Can Overvalued Factor Strategies Predict Returns?

There exists a great number of works addressing financial market anomalies. Hou et al. (2017) attempted to replicate all existing works on this topic. Their results showed most of the known anomaly variable to be insignificant. Yet, the paper didn’t address a number of anomalies reported by Arnott et al. (2016). We thus decide to analyze two particular anomalies from this work, in particular, we look at predictive ability of relative valuation with respect to subsequent performance of the long and short parts of Fama-French factor portfolios (HML, SMB). While we find that there is negative relation between current valuation and a subsequent return over the period between 1945-2017, only HML factor exhibits significant coefficient on relative valuation over various horizons for subsequent return. Significance for SMB portfolio is achieved only at long horizons for a subsequent return of 2-5 years. We show that OOS predictive ability of relative valuation is no better than that of mean-based forecasts. An attempt to improve performance of the initial factor portfolios through an investment strategy that would account for relative valuation of the long and short parts didn’t yield success.