Comparison of Portfolio Optimization Models with Real Features : An Empirical Study Based on Chinese Stock Market

In this paper, we compare the performance of four single-period portfolio selection models: Mean-Variance model (MV), Mean Absolute Deviation model (MAD), Conditional Value-at-Risk model (CVaR) and Minimax model (MM). Real features such as transaction cost and cardinality constraint are considered. Systematic in-sample and out-of-sample analyses are carried out to evaluate the efficiency of these four models using historical data from periods with different markets trends in Chinese stock market. Our empirical results reveal that portfolio strategies generated by these four models outperform the market portfolio with higher Treynor ratio and positive information ratio in the bull market and less volatile market, while they generally underperform the market portfolio in the bear market. Moreover, we find that the MM model demonstrates an ability of generating superior portfolio in volatile market with high degree of uncertainty