Bayesian Dynamic Panels for Mutual Fund Performance Persistence

Summary

The article summarizes research that models mutual fund returns and performance persistence with a Bayesian dynamic panel. The model allows fund returns to have time-varying variances and covariances, autocorrelation, stochastic volatility, and measurement error, relaxing the assumption that different funds’ errors are independent. It estimates the model with sequential Monte Carlo and particle filtering, and examines how alternative prior assumptions affect the results.

Using US mutual fund data from 2000 to 2014, the study reports differences in performance and persistence across fund categories. Risk and most of the Fama–French five factors are associated with higher measured performance, while fees are generally negative; fund size, turnover, and distribution fees also show positive relationships in reported specifications. Negative performance tends to persist, especially during the financial crisis, and volatility rises before the crisis in the sample. These are historical findings from one market and period, and some relationships vary across model specifications; they do not establish reliable forecasts or causal effects.

Key ideas

• The model captures time-varying return variance and covariance, autocorrelation, volatility, and measurement error across funds.
• Bayesian estimation allows the researchers to assess sensitivity to prior assumptions.
• The reported US mutual fund results vary across categories and model specifications.
• Negative performance tends to persist, particularly during the financial crisis.
• The study reports a rise in volatility before the crisis, but this historical pattern does not establish forecasting reliability.

Tags

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.