3.1 The Efficient Markets Hypothesis

The Efficient Markets Hypothesis (EMH) is the academic claim that sits across the doorway of technical analysis. Theory and History is 38% of CMT Level I, and this unit is where candidates must define the hypothesis, state how it is tested, and explain why chart-based work is still worth studying without pretending that a moving average is a guaranteed edge. Independent OpenExamPrep material for CMT Level I treats EMH as an argument about information and expected profits—not as a slogan to cheer or to dismiss.

Defining the Efficient Markets Hypothesis

In Eugene F. Fama's classic formulation, a market is informationally efficient when prices fully reflect available information. The operational, commonly accepted implication is narrower and more testable: you cannot consistently earn abnormal risk-adjusted returns from public information after costs. Abnormal means return beyond what an asset-pricing model says you should earn for the risk you took. After costs means commissions, bid–ask spreads, slippage, borrow fees, and short-sale frictions eat the paper profit. EMH does not say prices never move, that no one ever makes money, or that every investor is perfectly rational at every moment. It says that, in a competitive market, information you can all see should already be in the quote, so a simple, repeatable rule that uses that information should not print economic profit after risk and costs.

That implication is why EMH matters to a technician. Weak-form efficiency says past prices—and, in the usual technical-analysis reading, past volume—are already reflected. If the weak form held perfectly, reading a chart of yesterday's ticks would not produce durable excess returns. Level I still asks you to study price because real markets can be adaptive, behavioral, and inefficient in practice. Hold both ideas at once: the hypothesis is a serious null, and the null is not a photograph of every trading day.

The Joint Hypothesis Problem

Every empirical test of EMH needs a benchmark for what return should have been. Researchers subtract a model-required return (the Capital Asset Pricing Model (CAPM), Fama–French factors, a market index, and so on) from the realized return and call the residual alpha. If alpha looks reliably positive, two stories remain: prices did not fully reflect information, or the benchmark model mis-measured risk (or both). That inseparability is the Joint Hypothesis problem: tests of EMH are joint tests of informational efficiency and the asset-pricing model used to define abnormal returns.

On the exam, do not treat a famous anomaly paper as a courtroom conviction of inefficiency. A size premium might be compensation for distress risk that CAPM missed. Momentum might be a risk factor or a behavioral underreaction. The Joint Hypothesis problem is why efficiency and equilibrium pricing cannot be tested in isolation. A technician who beats the S&P 500 has not automatically beaten a risk-adjusted efficient-market test; the comparison index may not match the strategy's beta, sector mix, or drawdown profile.

Three Forms of Tests

Fama (1970) organized tests by how large the information set is.

Weak-form tests are the ones that most directly challenge classical technical analysis. Semi-strong tests are usually event studies: average the abnormal return around a public announcement and see whether the jump is immediate or whether a slow drift remains. Strong-form tests look at corporate insiders, specialists, or managers with non-public access. Empirically, strong-form efficiency is the easiest version to reject: legal insider filings and illegal insider-trading cases both show that private information can be valuable. Weak- and semi-strong-form evidence is mixed, sample-dependent, and always tangled with the Joint Hypothesis problem.

Three Anomalies That Challenge EMH

An anomaly is a pattern in returns that a benchmark efficient-market-plus-pricing-model story did not predict. Three families appear again and again in this debate.

Momentum. Over intermediate windows (often about 3 to 12 months), recent relative winners have tended to keep outperforming recent losers. That continuation is awkward for a naïve reading of weak-form efficiency, which expected little helpful serial correlation in returns. Momentum can fade at very short horizons because of microstructure noise and at very long horizons where reversal sometimes appears. It is not a promise that last week's hottest ticker is a buy; it is evidence that past return ranking has carried information in many samples.

Value and size. Stocks with cheap valuations (high book-to-market, low price-to-cash-flow, and similar ratios) and, in many historical samples, smaller capitalizations have shown higher average returns than CAPM-style betas predicted. Fama and French folded size and value into a pricing model, which illustrates the Joint Hypothesis problem in one stroke: the same pattern can be labeled mispricing or missing risk. For a technician, value and size effects are a reminder that cross-sectional cheapness and capitalization have historically moved expected returns—another public-information pattern that a strict semi-strong story has to explain as risk, cost, or chance.

Calendar effects and post-earnings-announcement drift (PEAD). Calendar anomalies include historically stronger January returns (especially among small stocks), weekend or day-of-week patterns, and related seasonal regularities. Many of these effects shrank after they were published, which is consistent with crowding, changing market structure, or data mining. PEAD is more stubborn: after an earnings surprise, average prices have continued to drift in the direction of the surprise for weeks to months instead of jumping once and stopping. A slow drift after a public number is a semi-strong-form challenge: the filing was public, yet the adjustment was not instantaneous.

