Three Forms Of Stock Market Efficiency
In its general form, EMT explains more than the random walk model. That model says simply that successive price changes are independent or uncorrelated, whereas EMT explains that by saying that stock prices fully reflect all the information (not just price histories) about a stock. As a result, virtually since the emergence of EMT as an explanation of the random walk model, EMT has been divided into three forms defined in terms of specified categories of information.
The three forms were first proposed to classify empirical tests of price behavior given specified kinds of information. The weak form tested the random walk model itself, using correlation and run tests like those just described to investigate whether past prices indicate anything about future prices. Semistrong-form testing investigated whether publicly available information other than prices was reflected in prevailing prices, and strong-form testing investigated whether private information was reflected in prevailing prices.
As the wealth of tests and discussion proceeded in the 1970s, the three forms of EMT came to be used to refer to the conclusions those tests suggested. Thus, the forms of EMT have since then been specified as follows: The weak form says that stock prices fully reflect all information consisting of past prices, the semistrong form says that stock prices fully reflect all information that is currently publicly
available; and the strong form (hold your breath) says that stock prices fully reflect all existing information, whether publicly available or not.
There is thus a direct and logical link between the random walk model and weak-form efficiency but a more attenuated and contingent link between the random walk model and stronger forms of EMT. Recall that the random walk model holds that price changes are independent of or uncorrelated with prior price changes. That means that technical analysis of past price changes—sometimes called chartist analysis—cannot aid the prediction of future price changes in any systematic way.
Weak-form efficiency explains this independence and its implications for prediction by hypothesizing that the current price impounds all information contained in prior prices. Thus, any price change can only be the result of new information, the production of which is itself assumed to be random. This process of information absorption continues and thus explains the absence of substantial linear dependence in successive price changes discovered in the correlation and run tests of the 1960s. It also leads to the stronger forms of the hypothesis.
The semistrong form of EMT posits not only that current stock prices reflect all information consisting of prior stock prices but also that they reflect all publicly available information about the stock in question. Testing this more ambitious claim requires a focus not on correlation analysis of price changes but on the relative swiftness with which prices change given new information.
Despite this different testing methodology, semistrong efficiency depends on the validity of the random walk model, which depends in turn on empirical conclusions concerning the absence of statistical dependence in stock price data. In other words, if future price changes depend on prior price changes, any price change the semi- strong form tests cannot be attributable solely to the new information the test is evaluating. Thus, both weak efficiency and semi- strong efficiency depend on the proof provided by linear testing models.
The strong form of EMT extends much farther than the random walk model suggested. Indeed, the strong form is a theological proposition, holding that public capital markets are infinitely wise: even nonpublic information is reflected in public stock prices. Abundant evidence has decimated the strong form by showing that people who possess nonpublic information can use it to make abnormally high market returns, with the apotheosis being the insider trading scandals of the 1980s.
Since the strong form of EMT has been discredited, debate concerning EMT centers on the semistrong and weak forms. Debate over the weak form generally is defined by analysis of the random walk model itself, usually in terms of the linear empirical models used to test the relationship between successive price changes; this has cut into the random walk model.
A widely noted study by MIT's Andrew Lo and Wharton's A. Craig MacKinlay demonstrated strong positive serial correlation in stock prices for weekly and monthly holding period returns. Using 1,216 weekly stock return observations from 1962 to 1985, they found a weekly correlation coefficient of 30%, an extremely high level of correlation. These researchers point out in their book, A Non- Random Walk Down Wall Street, that this does not mean necessarily that the stock market is inefficient but that the random walk model cannot be the basis for any theory of efficiency.
Though not conclusive, evidence like this led even Eugene Fama—a chief architect of EMT—to conclude that daily and weekly stock returns are predictable from past returns, thus rejecting the random walk model on a statistical basis. 9 Even so, Fama and the other fathers of EMT cling to the view that such discrepancies are merely anomalies that do not impair the basic model's validity, though others try to explain these results in a different way called noise.
