Analyzing the Performance of a Two-Tail-Measures-Utility Multi-objective Portfolio Optimization Model

This research paper proposes and experimentally investigates the out-of-sample performance of a multi(three)-objective portfolio optimization model. The three objectives used to evaluate the return distribution of the portfolio are two tail performance measures and a utility function in order to evaluate the middle part of the return distribution. Five different utility functions are considered, thus forming five instances of the proposed multi-objective portfolio selection model. For solving the problem, a multi-objective evolutionary algorithm, namely Strength Pareto Evolutionary Algorithm 2 (SPEA2), was employed. The results show that the majority of the portfolios generated by the solution technique produce better portfolios than S&P 500 Index considering the final wealth, Sharpe ratio, and Sortino ratio. The five portfolio models defined by different utility functions return, on average, approximately 10% more than the return on S&P 500 Index for an evaluation period of one and a half year. Furthermore, the computational results show that the proposed multi-objective portfolio optimization models are competitive to portfolios that have shown good out-of-sample performance in past studies, like the minimum variance portfolio with and without short sales and the second-order stochastic dominance portfolio.