Excess Return Forecast Using a Dynamic Asset Class Factor Model

We propose a Dynamic Hierarchical Factor Model using Asset classes
to predict mutual funds excess returns. We use different forecast combination
schemes of bivariate model considering each asset class factor
in isolation. Primary analysis highlights the importance to account for
asset class specific variations together with between classes or common
variations. Further refinements of the a priori repartition are however in
order. Forecasting performance of the model outperforms the historical
mean benchmark both in terms of MSPE and utility based criteria. A
forecasting exercise matching more closely real-time conditions must be
undertaken to validate these initial results.