On the use of factor betas versus characteristics in the performance evaluation of mutual funds

This paper introduces a new approach for estimating mutual fund performance that simultaneously controls for both factor model betas and rm characteristics. Our double-adjusted performance measure is motivated by growing concerns in the literature regarding the ability of traditional Fama-French factor models to adjust returns for the main anomalies. We propose a hierarchical Bayes approach to model conditional multi-factor model alphas as a linear combination of rm characteristics. We nd that traditional alphas are signicantly associated with rm characteristics, even when the factor model explicitly adjusts returns for those characteristics. Double-adjusted performance provides new evidence of mutual fund performance persistence, as we remove the noisy component of performance associated with passive loadings on characteristics. Moreover, we nd that previous relations between traditional performance estimates and specic fund features, e.g., mutual fund selectivity and mutual fund ows, are partially driven by the characteristic component of performance.