Investor Learning and the Aggregate Allocation of Capital to Active Management

Abstract

We estimate a model in which Bayesian investors learn about parameters governing mutual fund performance in real time and competitively allocate capital to funds, conditional on their current beliefs. The model-implied aggregate allocation of capital in response to the history of observed returns closely approximates the observed allocation over time. Key to this result is that investors learn not only about differential ability across funds but also about the nature of returns to scale---how a fund's performance depends on its size versus the size of its competition. Overall, our results support that mutual fund investors are not naive.

Keywords: Active Management, Returns to Scale, Learning, Mutual Funds, Bayesian Methods