Bayesian Approaches for Portfolio Construction : A Review

Determining the optimal mix of assets in the context of a portfolio construction involves “smart” forecasts of asset returns as well as good estimates of the asset return variances and covariances. Typically, sample moments are used as best estimates of the population moments. Several researchers have pointed out, however, that this generally acceptable practice introduces uncertainty, which can degrade the desirable properties of the constructed portfolio. In this chapter we discuss several approaches that use Bayesian principles to address uncertainty in the context of portfolio construction. The literature we review covers a period of about 30 years.We start with a detailed overview of the problem. We then present remedies that have been proposed to address various aspects of uncertainty in the empirical finance literature