We investigate the factor exposure of smart beta ETFs under model uncertainty using Bayesian variable selection. We find that smart beta ETFs have exposures to several factors, including size, value, momentum, quality, volatility/low beta, and dividend yield. The average return contribution of...

We pit individual theoretical predictors of the equity premium against a variety of data-driven statistical methods. Theoretically motivated predictive regressions outperform conventional penalised regressions but have similar out-of-sample R2 and lower economic gains relative to more agnostic...