Black Monday 1987 and the machines — a crash with no news
The answer
On Monday 19 October 1987 the Dow Jones Industrial Average fell 508 points, or 22.6%, the largest single-day percentage fall in its history. There was no war, no bankruptcy, no policy announcement large enough to explain it. The crash was produced mostly by the structure of the market itself.
Why this costs you money
Most investors protect themselves with an instruction that sounds sensible: "if it falls to a certain price, sell."
The instruction is fine. The price is usually the problem, because almost everybody picks the same kind of number. Round numbers. ₹1,000. ₹500. $50. The level from a chart that thousands of people are looking at. The 200-day moving average. Exactly 10% below the purchase price.
When many people place the same instruction at the same level, that level stops being protection. It becomes a supply of guaranteed sell orders sitting in one place. Price arrives, the orders fire together, and the price that the sellers actually receive is far below the level they chose.
This is not a theory about 1987. It is what a stop order does on any bad day, in any market, and it is why the difference between the price you set and the price you get is largest exactly when you needed the protection most.
The cost is specific: you take the loss you planned, plus an extra amount you did not plan, because you were standing where everyone else was standing.
How it works
The 1987 mechanism has a name: portfolio insurance.
The idea was reasonable. A large fund wants to limit its losses. Rather than buy protection from somebody else, it follows a rule: as the market falls, sell a portion of the holdings; as it rises, buy back. Done continuously, this approximates the payoff of an insurance contract without paying a premium for one.
Now notice the 3 properties that made it dangerous.
1. The rule is automatic. It does not ask whether the fall is justified. It sells because the price fell.
2. Many funds followed the same rule. Portfolio insurance was sold as a product by a small number of firms to a large number of institutions. Estimates put the assets following such strategies in 1987 at $60 billion to $90 billion .
3. The selling was concentrated in one instrument. Rather than sell thousands of individual shares, the strategies sold stock index futures, because futures were cheaper and faster. Selling futures pushed futures prices below the index. Other traders then bought futures and sold the underlying shares to capture the gap, which pushed the shares down, which told the portfolio insurance programs to sell more futures.
That is a loop, and here is the crucial property: the loop needs no news to run. Each step is a rational response to the previous step.
Two more things made 19 October worse.
The systems for processing orders were overwhelmed. Quotes on the ticker ran far behind actual trades, so buyers could not tell what the real price was. When you cannot see the price, you stop bidding. Buyers withdrew, not from panic, but from blindness.
And the derivatives markets and the share markets stopped agreeing. At points during the day, futures traded at a very large discount to the shares. That gap is a signal that 1 of the 2 markets has stopped functioning.
The Brady Report of January 1988 recommended coordinated circuit breakers across markets, unified clearing, and a single regulator with an overall view. Circuit breakers were introduced in 1988, and every market-wide trading halt you have seen since is a descendant of that day.
What it tells you, and what it does not
What it tells you. Crashes do not require a cause in the ordinary sense. A market can break because too many participants hold the same instruction. This is a structural risk, and it is invisible in any analysis of companies.
It also tells you something specific about protection. Any protection that requires you to sell into a falling market is not protection. It is a promise to become a seller at the worst moment, along with everyone else who made the same promise.
What it does not tell you. It does not tell you that portfolio insurance alone caused the crash, and this is where the popular account overreaches. The Chicago Mercantile Exchange's own study argued that other investors — mutual funds, dealers, individuals — sold 3 to 5 times as much as the portfolio insurers. And here is the fact that most accounts leave out: markets that had no portfolio insurance at all crashed too, and some crashed harder. Hong Kong fell about 45.8%, Australia over 40%, the United Kingdom about 26% over 2 days, and New Zealand roughly 60% by its 1988 trough. If computerised selling in New York were the sole cause, Hong Kong falling further is hard to explain.
So the honest statement is narrower than the legend. Portfolio insurance amplified and accelerated a fall that was global and already under way. That is still the lesson — amplification by identical rules — but "computers caused it" is not supportable.
The decision rule
A stop is only protection if the price you choose is not the price everybody else chose.
