2. Crowding: when everyone holds the same position

Level 10: Why it runs out

The previous module described a slow erosion, over years. This one describes a collapse over days. It is worth not confusing the two mechanisms: a published idea wears out little by little; a crowded idea can cost you a third of the year in three sessions without having stopped being correct.

Being right is worth nothing if everyone is right

When you buy something because it is cheap, your profit depends on the price rising toward what you think it is worth. If a lot of people got there before you for the same reason, a good part of that rise has already been collected. That is the obvious part. What almost nobody anticipates is the other part: when many hold the same position for the same reason and with the same kind of money, it stops mattering who is right and starts mattering who can hold on.

Think of a room's capacity. A one-billion position spread across thousands of investors with no debt, with different horizons and needs, is unwound little by little and almost silently. The same position in the hands of twenty leveraged funds with similar risk rules is unwound all at once, because those rules trigger at more or less the same threshold. The door has not shrunk: what has changed is the number of people wanting to get through it in the same second.

The mechanics of the collapse, step by step

• Somebody large needs money for a reason unrelated to the strategy: losses on another desk, client redemptions, a margin call.
• They sell the most liquid thing they have, which tends to be precisely their equity portfolio, not the asset causing the problem.
• That selling pushes prices against the common position: the cheap gets cheaper still and the expensive gets dearer.
• The other funds see losses, their risk models measure more volatility and force them to cut exposure. They sell exactly the same thing.
• Each round of selling manufactures the loss that triggers the next round. The loop feeds itself until the forced seller has finished selling.

Note what does not appear in that list: any news about the companies. The price moves because of investors' balance sheets, not because of the companies' businesses. That is why, when the liquidation ends, prices usually come back: if the signals had stopped working, there would be no rebound.

Leverage as a multiplier

A market-neutral portfolio earns little per euro invested, so it is financed with borrowed money to make the final figure interesting. With 500 million of capital and gross exposure of 4 billion, a 3% adverse move in the unleveraged portfolio becomes a loss of 24% of capital.

And here is the subtle trap: risk models estimate volatility from the recent past. The calmer a strategy has been, the less risk the model measures, the more debt it allows and the bigger the blow when it comes. Calm is not proof of safety; it is, frequently, the condition that manufactures the disaster.

How to detect that an idea is crowded

No single signal is enough, but together they map it out fairly well:

• Portfolio overlap. The declared positions of funds in your same style look more and more like yours.
• Days of volume to get out. How many average trading sessions you would need to unwind your positions without shouting. If it is more than two or three, you no longer control your exit.
• Short interest and borrow cost. In the leg that has to be sold, a rising borrow cost indicates competition for the same slot.
• Money in the wrapper. Total assets managed under that same idea, including the index products replicating it. It is the crudest thermometer and one of the most useful.
• Compression of the valuation gap. The distance between what you buy and what you sell narrows. If the cheap is no longer that cheap, the future edge is smaller however good the past was.
• Conditional correlation. Measure your resemblance to peers on the worst days, not on average. A 0.2 in calm and a 0.9 in the bad decile is the exact signature of crowding.
• The shape of the gains. Winning a little and often, and losing a lot all at once, is the typical profile of someone paid for providing liquidity to others and withdrawing it all at the same time.

Real case: August 2007

During the week of August 6, 2007, a large group of market-neutral quantitative funds suffered the worst sessions in their history while the general index barely moved and, on some days, even rose. In Khandani and Lo's simulation of a standard contrarian strategy, losses piled up on consecutive days to around 11% in a single session, with a rebound of the same order of magnitude on the 10th. With the sector's usual leverage, that translated into high double-digit falls within days.

The damage was very concrete. A large investment bank's quantitative fund lost around 30% in a week and received a capital injection of close to 3 billion dollars. None of that came from the companies in the portfolio. The most widely accepted reconstruction is that somebody with problems in mortgage credit force-liquidated their equity portfolio, which turned out to be everyone's portfolio.

The rebound on the 10th is the key piece of evidence: if the signals had died, prices would not have come back. In January 2021 the same mechanics were seen from the short side, with a well-known US fund losing around half its assets in a month. The practical lesson is uncomfortable: spreading your money across several managers who do the same thing is not diversifying, it is concentrating with more bills to pay.