A behavioral model for mutual fund dynamics

Based on Berk and Green (2004), I develop a model that explains the following well-known stylized facts on mutual funds in a unified framework: (i) a negative aggregate return, (ii) a short-term return persistence, and (iii) a convex return-flow relationship. In the model, agents learn about managers' time-varying abilities from fund returns and non-return information signals. Under decreasing returns to scale, investors equilibrate expected fund returns through fund flows, but their expectations are biased due to overconfidence about precision of non-return signals and overextrapolation of past return trends. I employ a Simulated Method of Moments (SMM) to estimate the model parameters. The model matches most of the 15 moments, and is not rejected at the 10% level. I run a horse race between rational equilibrating forces and behavioral inefficiencies by allowing parameters for biases determined by data. As a result, both information processing biases appear to be important to generate a negative aggregate return and short-term return persistence, and to improve a model fit.