MEAN-VARIANCE OPTIMIZATION AND PORTFOLIO CONSTRUCTION: A SHORT TERM TRADING STRATEGY

Mean-variance optimization, in theory a very powerful and intuitive tool, has failed to provide meaningful solutions in practical settings, and indeed, in theoretical settings in much past research. Whereas inaccurate statistical estimates for inputs provide even more erroneous outputs, the modeling errors determine outputs that are nothing short of extreme. In this study, we employ two different models based on the mean-variance framework, with one portfolio seeking the highest return given a risk target while the other portfolio seeks the lowest risk given a desired level of return. In unconstrained form, our results confirm to be acutely departed from past experience in this subject matter and contrary to the known literature on modeling errors, our portfolios remain solvent. In constrained form, our portfolios outperform the benchmark and market portfolios while maintaining at least some diversification; in unconstrained form, our portfolios provide surprisingly high absolute and risk adjusted returns with betas less than the benchmark and market portfolios.