Large Scale Covariance Estimates for Portfolio Selection
CEIS Research Paper
We propose an estimator of the Covariance Matrix (SWSE) of a large number of assets. This estimator improves the SimilarityWeighted Estimator (SWE) introduced in Munnix et al. (2014), by combining it with the shrinkage estimator of the sample covariance matrix towards the market factor developed by Ledoit and Wolf (2003). We compare the performance of our estimator to some alternatives already available form the literature and the industry. For this purpose we analyse both statistical and economic measures associated to the Global Minimum Variance (GMV) Portfolio, composed by the stocks included in the S&P 500 index and computed using the different estimators considered in our comparison.
Keywords: Portfolio selection, large scale covariance matrix, precision matrix, shrinkage, minimum variance, market dynamics
JEL codes: C55, C58, G11
Date: Friday, August 7, 2015
Revision Date: Friday, August 7, 2015