NYU Courant Professor || Machine Learning & Quantitative Finance || Author, Expert & Speaker || Quant of the Year
I'm excited to share our article "Untangling Universality and Dispelling Myths in Mean-Variance Optimization" with you. This is joint work with Jerome Benveniste and Gordon Ritter. Key highlights: ■ We introduce the notion of mean-variance equivalence (MVE) and prove a new result in portfolio theory, establishing sufficient and necessary conditions for a distribution to be MVE. ■ Perhaps surprising to many, the family of MVE distributions is remarkably expansive, encompassing not only elliptical distributions (a class of symmetric distributions, such as the multivariate normal, multivariate Student-t, and symmetric multivariate stable distributions) but also fat-tailed and asymmetric distributions. ■ We address and dispel a number of myths and misconceptions surrounding mean-variance optimization. The article is available on SSRN: https://lnkd.in/eSMEEqT9 #nyu #nyucourant #trading #quantitativefinance #portfoliomanagement #portfolioconstruction #Markowitz #portfoliotheory #investmentmanagement #meanvariance #optimization #equivakence #universality NYU Courant Institute of Mathematical Sciences NYU Courant Institute of Mathematical Sciences M.S. in Mathematics in Finance, NYU Courant New York University International Association for Quantitative Finance IAQF Society of Quantitative Analysts (SQA) Fabrizio Lillo Charles-Albert Lehalle jean-philippe bouchaud Marcos Lopez de Prado José A. Gutierrez, Ph.D. Ernest Chan Mark Kritzman David Turkington, CFA Will Kinlaw, CFA Dan diBartolomeo Robert Almgren Thierry Roncalli Marie Brière Benjamin R. Auer Prof. Dr. F. Schuhmacher Hendrik Kohrs Bastien Baldacci Bernd Dr Scherer Kent Daniel Raman Uppal Dessislava Pachamanova Reha Tutuncu Raul Leote de Carvalho Lee Maclin Christos K. Giovanni Beliossi Richard Lindsey PhD Keeyan Ravanshid Denis Dariotis Nino Antulov-Fantulin Lukas Sieber Stefan Klauser Amir Sadr
Petter Kolm
2mo
I'm pleased to announce the release of a new version of our article, accessible at: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4747461 It has been accepted for publication in the special issue of The Journal of Portfolio Management in honor of Harry Markowitz.
thanks Petter Kolm. Your paper is very good and easy to read. I found Myth 2 particularly interesting. However, I regret that you did not address the hidden assumption about the correlation between forecasts needing to be the same as the correlation between returns to obtain the solution of Markowitz. In my opinion, for completeness, you should have at least mentioned the article https://arxiv.org/abs/1610.08818 and my paper https://arxiv.org/abs/2201.06635 published in 2024, which focus on that assumption. Additionally, I wrote an article specifically about that hidden assumption, which you can find here: https://www.linkedin.com/posts/sebastien-valeyre-81810041_lhypoth%C3%A8se-cach%C3%A9e-dans-loptimisation-de-activity-7168203735104528384-mt_q?utm_source=share&utm_medium=member_desktop.
Anton Vorobets
4mo
Interesting defense of MVO. How do we assess the quality of the MVE-distributions in relation to their agreement with reality? Empirical return distributions are skewed and fat-tailed in complex ways with asset dependencies being far from constant and linear. If MVE-distributions are not able to approximate that well, it doesn’t really matter that they are skewed and fat-tailed. In relation to the necessity of the maximum expected utility justification, I add this reference: https://www.amazon.com/Risky-Curves-Empirical-Failure-Expected/dp/1138096466 My comment is meant in the most constructive way. I just don't see how the new result addresses the disagreement with reality issues, although I find it interesting.
Ferdinand Haas
4mo
Thank you Petter Kolm! I wonder when people will finally drop the nonsensical statement that MVO was based on the assumption of normality. It was clarified by Markowitz 70 years ago and It must have been so frustrating for Harry to have his arguments so stubbornly ignored. But he was to nice a person to make his point more aggressively.
Steven Pav
4mo
Thank you for sharing the paper, I wish I had written it! I suspect, though, that you will get some pushback about your results from everyone with a pet theory. One difference I have is that I view the "1/n" portfolio as a cautionary tale about parameter uncertainty, and should not be considered outside the framework of noisy estimates. Because of that, it should be analyzed with statistical decision theory, where it can easily be shown to be suboptimal. Seems like an open question whether Markowitz is optimal in that framing, though.
Gordon and Jerome. It’s great to see you two still working together after all these years!
Andreas C.
4mo
Interesting
Ernest Chan
4mo
Enjoyed reading it very much Petter Kolm!
NYU Courant Professor || Machine Learning & Quantitative Finance || Author, Expert & Speaker || Quant of the Year
4moRoberto Violi Dan Cardell CFA Rama CONT Daniel LIEBAU Kenneth A. Blay Eran Fishler Sanjiv Das Ivailo D. Lisa Goldberg Thomas Asikis Richard Berner Ronald Kahn Bastian Bergmann Lionel Martellini Alexander Fleiss Ananth Madhavan Deep Ratna Srivastav Ken Winston Gifford Fong Giuseppe Paleologo Álvaro Cartea Erik Lindstrom Peter Nystrup Nicholas Westray Andrew Chin Hilary Till Harvey Stein