Performance Evaluation of Risk-Based Portfolio Optimisation Models Across Different Market Conditions

Abstract

Risk-based portfolio optimisation has seen a marked increase in popularity in recent years, driven by the notorious unpredictability of asset returns. This dissertation evaluates the out-of-sample performance of six risk-based portfolio optimisation models across two market crises and one period of relative stability, using portfolios composed of the top 30 and 100 stocks by market capitalisation in the United States stock market. Results reveal no significant variation in model performances across different market conditions. Notably, the Hierarchical Equal Risk Contribution (HERC) model with average linkage excelled in the 30-stock portfolio, generating the highest returns but with elevated portfolio concentration and turnover. However, its performance declined sharply in the 100-stock case, indicating its vulnerability to larger, more diversified asset pools. In contrast, the Hierarchical Risk Parity (HRP) and Equal Risk Contribution (ERC) models demonstrated more balanced outcomes, with HRP consistently outperforming ERC in terms of risk-adjusted returns, albeit with higher turnover. Additionally, the Maximum Diversification (MDV) and Maximum Decorrelation (MDC) models consistently outperformed the HRP and ERC models in terms risk-adjusted returns, but at the cost of higher portfolio concentration and turnover.