Practically, this gives 3 conditional rules.
If the level you have chosen is a round number, a widely published chart level, or exactly 10% below your entry, move it. Not far — a few percent below the crowd's level is usually enough, because the cascade happens at the crowd's level.
If you use a stop order that becomes a market order when triggered, understand that you have accepted whatever price exists at that moment. In a cascade, that price can be far away. A stop with a limit avoids the bad fill but may not execute at all. There is no free version. Choose which failure you prefer, in advance.
And if a position is large relative to the stock's daily volume, a stop is close to useless, because your own order becomes part of the cascade. For those positions, the protection is a smaller position, not a cleverer order.
Try this now
Five minutes, on your own open positions.
- Open your orders. List every stop, stop-loss, stop-limit or GTT (good-till-triggered) order you have placed, with its trigger price.
- Look at each trigger price and ask: is this a round number? Anything ending in 00 or 50 counts. Anything that is exactly 5%, 10%, 15% or 20% below your buying price counts too.
- For each round-number level, open the stock's chart and look at what is there. Is that level a visible previous low, a previous high, or a level that any chart-reading person would mark? If yes, you are standing in a crowd.
- Now do the test that makes it real. Look at the stock's average daily traded volume, then look at the depth of the order book — the quantity of buy orders sitting below the current price, which most apps show as market depth. Add up the buy quantity in the 5 levels below the price.
- Compare that total to your own position size.
What you should see. Most readers find at least 1 stop at a round number, and usually more than 1. Round numbers are not chosen deliberately; they are chosen because they are easy to say.
Step 4 usually produces the second surprise. In a mid-sized or small company, the visible buying support a few levels below the price is often small — smaller than a single serious seller. On a calm day this does not matter, because new orders arrive constantly. On a bad day, they do not arrive, and the depth you looked at is roughly what is actually there.
Then do one thing: for any stop at a round number, move it. Down a little if you are protecting a long position. The point is not precision. The point is not being in the queue.
Three real cases
1. The United States, 19 October 1987 — identical rules, one instrument The Dow fell 22.6% in a day, the S&P 500 fell about 20.5%, and the Dow did not regain its August 1987 high until 26 July 1989. What it turned out to be: not a repeat of 1929. The US economy grew in 1988. A 22.6% single-day fall with no recession afterwards is the strongest evidence that this was a market mechanism failure rather than an economic event.
2. The United States, 6 May 2010, the "flash crash" — the same shape, 23 years later Major US indices fell about 9% and largely recovered within roughly 30 minutes . Some individual shares traded at 1 cent and others at $100,000 before those trades were cancelled. The joint SEC and CFTC report attributed the start to a large automated sell order in index futures interacting with high-frequency traders who withdrew. What it turned out to be: proof that circuit breakers designed for a slow cascade needed a version for a fast one. The result was the limit up-limit down mechanism for individual stocks.
3. India, 6 October 2012, the NSE "freak trade" episode — one order, one market A brokerage entered erroneous orders in NIFTY constituent shares, and the index fell roughly 15% within seconds before trading was halted. Trading resumed after about 15 minutes. What it turned out to be: a demonstration that a single mistaken instruction can hit the circuit breaker in a large, liquid, modern market. SEBI subsequently tightened order-level risk controls and price-band checks for brokers.
The question that resolves it
A novice looks at a stop order and asks: how much am I willing to lose?
An expert asks the same question, and then a second one: where is everybody else's answer?
The first question sets your loss. The second question determines whether you actually get it.
What would make this wrong
The falsification condition. If stop orders clustered at obvious levels did not worsen fills, then the price received on triggered stops would be as close to the trigger on volatile days as on calm days. It is not. Any reader can verify this on their own trade history by comparing trigger price to executed price on the worst days.
The honest limit, and it matters. This article is not an argument against stops. An investor with no exit rule at all usually does worse, because they substitute hope. The claim is narrower: the level is a crowding decision, and almost nobody treats it as one.
There is a second limit. In a market-wide cascade, moving your stop a few percent does not save you. It only helps against the ordinary version of the problem — a single stock falling through a well-known level on an ordinary day. That is the common case, which is why it is worth fixing, but you should not believe it protects you on a 1987.
