Comment on: Kritzman, M. 2006, 'Are Optimizers Error Maximizers?'
Forthcoming in the Journal of Portfolio Management
3 Pages Posted: 12 Dec 2019 Last revised: 2 Jan 2020
Date Written: November 13, 2019
Abstract
Kritzman (2006) asserts that the widely cited Michaud (1989) study characterizing the impact of estimation error on Markowitz (1952) mean-variance (MV) optimization as “error maximization” is hype. His paper consists of two examples of MV portfolio optimizations: an eight-country index asset allocation with near identical returns and four-indices with very different estimates. In spite of the absence of standard references in estimation error, authors continue to cite the paper. In this brief note we demonstrate that even in these stylized and unrealistic examples, Kritzman’s MV optimized portfolios perform on average worse than equal weighting out-of-sample. A MV optimization worse than equal weighting has little practical investment value or interest. The impact of optimizer error maximization properly measured appears alive and very well.
Suggested Citation: Suggested Citation