Portfolio Optimization : A Comparative Study of Two Alternative Approaches with Unknown Risk

Sammanfattning

Portfolio optimization is the mathematical process of finding the set of assetsto maximize the returns given a target risk or to minimize the risk givena return target. The standard model used in portfolio optimization is theMean-Variance Optimization model. This paper specifically examines twoalternative portfolio optimization models when an accepted risk threshold isunknown. The first approach demands a diversification of assets based onthe sectors to which they belong. The second model is based on assessinghighly correlated assets, in order to diversify by avoiding to invest too muchin such assets. The purpose of this paper is to explore how different parameterconfigurations affect the return and risk of these two optimization models andhow they compare to the standard risk-based model formulation. Both approaches were tested on S&P500 data and implemented usingGurobi software. Both approaches proved to be satisfactory as diversification-based heuristics in cases where the accepted risk threshold is unknown. Theirperformance roughly aligned with the efficient frontier formed by the MVOmodel, but with higher risk. The results should be an acceptable and useful foundation for future work.Such work may include more advanced optimization approaches with the sameobjective, and may benefit from our results as reference data or inspiration.