Flexible Optimal Models for Predicting Stock Market Returns

Abstract

This study assesses the usefulness of flexible optimal models of business cycle variables for predicting stock market returns. We find that variable estimation periods identify structural breaks in months with large absolute returns and the optimal models recognize regime switches. Flexible optimal models have much greater predictive power for stock market returns than fixed univariate or multivariate models. The dividend yield has consistent predictive power for stock market returns, but different variables make significant contributions to predicting stock market returns in different periods. These findings highlight the importance of employing flexible optimal models to consistently predict stock market returns.

Keywords: Predicting Stock Returns, Optimal Models, Business Cycles, Dividend Yield

JEL Classification: G11, G12