Portfolio Optimization and Stability : Evidence from the OMXS30

Sammanfattning

This study looks at the performance outside the sample, stability, and robustness of four portfolio optimization models: The Mean-Variance model, the Ledoit-Wolf model, the Treynor-Black model, and an extended Black-Litterman model that includes Fama-French factors. The analysis uses data from the OMXS30 in both stable market conditions and volatile market conditions. A rolling window method is used, and the performance is assessed using risk-adjusted measures that include the Sharpe ratio, Sortino ratio, Jensen´s alpha, and Information ratio. The results indicate that there are clear differences among the models. The Black-Litterman model has the best overall performance and shows most consistency across different market conditions. The Mean-Variance model has the weakest performance and the highest level of instability. The Ledoit-Wolf model does not provide much improvement compared to the Mean-Variance benchmark. At the same time, the Treynor-Black model demonstrates some occasional resilience but lacks consistency. In general, the findings show that the choice of model has a significant impact on portfolio outcomes, especially when market conditions are changing. This study adds empirical evidence from a small open economy and emphasizes the need to address estimation error in the construction of portfolios. From a practical viewpoint, equilibrium-based methods like the Black-Litterman seem to deliver more robust performance in comparison to the traditional method.