Right, Early, and Fired: The Limits of Arbitrage

Every finance textbook opens the arbitrage chapter the same way. Arbitrage requires no capital and involves no risk: you spot the same thing trading at two prices, buy the cheap one, sell the expensive one, and pocket the difference with certainty. Because such a lovely thing cannot last, arbitrageurs eliminate mispricing instantly, and that is why prices are correct.

Shleifer and Vishny open their 1997 paper by noting, drily, that in the real world almost all arbitrage requires capital and is risky. Then they spend the rest of the paper working out what follows from that one small correction to reality.

What follows is devastating, and it is the reason a generation of quants understands why their best trade lost the most money.

The problem: the textbook arbitrageur does not exist

The textbook imagines arbitrage as something done by "the market", a vast, patient, infinitely deep pool of rational money that will do any trade with a positive expected value.

Look at who actually does it. Real arbitrage is done by a small number of highly specialized professionals: hedge funds, prop desks, a handful of people who genuinely understand convertible bonds or merger spreads or the on-the-run Treasury basis. And these specialists are not deploying their own money. They are running other people's money.

That one institutional fact, that arbitrage is a delegated business, changes everything. Because now there are two brains in the system, not one. There is the arbitrageur, who understands the trade. And there is the investor in the arbitrageur, who does not.

The key idea via analogy: the coach who gets fired mid-comeback

Imagine you are a brilliant basketball coach with a strategy that reliably wins games, but always by starting slowly and dominating the fourth quarter. Your strategy is genuinely, provably correct.

Now add the owner. The owner does not understand basketball. He can only observe one thing: the scoreboard. And his rule is simple: if we are badly behind at halftime, you're fired.

Notice what happens. The worse you are losing at halftime, the more certain your fourth-quarter comeback becomes, and the more likely you are to be fired before it arrives. Your greatest opportunity and your greatest danger are the same moment. So what do you do? You stop running the strategy that works. You play it safe, protect the halftime score, and keep your job.

That is performance-based arbitrage, and it is the engine of the paper.

Here is the chain, and every link is real:

1. An arbitrageur spots a mispricing. Say an asset is 10 percent too cheap. He puts on the trade.
2. The mispricing gets worse. Noise traders get more excited, or a forced seller shows up, and the asset falls to 20 percent too cheap. The trade is now twice as attractive as when he entered. Any textbook says: back up the truck.
3. But he is now sitting on a loss. And his investors see the loss.
4. The investors cannot tell the difference between "temporarily wrong" and "just wrong". They are not experts, that is why they hired him. All they have is the scoreboard, and the scoreboard says he is losing. So they withdraw their money.
5. He must now sell the asset to fund the redemptions. He is forced to liquidate the very position that has just become the best opportunity of his career.
6. And his selling pushes the price down further, widening the mispricing, causing more losses, triggering more redemptions.

Read step 6 again. Arbitrage capital gets withdrawn exactly when it is most needed, and its withdrawal makes the problem worse. The mechanism that is supposed to correct mispricing instead amplifies it. Arbitrage is not a stabilizing force in extreme conditions. It is a destabilizing one.

And a rational arbitrageur, knowing all of this in advance, holds back. He does not take the full-size position when he first sees the 10 percent mispricing, because he knows he has to survive the path to 20 percent. He keeps dry powder. He sizes small. Which means the mispricing is never fully corrected, even at the start.

The mispricing survives not because nobody noticed, but because everybody who noticed was rationally afraid.

Why it mattered

• It explains why mispricing can be large, visible, and durable. You do not need to argue that professional investors are stupid or blind. You can concede they are brilliant, well-informed, and perfectly rational, and still get persistent mispricing, purely from the agency relationship between a manager and his capital. That is a much stronger argument than any behavioral one, because it survives the "but professionals aren't biased" rebuttal.
• It predicted LTCM. Long-Term Capital Management, run by Nobel laureates, held convergence trades that were essentially correct. In 1998 the spreads went the wrong way, capital fled, they were forced to liquidate into a market that knew they were liquidating, and the fund died. The trades subsequently converged. Shleifer and Vishny published the manual for this the year before it happened.
• It is the reason "the trade is more attractive now" is the most dangerous sentence in finance. It is often true. It is also often the sentence people say on the way to being carried out. This paper explains why both can be simultaneously correct.
• It reshaped how anomalies are studied. After 1997, the standard question about any anomaly became: why hasn't arbitrage killed this? And the standard answer became: because the trade requires holding a nasty, volatile, illiquid, hard-to-short position through drawdowns that no capital will finance. Anomalies now live precisely where arbitrage is hardest, which is a prediction you can test, and it holds up.
• It rehabilitates behavioral finance from its biggest objection. Critics of behavioral finance always say: even if some investors are irrational, arbitrage will fix the prices. Shleifer and Vishny answer: no, it will not, and here is the mechanism.

The honest limitations

• It is a theory of frictions, not of psychology. Note that nothing in the model requires anyone to be biased. The noise traders could be rational investors with liquidity needs. This is a strength (the result is robust) and a limitation (it does not prove that behavioral biases matter, only that arbitrage is weak).
• The model is stylized. A handful of periods, one mispriced asset, a simple withdrawal rule. Real capital flows are stickier and more complicated: lock-ups, gates, side pockets, and permanent capital vehicles all exist precisely to fight this problem, and the paper does not model them.
• It does not tell you how far things can go. The theory says mispricing can persist and widen. It does not say how much, or for how long, which is unfortunately the only thing you actually need to know when sizing the trade. It is a warning, not a calculator.
• The escape hatches are real. Since 1997 the industry has built defences: longer lock-ups, permanent capital, less leverage on convergence trades, explicit drawdown budgeting. The limits of arbitrage are real but they are not fixed constants, they are a function of your funding structure. The paper's deepest practical lesson is that your edge is worthless if your capital structure cannot survive holding it.

The one-line takeaway

Shleifer and Vishny showed that real arbitrage is done by specialists using other people's money, and because those backers pull their capital precisely when the trade is deepest underwater and most attractive, the smart money is forced to sell at the bottom, which means mispricing can persist, widen, and be made worse by the very people paid to fix it.