This paper investigates the ability to improve traditional portfolio optimisation rules in practice. Specifically, I examine the effect of parameter uncertainty on Markowitz portfolio performance and quantify the corresponding losses. Frequentist methods, in the form of di-rect parameter shrinkage and assigned portfolio weight shrinkage are employed to suppress effects of estimation errors. A Bayesian approach of the traditional Markowitz portfolio is used to account for estimation risks implicitly and novel Bayesian portfolio combinations are defined in search of an optimal investment rule. Moreover, the assumption of normal re-turns is relaxed by considering Markov Switching Gaussian Mixture models. I demonstrate the mean-variance modifications to be effective in improving out-of-sample performance and show ability to beat the equally weighted benchmark for empirical samples of the 48 Industry portfolios and Fama-French 100 portfolios. Lastly, I validate the competence of producing robust portfolios over periods of changing economic times when parameter uncertainty is considered in the Markowitz model.

Additional Metadata
Keywords Portfolio Optimisation, Parameter Uncertainty, Shrinkage, Bayesian Statistics JEL Classifications: G11, C38, C11, C53
Thesis Advisor Naghi, A.A.
Persistent URL
Series Econometrie
Pelt, T.G. van. (2020, July 17). Constructing Robust Portfolios, the Role of Parameter Uncertainty in Dynamic Optimal Portfolio Allocations. Econometrie. Retrieved from