1. What the efficiency hypothesis says and what it does NOT say

Level 8: What is priced in and what is not

The efficient markets hypothesis has a bad name for a curious reason: almost nobody disputes what it says, people dispute a caricature of it. The caricature holds that a share price is always the correct one, that bubbles do not exist and that the market is a sort of infallible oracle. None of that appears in the original formulation. What Eugene Fama put forward in the late 1960s is a good deal more modest and a good deal harder to refute. And it is the obligatory starting point for anyone who wants to know where a real edge might still be left.

What the hypothesis claims, and what it does not

The useful version can be stated like this: prices reflect available information up to the point where exploiting what is left stops covering the cost of exploiting it. It is worth pausing on what that sentence does not say.

It does not say the price is correct. It says known information is already inside it. A price can incorporate absolutely all public information and still be completely wrong about the future, because the future is not yet information: it is uncertainty. That a company was worth a hundred and three years later is worth twenty does not prove the market was ignoring something then; it proves that things happened.

It does not say nobody ever wins. It says that winning systematically, repeatably and net of costs is difficult, and that a good share of those who manage it for a while manage it by chance or by collecting a risk they had not booked as such.

And it does not say the market is wise. It says something colder: that a lot of people are competing for the same thing, and that competition leaves little visible margin.

The three forms, and which one is actually in play

The hypothesis is formulated in three increasingly demanding degrees:

• Weak form. The price already incorporates the entire history of prices and volumes. Looking for patterns on a chart should not by itself produce a sustained edge.
• Semi-strong form. The price already incorporates all public information: accounts, news, reports, presentations. This is the version genuinely in play, and the one the serious debate is about.
• Strong form. The price incorporates even private information, the kind that has not yet left an office. This form is openly false, and there is an elegant proof of it: if inside information gave no edge, there would have been no need to ban it.

When someone says efficiency has been refuted, they are almost always refuting the strong form, which nobody defends, or confusing it with the idea that the price gets things right.

The joint hypothesis problem

Fama added an uncomfortable warning that is rarely quoted. To claim that a return is abnormal you need a model that says what the normal expected return was given the risk taken. So any test of efficiency is in reality a test of two things at once: of the market and of the model you measure it with.

The consequence is that no empirical result allows you to conclude cleanly that the market is inefficient. There is always the escape route that the excess return was compensation for a risk the model did not capture. This is not a debating trick: it is the reason the debate has been open for half a century and will remain open.

The Grossman and Stiglitz paradox

Here is the idea that turns all of the above into something operational. Sanford Grossman and Joseph Stiglitz set out, in 1980, an internal contradiction in perfect efficiency.

Suppose the price reflects all information. Then analysis is pointless, because the price already tells you everything you could find out. If analysis is pointless, nobody spends money or time analyzing. But if nobody analyzes, there is no mechanism putting the information into the price, and the price stops reflecting it. Perfect efficiency destroys itself.

Out of that comes the conclusion you need to internalize: a market where information costs money cannot be perfectly priced in. There has to be a margin. That margin is not a gift or a flaw in the system: it is the payment for the work of keeping prices informed. It is a salary, not a lottery.

How much has to be left over, and why no more

The size of that margin is not arbitrary. It works like a thermostat:

• If the return on analysis rises above its cost, more people and more capital come in to do the same thing, and the margin narrows.
• If it falls below cost, people abandon it, the price becomes less informative and the margin widens again.
• The equilibrium sits where the margin barely covers the cost of the marginal effort. Neither zero, nor a fortune.
• And the cost is not only money. It is time, it is patience, it is accepting uncomfortable positions and it is enduring bad runs. All of that gets paid for too.

Andrew Lo called this view the adaptive markets hypothesis: efficiency is not a switch that is on or off, but a degree that changes according to how much competition there is in each corner. A very crowded market is almost perfectly priced in; one nobody is looking at is not.

Real case: ten years of a public bet

In 2007 a ten-year bet was formalized, running from 2008 to 2017, between Warren Buffett and a manager specializing in selecting hedge funds. Buffett's thesis was simple: an index fund tracking the US market, with no selection and minimal fees, would beat a carefully chosen basket of funds of funds.

At the close in 2017, the index had accumulated around 125% and the basket of funds was around 36%. The difference is not explained by the managers being incompetent: it is explained by the thermostat. A great many capable people competing for the same margin, charging fees on that margin, until almost nothing reached the end client.

The periodic reports comparing active funds with their benchmark index point the same way: over ten- or fifteen-year horizons, most US large-cap equity funds come in below their index. The lesson is not that analysis is useless. It is that in the most crowded corner on the planet the margin is already spoken for.