Enhancing the Inverse Volatility Portfolio through Clustering

This article presents a novel approach to portfolio construction, termed cluster-enhanced inverse volatility, designed to enhance the effectiveness of traditional inverse volatility portfolios. The goal of the method is to cluster the data to meet the two conditions—the same Sharpe ratios across assets and equal pairwise correlations—under which the inverse volatility portfolio becomes theoretically equivalent to the mean–variance optimal portfolio. The authors show that, as the asset data increasingly meet these two conditions, the cluster-enhanced inverse volatility portfolio approaches the mean–variance optimal portfolio. Empirical evidence from various datasets indicates that the authors’ cluster-enhanced inverse volatility portfolios outperform their traditional counterparts, particularly in portfolios with a large number of assets.