Clive W. J. Granger and Oskar Morgenstern · 1970
This work is a co-authored research monograph. Its scope is both conceptual and empirical: Granger and Morgenstern examine what it could mean to predict stock-market prices, how that question had been confused in prior discussion, and what statistical evidence can and cannot show about prices, price changes, volume, and relations among series. The book’s central thesis is not simply that markets are “random,” but that much alleged predictability dissolves once prediction is defined carefully and tested against appropriate price-change data.
As the title of the book implies, we have particularly concerned ourselves with the question of the predictability of stock market prices.
The authors’ first conceptual move is to strip stock prices of intrinsic metaphysical content. A price is not a revealed essence but a transactionally sustained valuation, dependent on what market participants will exchange for the security. This makes prediction an empirical problem about market behavior, not a deduction from “true value.”
The value of a stock is only what someone else will pay for it — in cash, in another stock or whatever it may be.
From that point, the monograph distinguishes several questions often run together. To predict an absolute price level is not the same as to predict relative movements, and to predict volatility or magnitude is not the same as to predict direction. This distinction is crucial because a statistical relation may exist without creating a profitable directional forecast.
The distinction between the prediction of absolute prices and that of relative prices has not been kept sufficiently clear in the literature.
The book’s treatment of the random-walk hypothesis is similarly careful. Granger and Morgenstern resist the loose interpretation that a random walk means all forms of prediction are impossible. Instead, they define the claim in relation to particular information sets and techniques, especially linear combinations of past changes. The negative result is therefore conditional and methodological: it says which kinds of historical-price rules fail, not that all future-oriented knowledge is logically excluded.
The random walk hypothesis does not say that price changes are unpredictable: it says that they are not predictable using (linear) combinations of previous price changes.
The empirical structure of the work follows from this clarification. The authors investigate whether past price behavior, volume, or relations among different series yield usable forecasts. Their results emphasize the difference between price direction and the intensity of market movement. Volume does not appear to tell the analyst whether prices will rise or fall on a given day, but it is associated with the size of movement. That finding matters because it preserves an empirical regularity while denying that the regularity is a straightforward prediction device.
Daily volume is unrelated to daily price change but is related, in an apparently unlagged fashion, with the square of daily price change and other measures of the extent of price change rather than the direction of the change.
The same caution governs their treatment of inter-series relations. The authors do not find that one stock or price series reliably leads another in a way that would permit systematic forecasting. Here the monograph’s skepticism is not rhetorical; it is an outcome of testing predictive relations across series and finding little that survives.
No one series seems to have any predictive value for any other series.
The relevance of Predictability of Stock Market Prices lies in this disciplined narrowing of claims. It contributes to the early econometric literature on efficient markets and random walks, but its enduring value is methodological: it asks what kind of prediction is being claimed, what data would support it, and whether the discovered relation concerns direction, magnitude, timing, or cross-market dependence. Its core argument is that stock-market predictability must be specified before it can be measured. In doing so, Granger and Morgenstern turn a broad speculative question—can the market be forecast?—into a set of precise statistical questions, many of which receive negative answers, and some of which reveal weaker but meaningful regularities about volatility and trading activity.
This work was divided into 114 sections when it entered the library's research corpus—an apparatus for search and citation, not necessarily the author's own table of contents. Each title opens its summary.
Put a question to this work; the Librarian answers from its 114 sections and cites the passage.
Ask the Librarian