Deterministic Portfolio Selection Models, Selection Bias, and an Unlikely Hero

AbstractPortfolio selection models are programmed by their respective efficiency criteria to fall into a state of first-order condition love with the right sort of outliers. Nothing is changed, unfortunately, when, in general, a deterministic portfolio optimization model's inputs are stochastic rather than parametric. Distributional properties of the input estimator functions employed by four common portfolio selection models are reviewed and their solution algorithms studied in search of unique interactive effects that may mitigate the estimation error problem. Empirical and analytic support is provided for the conclusion that there is one model, an unlikely hero, that is least susceptible.