Imperfect Predictability and Mutual Fund Dynamics: How Managers Use Predictors in Changing Systematic Risk

Suppose a fund manager uses predictors in changing portfolio allocations over time. How
does predictability translate into portfolio decisions? To answer this question we derive a
new model within the Bayesian framework, where managers are assumed to modulate the
systematic risk in part by observing how the benchmark returns are related to some set of
imperfect predictors, and in part on the basis of their own information set. In this
portfolio allocation process, managers concern themselves with the potential benefits
arising from the market timing generated by benchmark predictors and by private
information. In doing this, we impose a structure on fund returns, betas, and benchmark
returns that help to analyse how managers really use predictors in changing investments
over time. The main findings of our empirical work are that beta dynamics are
significantly affected by economic variables, even though managers do not care about
benchmark sensitivities towards the predictors in choosing their instrument exposure, and
that persistence and leverage effects play a key role as well. Conditional market timing is
virtually absent, if not negative, over the period 1990-2005. However such anomalous
negative timing ability is offset by the leverage effect, which in turn leads to an increase
in mutual fund extra performance.