Zusammenfassung:
Building on the success of Ferreira and Santa-Clara (2011) in separately forecasting the return components of the stock market, this paper examines the links between economic regimes and these components to predict the aggregate U.S. stock market. We propose a three-step methodology that we call the flexible regime approach. First, we estimate the regime dynamics of ten macro-financial variables using Markov-switching regressions. Second, we treat the regime filtering results from the Hamilton filter as views and test the predicted regime classification, the predicted regime probabilities, and the conditional and mixture densities as view generators. We use entropy pooling to re-weight the historical distribution to derive posterior probabilities. Finally, we link these probabilities to the realized outcomes of earnings growth and changes in the price-earnings multiple to form the sum-of-the-parts forecast. Our results demonstrate significant predictability from a statistical and economic perspective. We emphasize the role of default spreads and interest rates in predicting earnings growth and stock market volatility and inflation in predicting multiple growth. Finally, our results suggest that the predictability of both return components varies over time and is affected by the business cycles. While earnings growth is more predictable during periods of expansion, forecasting multiple growth is more advantageous during recessions.