Forecasting Stock Market Returns: The Sum of the Parts is More than the Whole
Miguel A. Ferreira Universidade Nova de Lisboa
Pedro Santa-Clara University of California, Los Angeles
National Bureau of Economic Research (NBER) September 7, 2008
We propose forecasting separately the three components of stock market returns: dividend yield, earnings growth, and price-earnings ratio growth. We obtain out-of-sample R-squared coefficients (relative to the historical mean) of nearly 1.6% with monthly data and 16.9% with yearly data using the most common predictors suggested in the literature. This compares with typically negative R-squared coefficients obtained in a similar experiment by Goyal and Welch (2008). An investor who timed the market with our approach would have had a certainty equivalent gain of as much as 2.3% per year and a Sharpe ratio 82% higher than using the historical mean to forecast returns. We conclude that there is substantial predictability in equity returns and that it would have been possible to time the market in real time.