Do Arbitrage Pricing Models Explain the Predictability of Stock Returns ?

Recent studies find evidence that the rates of return to holding common stocks and bonds are, to some extent, predictable over time. While measurement errors can induce spurious predictability, there seems to be an emerging consensus that predictable variation is an important stylized fact about asset returns. There are two competing views of this predictability. Some interpret This article studies predictability in U.S. stock returns for multiple investment horizons. We measure to what extent predictability is driven by premiums for economywide risk factors, comparing two standard methods for factor selection. We study single-beta models and multiple-beta models. We show how to estimate the fraction of the predictability in returns captured by the model, simultaneously with the other parameters. Our analysis indicates that the models capture a large fraction of the predictability for all of the investment horizons. The performance of the principal components and the prespecified-factor approaches are broadly similar. * This article has benefited from the helpful comments of a number of colleagues, including Torben Andersen, John Cochrane, Eugene Fama, Kenneth French, Allaudeen Hameed, Lars Hansen, Campbell Harvey, Ravi Jagannathan, Narasimhan Jegadeesh, Andrew Lo, John Long, Jianping Mei, Kenneth Singleton, Charles Trzcinka, and two anonymous referees. This article was presented at the 1992 Berkeley Program in Finance; the 1992 Conference on Financial Economics and Accounting; the 1992 Pacific Northwest Finance Conference; the 1993 American Finance Association; the 1993 Western Finance Association; the 1994 Conference on Multivariate Times Series and Financial Econometrics sponsored by the University of California, San Diego; and at the following universities: University of California, Berkeley; University of Maryland; McGill University; University of Michigan; Northwestern University; Ohio State University; University of Rochester; University of Southern California; University of South Carolina; Tulane University; and University of Wisconsin-Madison. We acknowledge the financial support of the New York Stock Exchange. Ferson acknowledges financial support from the Pigott-Paccar professorship at the University of Washington, and Korajczyk acknowledges financial support from Merrill Lynch Capital Markets.