CFE 2015: Start Registration
View Submission - CFE
A0485
Title: Mutual fund dynamic risk allocation and skill level with a Bayesian seminonparametric change-point model Authors:  Mark Jensen - Federal Reserve Bank of Atlanta (United States) [presenting]
Mark Fisher - Federal Reserve Bank of Atlanta (United States)
Paula Tkac - Federal Reserve Bank of Atlanta (United States)
Abstract: The aim is to model the skill and risk-factor allocation of a mutual fund dynamically as a change-point process of the four factor Fama-French, momentum model of mutual fund returns. Each mutual funds change point probability is assumed to be a random draw from an unknown distribution that is modelled nonparametrically. In addition, each fund regimes latent risk factor beta and alpha coefficients are also assumed to be random draws from unknown meta-distributions. We nonparametrically model each meta-distribution as an unknown distribution and place a hierarchical prior on it. A nonparametric heirarchical prior for the regime coefficients enable our change-point model to learn from past fund regimes risk factor betas and performance alphas to better predict their out-of-sample values. We find dynamic factor allocation and changing levels of skill for mutual funds. An arbitrary fund will have a relatively flat duration distribution for a regime. In other words, a fund is more or less likely to experience a change point.