A1528
Title: Comparison between the Markowitz and Konno-Yamazaki Models for Portfolio Selection
Authors: Jose Luis Esteves dos Santos - Top Atlantico, Viagens e Turismo SA (Portugal) [presenting]
Helder Miguel Correia Virtuoso Sebastiao - University of Coimbra (Portugal)
Ana Francisca Ferreira Martins - University of Coimbra (Portugal)
Abstract: Two portfolio selection models are compared: the Mean-Variance (MV) model, introduced by Markowitz, and the Mean Absolute Deviation (MAD) model, proposed by Konno and Yamazaki. A large database is used for this comparison, consisting of the 500 largest U.S. stocks, analysed across different time frames. The study period spans from January 4, 2010, to November 30, 2021. After applying rebalancing strategies at various frequencies, an out-of-sample performance analysis is conducted for both models. Several performance metrics are employed in this evaluation, including win rate, cardinality, turnover, annualized mean return, annualized standard deviation, Sharpe ratio, Sortino ratio, Conditional Value at Risk (CVaR), and maximum drawdown. The results indicate that although the minimum-risk solution of the MAD model yields slightly worse values than the corresponding MV model, the differences are relatively small. However, a notable drawback of the MV model is that it tends to allocate capital across all 500 assets, many of which have extremely small weights, and this is an unrealistic assumption in practical investment scenarios. In contrast, the MAD model recommends investing in a more concentrated portfolio of approximately 40-50 assets, which aligns more closely with real-world portfolio management practices.
