Why did a published factor stop working in August 2007?
A factor is a rule, and a rule anyone can read is a rule many people follow. Between Monday 6 and Thursday 9 August 2007 the quantitative equity funds following the same rules lost, in some cases, a quarter of their capital in three days — and the market barely moved. That week is the definition of crowding, and this lesson is what it teaches about capacity.
What happened
Khandani and Lo reconstructed it within weeks, from returns and transaction data rather than from any fund's disclosed positions. Their inference: one or more sizeable equity market-neutral portfolios were unwound in a hurry — plausibly a multi-strategy fund or a proprietary desk raising cash against losses elsewhere — selling their longs and buying back their shorts, with the unwinding apparently starting in financial shares and running through books that were long book-to-market and short earnings momentum. Those longs were everyone's longs and those shorts everyone's shorts, because everyone was running the same published factors. The selling pushed the longs down and the shorts up, which produced losses at every other fund holding the same book, which forced further unwinding. Value, momentum and quality spreads all went sharply the wrong way at once, for reasons that had nothing to do with value, momentum or quality.
By the following week most of it had reversed. A fund that held on recovered; a fund that had to cut at the bottom did not.
Invisible from the index
The market's own record of that week is nearly blank. Measured on 2026-09-04 from adjusted closes, SPY.US printed an adjusted close of 101.40 on Friday 3 August, 102.52 on Thursday 9 August and 102.41 the following Monday; the value fund IWD.US went 52.91, 53.56, 53.60; the growth fund IWF.US 11.71, 11.84, 11.82. Up a percent, then flat. A crisis in long-short factor portfolios left no mark on the long-only funds that hold one side of each — which is the clearest demonstration in the record that a factor spread is a different asset from the fund that carries its name.
Capacity
Every factor has a capacity: the amount of money that can follow the rule before the following changes the prices the rule depends on. Small-cap value has a small one, because the shares are small. Momentum's is limited by turnover — a strategy that replaces its whole book yearly moves prices every time it trades. Large-cap value and low volatility have large capacities, and correspondingly smaller premia, because the money already there has taken the easy part.
Crowding is capacity exceeded: not one fund too large, but many funds with the same rule, so that a shock to any one is transmitted through prices to all. The tell is correlation between strategies that should be independent, and it is only visible in a crisis, which is the wrong time to learn it.
Why a published factor decays
Put the two together and the decay McLean and Pontiff measured has a mechanism. Publication brings money; money reduces the premium by pushing the cheap group's prices up; and beyond a certain scale the money makes the factor a crowded trade whose worst weeks are not in the backtest. A factor's published Sharpe ratio was earned by a portfolio nobody else was holding.
In the data
The same window, 3 to 13 August 2007, for the S&P 500 fund and the small-cap fund, as traded (the adjusted figures in the prose above are rescaled after every later dividend; the day-to-day moves are the same):
The crowding itself is in no public price or filing; it lived in the positions of funds that did not publish them. What public data can show is capacity's shadow: the size of the companies in a small-cap sort, and the daily volume in them that a rule has to trade against.
Try it now
From the two tables above, write down the largest daily move in either. Then explain how a week that looks like that in the index could be the worst week in a decade for a factor fund.
The small-cap fund at full history, where the week is invisible and the 2008 that followed is not:
Measure August 2007 and then September 2008 to March 2009. The first is crowding; the second is the market. Only one of them shows.