The objective of this study is to investigate portfolio selection using Markowitz - Sharpe mean-variance analysis by employing technology readily available to today's manager.
Despite having theoretical appeal, mean-variance optimisation has not been widely adopted by the investment community. This study explores possible reasons for this, and investigates some of the inherent limitations on practical application of mean-variance optimisation.
Portfolio optimisation generally requires sophisticated numerical methods, however. Analytical derivation of the special case unconstrained mean-variance frontier is presented, which is later used for portfolio comparison purposes.
One of the results of the efficient market hypothesis is that share returns are normally distributed. Actual share returns more closely follow a lognormal distribution. Simulation methods for share returns are also investigated, with methods explained to generate lognormal time series share returns.
James-Stein estimators provide dominate estimators over the mean for security returns. James-Stein estimators are applied to the portfolio selection problem in the context of mean-variance analysis.