The Efficient Market Hypothesis: What It Actually Forbids

The efficient market hypothesis gets argued about more than it gets read. It is produced as proof that chart reading is worthless, that index funds are the only rational choice, and that anyone who beat the market was lucky. It is produced just as often to prove the opposite, usually by pointing at a crash.

Both sides are frequently arguing about a claim the hypothesis never makes.

What follows separates the three versions of the hypothesis, each of which sets a different bar, from the conclusions attached to them afterwards. It also does two things the widely read explainers do not.

It checks what the 2013 economics prize was actually awarded for, against the announcement that awarded it. And it shows where weak-form efficiency stopped being a theory and became a disclosure rule that binds the firms this audience reads every day.

Which Claim Are You Actually Arguing With

The hypothesis is a statement about information. It holds that a defined body of information is already contained in the current price, so that body of information cannot be used to earn a return beyond what the risk carried justifies.

Read what that leaves alone. It does not say the price is right. It does not say the price will not fall a long way, or that it will not look absurd afterwards. It does not say returns are unavailable. It says one specific input has already been used.

Four separate propositions get carried under the same name, and only the first belongs to the hypothesis:

• The current price already reflects information set X.
• Nobody can beat the market.
• Prices are correct valuations.
• Markets behave sensibly.

The second needs trading costs added before it follows from the first, and the arithmetic does the work rather than the theory. The third is a much stronger claim about accuracy that the hypothesis never advances; a market can absorb every scrap of available information and still be wrong, because the information itself was wrong. The fourth is a description of temperament and belongs to a different argument.

Someone who says the theory is nonsense because of a bubble is refuting the third proposition. Someone who says it is obviously true because most funds trail the index is defending the second. Neither has touched the hypothesis.

The premise that the market discounts everything was in circulation as the first element of Dow Theory long before the hypothesis was formalised, which is part of why the idea feels older and vaguer than the testable statement actually is.

Before disagreeing with it, settle which of the four you mean. The rest of this page treats only the first.

Three Forms, Three Different Bars

The three forms are usually presented as a ladder of strength. They are more useful read as three different tests, because each one names a different information set and is therefore refuted by different evidence.

Weak form covers the price record itself: past prices, past returns, past volume. Its bar is that no rule built only out of price history earns a return above the risk it takes. Evidence against it is a rule using nothing but historical prices that keeps working after costs.

Semi-strong form adds everything public: filings, announcements, economic releases, news. Its bar is that public information is already priced by the time it reaches you, so acting on a published number is acting late. Evidence against it is a durable return earned from information anyone could have read.

Strong form adds private information as well, and claims even that confers no advantage. It is the version almost nobody defends, and the existence of insider dealing law is the practical acknowledgement that private information is worth something.

Reading them this way makes the common shortcut visible. A claim that the hypothesis is dead usually rests on evidence against the strong form, which was never widely held, and is then applied to the weak form, which is the version that would actually constrain a trader.

What the 2013 Nobel Record Actually Says

The prize is regularly cited as though it settled the question. The announcement that awarded it does not read that way.

The announcement is dated 14 October 2013. Three economists shared the award, and the Royal Swedish Academy of Sciences divided the SEK 8 million equally between Eugene Fama, Robert Shiller and Lars Peter Hansen. The citation covers their empirical analysis of asset prices.

The Academy’s own summary sets out why all three were recognised together. Across a horizon of days or weeks, it says, the direction of share and bond prices cannot be called in advance. Widen the horizon to three to five years and their broad path becomes partly foreseeable. The Academy describes that pair of findings as ones that may look surprising and contradictory, and hands out the prize for both.

The first belongs to Fama. His work from the 1960s onward established how fast fresh information reaches the price, and how little room that speed leaves anyone trying to call the next few days.

The second belongs to Shiller, working in the early 1980s. He found that share prices swing far wider than the dividends underneath them, and that a price-to-dividend ratio standing high tends to come back down while a low one tends to climb. The Academy records the same behaviour in bonds and in other assets.

