Margin Call Explained: Why a Crash Rebounds With No News

A margin call is a broker's demand: add cash or close part of the position, because your own equity has dropped below the line. It looks like a problem with one account. The thresholds, though, sit in roughly the same place for everyone, so a sharp fall sends the calls out in bulk and puts sellers on the tape who do not want to sell. Most crashes measured in days and weeks are exactly that: a forced clean-out of positions bought with borrowed money, not a change in the fundamentals.

What a margin call actually is#

A mortgage is leverage too. You put down a hundred thousand, buy a flat worth a million, the bank lends the rest, and nobody treats it as reckless. One difference matters: the bank does not revalue your flat every evening after the close, and it cannot put it up for auction on Wednesday morning.

A broker can.

The loan from a broker sits on shares it can see, price every second and sell itself. So the agreement says nothing about a repayment date and everything about a proportion: how much of the account has to be yours.

Two numbers set that proportion. At the purchase, a firm may lend up to half the price of the stock, which means a hundred thousand of your own carries two hundred thousand of exposure. After the purchase, your equity may not fall below 25% of the current market value of those shares, under rule 4210 of the US market regulator: "25 percent of the current market value of all margin securities (...) 'long' in the account". Below that level, the call arrives.

A call has two answers: you add cash, or you close part of the position. The third is that you do nothing, in which case the broker does it for you. "Firms don't have to issue a margin call before selling securities in your margin account to meet a margin call," the same regulator writes. Your consent is not required, because you signed it when you opened the account.

There is one more asymmetry in that agreement, and it is easy to miss. Share prices move every day. A loan does not move at all. The entire fall therefore comes out of your equity, to the penny, and the buffer melts faster than the chart suggests.

One hundred, eighty, sixty#

How far does the price have to fall before that phone rings? Say you hold round numbers. A hundred thousand of your own, another hundred thousand borrowed, two hundred thousand of stock, which puts your equity at 50%.

The price gives up 20%. Now the account is worth 160,000 against a loan still fixed at 100,000, so 60,000 is yours, which is 37.5%. Above the line, with the comfort gone.

Down 40% from the top, and the arithmetic turns. Your account is worth 120,000, the loan is still 100,000, and 20,000 of your capital is left, which is 16.7%.

The phone rings.

The same three levels in one table:

Look at what happened to the money along the way. The price fell 40% and you lost four fifths of what you put in. Two-to-one leverage does not double the fall in the price. It doubles the speed at which your buffer disappears, because every percent on the quote is two percent of your capital.

Why you sell three times more#

There are two roads back over the line and they cost wildly different amounts. That gap decides the size of the sell-off across the whole market, and almost nobody does the arithmetic.

Cash lands on both sides of the fraction: it adds to your equity and to the value of the account. A sale adds nothing to your equity at all, because it shrinks the stock and the loan by the same amount.

With an account worth 120,000 and a loan of 100,000, you clear 25% again by adding roughly 13,000 in cash or by selling 40,000 of stock. Which means one dollar of cash does the work of three dollars raised by selling, and a third of the account goes to market at the worst price of the cycle.

Whoever has cash pays thirteen thousand and keeps the whole position. Whoever does not hands over forty thousand of stock. And whoever took the loan usually has no cash, because if they had it, they would not have taken the loan.

Where the loop comes from#

All of that was still one account's arithmetic. The market turns it into something else, because there are millions of accounts and no two thresholds sit in the same place.

They scatter for two reasons. Everyone bought at a different price, so everyone enters the fall with a different buffer. On top of that, the regulator sets only a floor, and brokerage firms routinely require more, each at its own level. Calls therefore do not arrive at once. They spread down the price scale, one level under the next.

The rest writes itself. Forced selling at the first level knocks the price down to where the second account gets its call. That one sells, the price steps lower, the third comes in. Supply here is not produced by new information about the business. It is produced by the price, which manufactures its own next seller.

I know of no other mechanism in the market that feeds itself this precisely.

Which is why the news in a week like that reads absurdly. Everyone hunts for the reason a stock is down a third straight day when nothing has happened, and settles on whatever is available: politics, rates, a war, the weather in the Gulf of Mexico. The real reason is duller, and duller stories get written less often. What else moves a quote outside results season I take apart separately, in the piece on why share prices move between earnings.

