Asset Allocation with a High Dimensional Latent Factor Stochastic Volatility Model

Abstract

We investigate the implications of time-varying expected return and volatility on asset allocation in a high dimensional setting. We propose a dynamic factor multivariate stochastic volatility (DFMSV) model that allows the first two moments of returns to vary over time for a large number of assets. We then evaluate the economic significance of the DFMSV model by examining the performance of various dynamic portfolio strategies chosen by mean-variance investors in a universe of 36 stocks. We find that the DFMSV dynamic strategies significantly outperform various benchmark strategies out of sample. This outperformance is robust to different performance measures, investor’s objective functions, time periods, and assets.