A1072
Title: Enhanced covariance matrix estimators in portfolio management: Comparing parametric and nonparametric approaches
Authors: Sarah Alsaed - University of Essex (United Kingdom) [presenting]
Spyridon Vrontos - University of Essex (United Kingdom)
Abstract: Traditional mean-variance optimizers that rely on raw estimates of the covariance matrix tend to be unstable due to the substantial amount of noise in the sample covariance matrix. The aim is to refine and enhance the covariance matrix used in portfolio optimization problems. Shrinkage techniques, applications of random matrix theory, and a variety of hybrid approaches are examined. In addition, the parametric DCC GARCH model is examined to estimate dynamic conditional correlations. A comparative analysis using a rolling window approach is employed to test the efficiency of each covariance matrix estimator in achieving robustness and better performance metrics. A plethora of investment strategies is tested and evaluated.