How to spot forced selling#

So how do you know you are watching a loop rather than a deserved fall? You do not get certainty, but you do get three tells, and none of them needs data you cannot reach.

The first is the missing document. A fundamental fall usually leaves paper with a date on it: a filing, cut guidance, a note about a lost customer. If the investor relations page is silent through the window of the fall, the information the market is supposedly discounting does not exist anywhere you could check it.

The second is the company it keeps. Forced selling does not pick stocks by the quality of the business, it picks them by who was holding them, so names with nothing in common go down together apart from one thing: they sat on the same leveraged accounts. When a software vendor and a copper miner give up the same percentage in the same week, their results are not what they have in common.

The third is the shape of the day. A broker recalculates collateral after the close and gives a defined window to respond, so supply from calls returns in waves at similar hours. A fall on published results looks different: it happens in one minute and then goes quiet.

None of these is proof. Three at once is a strong hypothesis, and telling a hypothesis from a certainty is the only thing you genuinely control in that week.

How much fuel the loop has#

How big is the tank? It can be measured, because the US market regulator publishes the total of debit balances on brokerage accounts every month. It is one of the few numbers about market positioning that is nobody's estimate.

In July 2025 those loans stood at 1.02 trillion dollars. By June 2026 they reached 1.50 trillion, according to the monthly FINRA margin statistics. Which means borrowed capital on customer accounts grew by close to half in eleven months.

A month later the balance had fallen back to 1.42 trillion. Eighty-five billion dollars of loans disappeared in four weeks, which means collateral of that size had to be turned into cash by somebody, willingly or otherwise.

Read that series historically, never at face value. The level on its own says nothing, because it grows with the value of the market. The pace says plenty, and eleven months of growing by half followed by a month of retreat describes a market that borrowed first and had to repay afterwards.

Why the rebound arrives with no news#

If the loop has a tank, it also has a bottom. A forced seller is an exhaustible resource: the moment the position closes, they leave the market for good, because there is nothing left to liquidate.

At some point the last leveraged account is cleared and the supply that never asked about price simply stops. A rise then needs no good news. It needs the absence of sellers who have to sell.

Hence the V on the chart, and hence the feeling that the price bounced for no reason. There was a reason. It had nothing to do with the company. A second mechanism of the same family, the closing of short positions, adds buyers from the same category to the same move: the ones who buy because they must.

Where this story breaks#

We could stop here in theory. I have two objections to my own thesis and both are serious.

First: the low usually does have a date and a headline attached. On 9 April 2025 the S&P 500 rose 9.5% in a single session, having entered the day nearly 19% below its record set less than two months earlier. The rebound followed the announcement of a pause on part of the tariffs, not silence. I do not know how much of that session was an exhausted loop and how much was the announcement, and I doubt anyone can separate the two. The mechanism explains amplitude and speed. It does not explain the date.

Second: sometimes the fall is entirely deserved and leverage merely gives it speed. When a company is losing customers or drifting towards a covenant, forced selling is not the explanation, it is an addition to the explanation. The loop does not turn a weak business into a good one, and there is no reason for the price to return to where it started.

"It's only mechanics" is therefore a hypothesis to be checked, not a consolation.

What to check while the loop runs#

Since the price carries no information about the business in a week like that, you check what does. Three things, in this order:

1. Whether the company published anything during the window of the fall that changes its numbers.
2. Whether its own balance sheet survives a worse quarter: net debt, maturities, covenants. How to read one I take apart separately.
3. Which assumption of your thesis this week actually broke.

The third point is often the shortest, because the answer is "none". That is an answer, not the absence of one. The reflex after a heavy session I cover in more detail in the piece on what to do when a stock drops 20%.

In Taufolio this work sits in two places. The Investment thesis holds the reasons you own the company, broken into assumptions with a measurable condition attached, so in a week of forced selling there is somewhere to check whether one of them genuinely broke. Today's move answers the second question the same day: whether the portfolio was moved by one company or by something broader that hits everything at once. What that looks like on screen is on the product page.

The loop always ends in the same place: when the last forced seller has finished selling. The date that happens cannot be worked out in advance and I do not try. What can be worked out today, long before the next crash, is whether the companies in your portfolio walk into a week like that carrying leverage of their own. That single number decides whether forced selling is a story about investors' accounts or about the business as well.