Verification on the Performance of Classical and Modern Portfolio Optimization Models

Abstract

In the financial markets, people make portfolio investments in order to diversify the risk taken by investments to individual assets. However, due to the lack of professional knowledge, the investors may invest in the assets randomly or just with equal weight to each asset in their portfolio. Commonly, the strategies provided by portfolio managers are expected to earn higher returns than the random investments made by unprofessional investors. And also, since Markowitz proposed the Modern Portfolio Theory in 1952, various portfolio optimization models have been developed to address the portfolio selection problems. However, many proposed portfolio optimization models depend on the forecasting of the future returns’ probability distributions, which can lead to estimation errors. Thus, these optimization models may not provide an advantage compared to just randomly selected portfolio composition. The objective of the doctoral thesis is to make the verification on the historical performance of strategy portfolios obtained by applying the classical and modern portfolio optimization models. To make the verification, we compare the performance of the strategy portfolios with that of the random-weights portfolios, by this way, we draw the conclusion on whether the strategy portfolios obtained from the applied portfolio optimization models outperform the random investments. In our research, we apply the classical portfolio optimization models and the modern portfolio optimization models to generate the strategy portfolios. By applying the chosen risk and performance measures, as well as the proposed statistical testing methods, we verify the performance of the obtained strategy portfolios. In the verification procedure, we apply the random-weights portfolios generated by applying the Monte Carlo simulation method as the benchmark. By applying the historical daily adjusted closing prices of stocks in the Dow Jones Industrial Average as the datasets, we make three empirical case studies. Based on the results of the studies, we have the following main findings. Firstly, the strategy portfolios obtained from minimum-risk portfolio optimization models are verified as lowering the corresponding risk measure in the out-of-sample period. Secondly, the Minimum-CVaR strategies lead to both increased performance and low risk when the stock market changes from the uptrend to the downtrend. Thirdly, the maximum-performance strategy portfolios are not verified as having a stable advantage in improving the portfolio performance in the out-of-sample periods, and most of them react similarly to the changes of the analysis period.