Predicting the Market Using Information from Equity Portfolio Returns

Abstract

In this paper we provide new evidence on the predictability of aggregate stock market returns, and new time series of the expected excess returns on common stocks. We extract aggregate discount rate news from equity portfolio returns and use this information to construct estimates of expected excess market returns. We find that a linear combination of the lagged returns on the market portfolio, on small firms with high book-to-market ratios and on large firms with low book-to-market ratios contains information about future market returns over horizons of a few quarters. In contrast, the lagged returns on the market portfolio and 6 portfolios formed on the basis of dividend yield provide information that is useful for predicting market returns at business cycle horizons or longer. Moreover, the conditioning information we find is largely uncorrelated with commonly used predictor variables such as the market dividend yield and the CAY variable of Lettau and Ludvigson. Further analysis suggests that the level of predictability found cannot be attributed to data-mining bias.

Keywords: predictability, equity premium

JEL Classification: G10, G12