Optimal Portfolio Diversification via Independent Component Analysis

A natural approach to enhance portfolio diversification is rely on factor-risk parity, which yields the whose risk equally spread among a set of uncorrelated factors. The standard choice take variance as measure, and principal components (PCs) asset returns Although PCs are unique useful for dimension reduction, they an arbitrary choice: any rotation results in This problematic because we demonstrate that factor-variance-parity some PCs. More importantly, choosing does not account higher moments returns. To overcome these issues, propose using independent (ICs) factors, maximally independent, care about We IC-variance-parity helps reduce return kurtosis. also show how exploit near independence ICs parsimoniously estimate factor-risk-parity based value at risk. Finally, empirically portfolios outperform those PCs, several state-of-the-art benchmarks.