Modified Models for Constrained Mean Absolute Deviation Portfolio Optimization

Abstract

This study addresses portfolio optimization through a comparative analysis of models. Utilizing 15 stocks from S&P 500 over 12 years, it contrasts optimal weights and objective values of five models: Konno and Yamazaki (KY), Feinstein and Thapa (FT), Adjusted Feinstein and Thapa (AFT), and Adjusted ChiangLin et al. (ALC). Employing Pyomo in Python with GLPK solver, findings reveal KY and LC models have identical outcomes. AFT model approximates KY and LC solutions. Modified AFT and ALC models, with short selling and risk-neutral interest rates, closely mimic KY’s results. The study recommends KY’s model or a modified ALC model for portfolio optimization.