Portfolio Optimization Using the Atkinson Index

Abstract: Traditional mean-variance optimization of portfolios has received much criticism due to its inability to account for higher order moments and non-quadratic utility. In this thesis, the topic of portfolio optimization is studied using the Atkinson index with CRRA utility. We construct Atkinson-efficient portfolios using computer-generated data in Monte Carlo simulations, as well as using financial asset returns data of assets taken from American stock exchanges. The results show that when normality holds, there are no benefits to the usage of the Atkinson. Under non-normality, however, there are advantages to using the Atkinson instead of the Sharpe ratio. These advantages are great enough for researchers to disregard its greater complexity.