None of these anomalies is a turnkey system. Capacity, costs, regime change, and the Joint Hypothesis problem all still apply. They are why prices fully reflect public information immediately is a hypothesis, not a daily observation.

What the Arcsine Law Predicts for Asset Prices

Lévy's arcsine law is a random-walk probability result that contradicts a popular intuition. For a symmetric random walk, the fraction of time spent on one side of the origin has a U-shaped (arcsine) distribution: paths that spend almost all of their time in the lead, or almost all of their time behind, are more likely than paths that split time evenly. Translate that into prices and relative performance: leaders tend to remain leaders. Time-on-lead is long. The identity of who is ahead does not quickly mean-revert just because the process is fair.

That prediction is easy to misuse. Persistent leadership in an index or a relative-strength ranking is consistent with a random walk; it is not, by itself, proof of a skilled trend. It is also not a forecast that the leader must crash back to the pack on a timetable. For Level I, remember the directional claim: arcsine geometry produces time-on-lead persistence, not mean-reversion of who is ahead.

Four Prominent Challenges to EMH

Beyond named return anomalies, four theoretical and empirical challenges show up in serious discussions of efficiency.

1. The Grossman–Stiglitz paradox. If prices already contain all information, no one pays to collect information, so prices cannot contain it. Efficiency, if it exists, must be limited by the cost of research. Perfectly informative prices are internally unstable.
2. Excess volatility. Robert Shiller and others documented that prices often move more than news about dividends or cash flows can justify if discount rates are stable. Large swings with little fundamental news are hard to reconcile with a simple price-equals-value-given-public-cash-flow-news story.
3. Fat tails, dependence, and crashes. Benoît Mandelbrot emphasized that returns are not independent Gaussian draws. Extreme days cluster; 1987-style gaps occur more often than a bell curve allows. A market can be a fair game in expectation and still be a terrible home for models that assume thin tails and independent increments.
4. Limits to arbitrage. Even if a mispricing is identified, capital, horizon, funding, and noise-trader risk can keep a rational trader from forcing the price back. Andrei Shleifer and Robert Vishny's limits-to-arbitrage argument is why smart money will fix it immediately is not an engineering identity.

Four Alternatives to EMH

If the strict efficient-market null is incomplete, what replaces it?

These alternatives do not license a claim of guaranteed edge. They explain why studying price, volume, and crowd behavior can still be rational when the EMH null is only an approximation.

Random Walk versus Martingale

A random walk (in the strict textbook sense) says the next increment is independent of the past, often with an identical distribution: P_t = P_{t-1} + ε_t with ε independent, frequently assumed Gaussian. A martingale is a fair-game statement: the conditional expectation of the next price, given today's information, is today's price. Variance may change; tails may be thick; independence of squared returns is not required. With a positive required return, prices are often modeled as a submartingale (E[P_{t+1} | information_t] ≥ P_t).

EMH is closer to a martingale / fair-game claim after risk adjustment than to a claim that ticks are independent, identically distributed, and normal. You can reject a naïve random walk (because volatility clusters) without automatically rejecting informational efficiency. Mixing the two terms is a common Level I trap.

Fama's Suggested Revision to the Three Forms

In Efficient Capital Markets: II (1991), Fama kept the research program but renamed the bins. Weak-form tests become tests of return predictability—not only past prices, but dividend yields, interest rates, seasonals, and cross-sectional forecasts. Semi-strong-form tests become event studies. Strong-form tests become tests for private information. The second and third changes are mostly labels; the first change widens the weak-form bucket so that past prices only is no longer the whole predictability literature.

The Level I Tension, Taught Honestly

Weak-form logic says past price and volume alone should not yield durable excess returns after costs and risk. CMT candidates still study those series because markets can be adaptive, behavioral, and inefficient in practice, because anomalies persist in some samples, and because limits to arbitrage keep mispricings alive. Independent study of these topics is not a warranty. A measured-move target, a Fibonacci grid, or a breakout rule is a hypothesis that still needs costs, risk, and invalidation. Treat EMH as the null you must understand, not as a reason to skip the chart—and not as a reason to skip testing.

Key Takeaways

• EMH: prices fully reflect available information; commonly, you cannot consistently earn abnormal risk-adjusted returns from public information after costs
• Joint Hypothesis: efficiency tests are always paired with an asset-pricing model
• Three test forms: weak (past price/volume), semi-strong (public information), strong (including private)
• Arcsine law: leaders tend to remain leaders (time-on-lead), not quick mean-reversion of who is ahead
• Fama 1991: return predictability, event studies, tests for private information