Covariance Estimation Methods for Mean-Variance Portfolios
Summary
The document surveys alternatives to the sample covariance matrix used in Markowitz mean-variance portfolio construction. It names RiskMetrics exponentially weighted moving average estimators from 1996 and 2006, multivariate DCC-GARCH, and Ledoit-Wolf shrinkage estimators. The latter include a 2003 approach based on a single-factor index model, a 2004 estimator that shrinks toward the identity matrix, and a later nonlinear shrinkage paper.
The response points readers toward software implementations and publications, but does not explain estimator formulas, compare their assumptions, or report empirical results. It is therefore a list of candidate methods rather than guidance on which method performs best. Choice depends on the asset universe, data, and portfolio objective; the document offers no comparative evidence or practical selection criteria.
Key ideas
- EWMA covariance estimates give greater weight to recent observations.
- Multivariate DCC-GARCH is presented as another model-based alternative to sample covariance.
- Ledoit-Wolf methods use shrinkage, with examples targeting a factor model or the identity matrix.
- The cited discussion lists implementations and papers but supplies no empirical comparison or selection rule.
Tags
Full text
# Widely accepted methods for coming up with the co-variance matrix of assets? # Widely accepted methods for coming up with the co-variance matrix of assets? ### Question What are the widely accepted ways for coming up with co-variance matrix of assets after the Markowitz's modern portfolio theory? ### Question explained in more detail - After Modern portfolio theory was introduced, to my best knowledge, there are bunch of new theories of co-variance methods came out. - Exponential co-variance matrix and Ledoit-Wolf are two examples of those new methods. - Can somebody tell me the more recent advancement of the co-variance matrix to create a portfolio? If the paper has a github code written in Python, it would be more than welcome. ## Answer by develarist (score 6, accepted) https://quant.stackexchange.com/a/46058 Multivariate volatility models for replacing the sample covariance matrix with in the mean-variance portfolio selection model: - RiskMetrics 1996 EWMA (Exponentially weighted moving average) covariance matrix - RiskMetrics 2006 EWMA covariance matrix - Multivariate DCC-GARCH covariance matrix Jon Danielsson "Financial risk forecasting" has EWMA and GARCH for R and Matlab and looks like Python now too. Kevin Sheppard's MFE toolbox for Matlab and Arch package for Python have EWMA and GARCH. RiskMetrics 2006 EWMA for Python is here. - Ledoit and Wolf (2003) covariance matrix based on the single factor index model: Ledoit, O. and Wolf, M. (2003). Improved estimation of the covariance matrix of stock returns with an application to portfolio selection. Journal of Empirical Finance, 10:603-621. Matlab code - Ledoit and Wolf (2004) covariance matrix based on the identity matrix: Ledoit, O. and Wolf, M. (2004). A well-conditioned estimator for large-dimensional covariance matrices. Journal of Multivariate Analysis, 88:365-411. Matlab code Python code (Sci-kit learn package) See the second author's publication page if links change at https://www.econ.uzh.ch/en/people/faculty/wolf/publications.html ## Answer by babelproofreader (score 2) https://quant.stackexchange.com/a/45952 Ledoit and Wolf have a new paper ( November 2018 ) called "Analytical Nonlinear Shrinkage of Large-Dimensional Covariance Matrices" which has MATLAB code for the procedure at the end of the paper. The paper can be downloaded at SSRN.
Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.