What exactly is efficient?
Separate three questions: is the message new information; has it reached the quote; and can the observed pattern be used after costs and risk? A yes to the first is not a yes to the third. Information may be public but difficult to process. Processing cost belongs to the economics of the test even when it is invisible on a chart.
Three information sets
The weak form concerns past prices; the semi-strong form public information such as published reports; the strong form also considers information to which access may be monopolistic. These are nested but different claims. Testing daily-return patterns does not automatically test every consequence of financial reporting disclosure. The existence of private information cannot be turned into a justification for using it: this is a research classification, not a statement about permitted conduct.[1]
For [fundamental analysis](/guides/fundamental-analysis/), the distinction matters. Business analysis can help understand risk and assumptions even if it does not deliver excess performance against a comparable alternative. Do not choose between “analysis is useless” and “effort must be rewarded”. An informed decision and additional return are different quality criteria.
An educational earnings announcement
Suppose a fictional share is 100 before publication, 106 immediately after and 105 at day end. An observer says, “After good news the share rose, so buy good reports.” But a buyer who received the news at the same time as the market and bought at 106 would see a move to 105, about −0.94% before costs. The rise from the old price was unavailable to that buyer. The figures are fictional and show only the error in choosing the starting point.
A substantive test would define unexpected earnings in advance, collect all suitable publications, record exact times and compare subsequent results with risk and market movement. Unsuccessful observations and excluded companies must remain in the journal. Selecting only successful examples tests the author’s ability to choose illustrations, not market efficiency.
Why one test does not settle the dispute
Results depend not only on the information hypothesis but also on the model of normal return. If a strategy beats an index, explanations include additional risk, a matching error, chance or genuinely unreflected information. Summers separately shows the limited power of common tests when valuation deviations persist for long periods. A lack of statistically visible dependence does not prove prices are error-free.[2]
The reverse error is treating every bubble as proof of a tradeable known-ahead profit. That would require an observable recognition rule, executable trades and accounting for possible continued rises. Post-event history does not satisfy those conditions. [Technical-analysis evidence](/guides/technical-analysis-evidence/) continues the rule-testing discussion.
Practical reading result
Record the information set and access time; name the alternative and risk model; separate gross from net performance; test a new sample. In the primary source, look for these conditions, not only the abstract’s conclusion. The efficiency hypothesis is useful discipline for testing bold promises, not a certificate of no risk or a reason to buy any asset at any price.
Test an alternative explanation
Formulate two explanations for one result. In the first, price rapidly reflects known facts but the required risk premium changes. In the second, some information is absorbed with delay. Which observations would distinguish them? If the answer is only the same chart, identification is incomplete. Name why the selected measure captures information delay rather than risk or participant composition.
Also test the opposite conclusion. If a rule failed, might the test have been too short, costly or poorly implemented? This is not a reason to rescue every promise; it separates rejection of one implementation from rejection of the whole idea. State what the result did not confirm and what it cannot exclude.
Sources and verification limits
[1] Efficient Capital Markets: A Review of Theory and Empirical Work, Fama 1970: classification of information sets; not a claim about current efficiency of every market.
[2] Do We Really Know That Financial Markets Are Efficient?, NBER working paper 0994: limits of long-horizon empirical tests of efficiency.