Bayesian Alphas and Mutual Fund Persistence




    Search for more papers by this author
    • Busse is from the Goizueta Business School, Emory University, and Irvine is from the Terry College of Business, University of Georgia. We thank an anonymous referee, Stephen Brown, Richard Green (the editor), Lubos Pástor, John Scruggs, Jay Shanken, Chris Stivers, and seminar participants at University of Cincinnati, University of Colorado, Erasmus University, University of Georgia, Simon Fraser University, University of Texas at Dallas, Tilburg University, Vanderbilt University, Yale University, the 2003 American Finance Association meetings in Washington, DC, the Berkeley Program in Finance, and the CIRANO Fund Management Conference for useful comments. We thank Rob Stambaugh for preliminary discussions on this topic. Thanks to Ron Harris for research assistance.


We use daily returns to compare the performance predictability of Bayesian estimates of mutual fund performance with standard frequentist measures. When the returns on passive nonbenchmark assets are correlated with fund holdings, incorporating histories of these returns produces a performance measure that predicts future performance better than standard measures do. Bayesian alphas based on the Capital Asset Pricing Model (CAPM) are particularly useful for predicting future standard CAPM alphas. Over our sample period, priors consistent with moderate to diffuse beliefs in managerial skill dominate more skeptical prior beliefs, a result that is consistent with investor cash flows.