Mutual Fund Families and Performance Evaluation

We develop a continuous-time Bayesian learning model to evaluate the composite skill of a mutual fund manager and a fund family. Our model estimates the composite skill of each fund as a function of its own performance and family performance. We show two competing effects of the family performance on the evaluation of a member fund: a positive common-skill effect and a negative common-noise effect. The overall effect increases with the correlation of unobservable skills, and decreases with the correlation of unobservable noise. This pattern is stronger in larger families. Consistent with our assumptions, we find empirically that funds within the same family show higher correlations of estimated alphas and of residual returns. We also find that the effect of family performance on flows to a member fund exhibits strong cross-sectional patterns that are consistent with our model predictions