Comparison of the Effectiveness of Nonlinear Shrinkage and Linear Shrinkage Toward Constant Correlation Model in Portfolio Optimization

Abstract

Modern Portfolio Theory (MPT), established in 1952, laid the foundation for portfolio optimization, emphasizing diversification and the balance between risk and return. However, the traditional mean-variance model often needs to improve when the number of assets is large. Shrinkage estimation, especially nonlinear shrinkage, was developed to address this limitation. This study directly compares the effectiveness of nonlinear shrinkage with linear shrinkage, specifically on the Shrinkage toward Constant Correlation Model (SCCM), which has been proven to provide superior investment results compared to other linear shrinkage models in the Vietnam stock market from 2019 to 2023. Using stock price data from HOSE, the minimum variance portfolio (GMVP), and backtesting techniques, this study evaluates portfolio performance based on cumulative return, annual volatility, Sharpe ratio and Sortino ratio criteria. The results indicate that nonlinear shrinkage outperforms the SCCM across the criteria of annual volatility, Sharpe ratio, Omega ratio, and Sortino ratio. The research affirms the superiority of the nonlinear shrinkage technique in portfolio optimization on the Vietnam stock market, reinforcing prior studies and aiding investors in making effective investment decisions.