The hindsight trap. It is tempting to say that portfolio insurance was obviously dangerous. Look at what was actually knowable in 1986.
The strategy had a solid mathematical basis, derived from option pricing theory published a decade earlier that had already won wide acceptance. It had worked in live use for several years. It was sold by respected firms to sophisticated institutions with risk committees. And its core assumption — that you can always sell a small amount at close to the current price — had been true in every market anybody had traded in.
That assumption is where the failure lived, and it was stateable in advance. A careful person could have asked 1 question: what happens if everybody using this strategy needs to sell at the same moment? The answer follows from the strategy's own description. It requires a buyer. It does not create one. If the users are large relative to the market, they are trading against themselves.
Some people did ask. There were warnings in 1986 and 1987 about the size of portfolio insurance relative to market liquidity. They were not secret and they were not popular.
What a careful person could not have seen was the date, the scale, or the fact that the same day would produce larger falls in markets without any of the machinery. Anybody who says the size of the fall was foreseeable should be asked why Hong Kong, with no portfolio insurance, fell twice as far.
In India
India's protection against this failure is a set of rules that most investors have seen without knowing what they are for.
Market-wide circuit breakers. SEBI's index-based framework halts trading across the equity and derivatives markets if the NIFTY 50 or the SENSEX moves 10%, 15% or 20% from the previous close. The halt length depends on the level and the time of day, and there is a pre-open call auction when trading resumes . This is the direct descendant of the Brady Report.
Individual price bands. Most shares have a daily band — commonly 5%, 10% or 20% — beyond which the price cannot move that day. A stock "at upper circuit" or "at lower circuit" has not settled. It has run out of permission. Dynamic price bands apply a narrower intraday band, flexed by the exchange during the day, to stop instantaneous freak moves.
A practical consequence for stops in India. If a stock hits its lower circuit, there are no buyers at any price within the band. A stop order does not execute. It sits, and executes the next day, lower. Anybody holding a small or mid-sized Indian company should assume their stop does not work in the situation the stop was bought for.
In the United States
The US system has 2 layers and it is worth knowing both.
Market-wide circuit breakers, based on the S&P 500. Level 1 is a 7% fall, Level 2 is 13%, both producing a 15-minute halt if they occur before 3:25 p.m. Eastern time. Level 3 is a 20% fall, which closes the market for the rest of the day. These thresholds were revised after 2010; the original 1988 version used Dow point levels, which became meaningless as the index rose.
Limit up-limit down (LULD) for individual securities, introduced after the 2010 flash crash. A stock cannot trade outside a band around its recent average price. If it stays outside for 15 seconds, there is a 5-minute pause. This is the rule that stopped the 1-cent trades of 2010 from recurring.
Order type consequences. US brokers generally support stop, stop-limit and trailing stop orders. Many US brokers publish guidance recommending against plain market orders around the open and close, for exactly the reason this article describes.
Where they differ, and what that tells you
The 2 systems stop different things, and the difference tells you where to be careful.
India's individual price bands stop a stock from moving more than a fixed percentage in a day. The United States has no equivalent daily cap for most stocks; LULD stops instantaneous moves but a US stock can fall 40% in a day through a series of legal price steps.
So the failure modes are different. In the United States, your stop will execute, and it may execute a long way below your trigger. In India, on a stock at its lower circuit, your stop will not execute at all, and you will watch the position for days without being able to leave.
What that tells you is that "I have a stop" means 2 different things in the 2 countries. In the United States, the risk to plan for is a bad fill: use a limit on the stop, and accept the risk of not executing. In India, the risk to plan for is no fill at all: the real protection is position size relative to the stock's liquidity, checked before you buy, because after the circuit hits there is nothing to do.
Neither system protects you from the 1987 problem, which was everybody holding the same instruction. Both systems only buy time.
Carry this
- A stop at a round number is a queue, not a plan.
- Any protection that requires selling into a fall is a promise to sell at the worst price.
- In India the circuit stops your exit. In the United States it only slows it.