Spectrally-Corrected and Regularized Global Minimum Variance Portfolio for Spiked Model

TL;DR

This paper introduces SCRGMVP, a novel portfolio optimization method that improves risk estimation by combining spectral correction and regularization under a spiked covariance model, outperforming traditional methods.

Contribution

The paper proposes a spectral corrected and regularized global minimum variance portfolio model that enhances risk estimation accuracy under the spiked covariance model.

Findings

Outperforms traditional covariance estimation methods in simulations.

Has lower computational complexity than existing methods.

Demonstrates improved portfolio risk management in synthetic and real data.

Abstract

Considering the shortcomings of the traditional sample covariance matrix estimation, this paper proposes an improved global minimum variance portfolio model and named spectral corrected and regularized global minimum variance portfolio (SCRGMVP), which is better than the traditional risk model. The key of this method is that under the assumption that the population covariance matrix follows the spiked model and the method combines the design idea of the sample spectrally-corrected covariance matrix and regularized. The simulation of real and synthetic data shows that our method is not only better than the performance of traditional sample covariance matrix estimation (SCME), shrinkage estimation (SHRE), weighted shrinkage estimation (WSHRE) and simple spectral correction estimation (SCE), but also has lower computational complexity.