Quantitative Wealth and Investment Management (QWIM): Advanced Portfolio Diversification

Abstract

Within the realm of Portfolio Diversification, the classic Mean-Variance (MV) Analysis framework faces scrutiny for its stringent assumptions and the high propensity for creating portfolios that are heavily concentrated in a limited selection of assets. Our study explores various Advanced Portfolio Diversification (APD) techniques-Hierarchical Clustering (HC) with Hierarchical Principal Component Analysis (HPCA) and Dynamic Time Warping (DTW)-to address the inherent estimation challenges associated with MV. These innovative methods refine the diversification process by utilizing the hierarchical and centroid clustering network approaches, thereby mitigating most of the problems incurred by the MV framework and problematic portfolios. When tested against various diversification models and anchoring strategies, these diversified portfolio management methodologies demonstrate superior performance over the traditional Mean-Variance (MV) strategy in the long-term horizon, across multiple risk-adjusted evaluation metrics. The reasons for this result are twofold: (1) the more diverse weight allocation of HC models, and (2) the flexibility of HC models in selecting different risk measures.

Keywords: Quantitative Wealth and Investment Management, Portfolio Diversification, Hierarchical Clustering, Hierarchical Risk Parity (HRP), Hierarchical Equal Risk Contribution (HERC), Hierarchical Principal Component Analysis (HPCA), Dynamic Time Warping (DTW)