalf of other people who can ask for their money back at any time, which produces the same effect by a different route. And a falling price is itself information: it tells everyone watching that others are selling, which is a reason to sell. None of this requires anybody to behave irrationally, and none of it can be regulated away without also removing the market.

This is why falls are a permanent feature rather than a malfunction, and it is also why the explanations only ever arrive afterwards. The trigger can be small, because the mechanism supplies the rest of the fall by itself.

How often, and how deep

That is as far as reasoning gets. The rest is the record.

High point	Low point	Fall	Time falling	Time to get back
11 February 2000	21 September 2001	-50.2%	1.6 years	2.2 years
14 January 2004	17 May 2004	-29.8%	4 months	6 months
10 May 2006	14 June 2006	-29.7%	35 days	4 months
8 January 2008	27 October 2008	-59.5%	10 months	1.9 years
5 November 2010	20 December 2011	-27.2%	13 months	17 months
3 March 2015	25 February 2016	-21.7%	12 months	6 months
14 January 2020	23 March 2020	-38.3%	2 months	7 months

That is 7 falls of 20% or more in 27.2 years, or roughly one every 3.9 years. Lower the bar to a fall of 10% and the count rises to 19, about one every 1.4 years. The deepest of them took 59.50% off the index between 8 January 2008 and 27 October 2008.

The last column is the one to sit with, and it needs reading carefully. Recovery there is measured from the low point, and nobody can identify a low point at the time. Somebody who was invested at the high — which is to say almost everybody, because a high is when the most money is in — waited the fall and the recovery added together. That total ran from 5 months at the mildest to 3.8 years at the worst, with a middle case of 1.5 years.

There is no useful average in that list. The honest answer to "how long does it take to recover" is that it has ranged from a few months to nearly four years, and nothing told them apart in advance. A plan that needs a recovery to arrive within some particular period is not a plan this record supports.

The number nobody quotes

Counting crashes understates the experience, because it treats everything in between as normal. It was not. Below is every trading day in the record, measured against the highest point the market had reached by that day.

On a given trading day the market was	Share of all days
Below its previous high at all	91.3%
More than 5% below it	57.1%
More than 10% below it	38.5%
More than 20% below it	20.4%
More than 30% below it	12.4%
More than 40% below it	4.7%
More than 50% below it	1.5%

The market was below some earlier high on 91.3% of all trading days. It was more than 10% below on 38.5% of them and more than 20% below on 20.4%. The typical day sat 6.78% beneath a level the market had already touched.

Show these numbers as a table

How far below its previous best the market was	Number of periods	Share
-59.5% to -56.5%	21	0.3%
-56.5% to -53.5%	45	0.7%
-53.5% to -50.6%	35	0.5%
-50.6% to -47.6%	16	0.2%
-47.6% to -44.6%	27	0.4%
-44.6% to -41.6%	116	1.7%
-41.6% to -38.7%	104	1.5%
-38.7% to -35.7%	211	3.1%
-35.7% to -32.7%	149	2.2%
-32.7% to -29.7%	122	1.8%
-29.7% to -26.8%	120	1.8%
-26.8% to -23.8%	161	2.4%
-23.8% to -20.8%	200	3.0%
-20.8% to -17.8%	248	3.7%
-17.8% to -14.9%	323	4.8%
-14.9% to -11.9%	402	5.9%
-11.9% to -8.9%	523	7.7%
-8.9% to -5.9%	779	11.5%
-5.9% to -3.0%	981	14.5%
-3.0% to +0.0%	2176	32.2%

Set that beside what the same record returned overall and the two facts sit oddly together until you accept both. This is the market used for years to argue that shares are the best long-run home for money, and it spent roughly nine days in ten below its own previous best. Those are not competing findings. Being below a previous high is the ordinary condition of a rising market, not a sign that something has gone wrong.

It also accounts for a particular kind of unhappiness. An investor who judges a portfolio against its peak value is, on this evidence, going to feel behind almost every time they look, however well the investment is actually doing. That feeling is produced by the comparison rather than by the outcome.

The record ends inside one

At the time of writing the market has not regained the high it reached on 2 January 2026. At its lowest point since, it stood 15.08% below it.

That is worth stating rather than quietly leaving out, because it is the ordinary situation. Every table of completed falls is a table of episodes that finished, assembled by somebody standing safely on the far side of all of them. Nobody living through any one of them had that table. Whether the current fall becomes another line in it, or something worse, is not knowable from this data and is not guessed at here.

What this cannot tell you

7 events is not a sample. Every statistic here about the depth or duration of large falls rests on that handful of observations from a single market. That is enough to establish that a fall of 59.50% is possible in India. It is nowhere near enough to say how likely one is in a given decade.

It is one country over one stretch. India grew quickly across this period. A record from a market whose economy disappointed would look different, and the people inside it had no way of knowing which kind they were in. The same caution is set out at greater length in why long-term equity returns remain uncertain.

The figures include dividends. These are total-return figures, so they recover somewhat faster than the headline index a reader sees quoted. Measured on price alone, each fall would look a little deeper and take a little longer to undo.

Nothing here predicts anything. The record describes falls that happened. It contains no signal for when the next one begins, and the attempt to find one is tested directly, and fails, in why market timing is unreliable.

What to take away

Falls are not a defect in the market. They are a consequence of how a price gets made, they have arrived in this record every few years, they have gone deep, and the wait to get back to level has been anything from months to years.

The useful response is not to predict them but to arrange your money so that one arriving at a bad moment does not force a decision. That means knowing what you will need within the next few years and keeping it somewhere a fall cannot reach, so the money in shares is money you can genuinely leave alone — the argument made in full in how to choose an asset allocation.

And when you next check your portfolio and find it below its best, remember what the second table says. That is not the exception. That is nine days out of ten.