In this paper, we have investigated how portfolio allocations from the traditional mean-variance framework compare against Bayesian portfolio compositions. We show that the Bayesian allocation should always be preferred over the traditional mean-variance allocation for its improved out-ofsample performance and its more reliable prediction of the latter. Supplementing previous investigations on robust mean-variance optimization and the brief study by Meucci (2005), we provide to our best knowledge the first detailed study on the out-of-sample performance of robustifed Bayesian allocations. Surprisingly, we can observe that a robustifcation of Bayesian allocations does not yield signifcant additional improvement. This result is especially interesting because of preceding positive results in the literature on the robustifcation of the traditional mean-variance framework. Therefore, we believe that further research is required to understand this phenomenon in full detail.      «

In this paper, we have investigated how portfolio allocations from the traditional mean-variance framework compare against Bayesian portfolio compositions. We show that the Bayesian allocation should always be preferred over the traditional mean-variance allocation for its improved out-ofsample performance and its more reliable prediction of the latter. Supplementing previous investigations on robust mean-variance optimization and the brief study by Meucci (2005), we provide to our best knowledg...     »