Vol. 21, No 2, 2023
by Eva A. Xu and Eric L. Tarkin
Standard models of risk and return are known to underestimate the frequency of extreme events and cannot account for the observed phenomena of increasing correlations in times of stress. This was most salient during the global financial crisis. Despite all of this, practitioners still rely heavily on the ubiquitous mean–variance optimization (MVO) for portfolio construction. This paper proposes a flexible framework, based on an explicit parametric model, that can address many of the shortcomings of MVO. We demonstrate that the proposed CVaR optimization is a superior descriptor of multi-asset returns and downside risk, and can lead to improvements in investment performance.