So a single committee, in a single citation, recognised unpredictability at short horizons and partial predictability at long ones, and paid equally for both. The award settled nothing in favour of either side.

Of the four widely read explainers read in full while preparing this page, none states that the 2013 prize was shared with Shiller, whose finding cuts against the reading they use it to support. One names it as Fama’s prize with no source attached to the claim at all.

The Regulator Already Turned Weak-Form Efficiency Into a Rule

Weak-form efficiency is discussed almost everywhere as an academic proposition. In the United Kingdom it is also an enforceable disclosure requirement, and the firms this audience deals with are already bound by it.

COBS 4.6 of the FCA Handbook governs how past performance may be presented in a financial promotion likely to reach a retail client. COBS 4.6.2R sets the conditions that presentation must satisfy. The section was last updated on 6 April 2026, and COBS 4.6.2R itself carries the date 1 January 2021.

Six conditions sit in that rule:

• A warning has to appear, prominently, carrying the point that the numbers are historical and are not a reliable indicator of future results.
• Those numbers cannot be the loudest element in the promotion.
• Coverage has to reach five years back, or across the whole life of the investment or service where that is shorter, and has to be cut into complete 12-month blocks.
• The window used, and where the numbers came from, both have to be named plainly.
• Gross figures have to be accompanied by what commissions, fees and other charges take out of them.
• Figures in a currency other than sterling have to say which currency, and warn that the exchange rate moves the outcome.

The five-complete-periods condition is the one carrying the weight. It removes the firm’s ability to choose the window, which is the cheapest way to make a record look predictive. That is the same defect a trader introduces by selecting the stretch of history a backtest is run over, and the rule exists because a chosen window is not evidence.

Notice what the rule stops short of. It does not declare past performance uninformative. It forbids presenting it as a reliable indicator of what comes next. That is weak-form efficiency written as an obligation rather than as a proposition, and it applies whether or not the firm or the reader accepts the theory behind it.

What This Changes About a Trading Strategy

Less than the volume of the argument suggests.

The hypothesis addresses whether an information set can produce a return. It says nothing about the costs subtracted from whatever return is produced, and those costs apply identically under either answer. The spread paid on entry and exit, financing on positions held overnight, and slippage between the price seen and the price filled are arithmetic, not theory.

A strategy that survives those costs was never going to be rescued by markets turning out to be inefficient, and one that does not survive them is not saved by proving that they are.

The second point is about evidence rather than markets. If a rule appears to work on historical prices, the weak form supplies a specific alternative explanation to rule out, which is that the rule was fitted to the sample. That is a reason to test differently rather than a reason to stop.

And the part of the argument that gets least attention is the part that binds hardest, because a conviction that the market is wrong is very hard to distinguish from the wish to be right about a position while it moves against you.

Which Figures This Page Does Not State, and Why

No percentage of active funds trailing an index appears above. Two of the four explainers read for this page state such figures, and neither attaches a report title, a date or a link to them, so there is nothing to check them against.

No figure for how quickly prices incorporate news appears either. The Academy’s announcement describes the incorporation as very quick and states no interval, and inventing one would put a number where the source declines to.

Where a number is not published by a source that can be named and dated, this page states the absence instead.

Who This Page Is Not For

Anyone looking for a verdict on whether technical analysis works will not find one here. That question is decided by evidence about particular rules on particular instruments after particular costs, and not by a hypothesis about information.

Nor is this a page about picking funds. The explainers that dominate this subject are written for an investor choosing between active and passive management, which is a different decision from the one a leveraged retail account presents.

Which part matters to you depends on what you are actually deciding. If you are weighing a rule built from price history, the weak form is the version to argue with, and the useful test is whether that rule survives out of sample and after costs.

If you are trading around scheduled releases, the semi-strong form is yours, and the question is whether you are acting on the number or on everyone else’s reaction to it. And if you are deciding whether the theory makes the whole exercise pointless, it does not say that. The costs would have been the binding constraint either way.