Out-of-sample equity premium prediction in the presence of structural breaks

Abstract

This study comprehensively investigates the uncertainty on parameter instability and model selection when forecasting the equity premium out-of-sample. We employ the robust optimal weights methodology proposed in Pesaran et al. (2013) to construct out-of-sample forecasts in the presence of possible structural breaks. While we find that parameter instability alone cannot fully explain the weak predictive performance of many variables considered in Goyal and Welch (2008), our empirical results show that some models, particularly the one with the stock market variance, can consistently generate superior statistical and economic gains relative to the historical mean benchmark and other competitors when estimated by the robust optimal weights. Furthermore, we discover that the stock market variance seems to be more powerful when forecasting the equity premium during periods of financial crisis.