A Fresh Look at Return Predictability Using a More Efficient Estimator

Abstract

I assess time-series return predictability using a weighted least squares estimator that is around 25% more efficient than ordinary least squares (OLS) because it incorporates time-varying volatility into its point estimates. Traditional predictors, such as the dividend yield, perform better in- and out-of-sample when using my estimator, indicating the insignificant OLS estimates may be false negatives driven by a lack of power. Some newer predictors, such as the variance risk premium and the president’s political party, are insignificant when using my estimator, indicating the significant OLS estimates may be false positives driven by a few periods with high expected volatility. Received March 31, 2018; editorial decision September 26, 2018 by Editor Jeffrey Pontiff. Authors have furnished an Internet Appendix and supplementary data and code, which are available on the Oxford University Press Web site next to the link to the final published paper online.