Stock Market Predictability : Is it There? A Critical Review

This paper aims to survey selected recent papers presenting new evidenceon an age-old question in financial economics: Are stock market returnspredictable?. The hypothesis that equity returns are predictable (specificallyat long horizons) has been called a new fact in finance by Cochrane(1999). Thus, stock market predictability is now taken almost as a featureof the data. However, there is less consensus on what drives thispredictability. It may reflect time-varying risk premiums, it may reflectirrational behavior on the part of market participants, or it may simplynot present in the data - a statistical fluke due to poor statistical inference.It is this last possibility that seems to gain increasing credibilityconsidering the long list of authors criticizing the statistical methodologiesin the predictability literature. The results of these studies typically showthat findings against the constant expected excess return hypothesis basedon standard statistical inference can appear much more significant thanthey really are. Of course, the strength and usefulness of these criticalresults are still hotly debated - consistently evaluating their significanceis thus no easy task. Still, they raise questions about the implications ofthese findings for the way academics and practitioners use financial theory.In particular, the implications for the asset management industry(and dynamic asset allocation strategies) may be serious.