Hierarchical PCA and Modeling Asset Correlations

TL;DR

This paper introduces Hierarchical PCA and a clustering algorithm to better model and understand cross-sectional stock correlations across global markets, improving upon traditional PCA methods.

Contribution

The paper presents a novel hierarchical PCA approach and a clustering algorithm for identifying homogeneous stock groups, enhancing correlation modeling across diverse markets.

Findings

Hierarchical PCA outperforms classic PCA in modeling correlations.

The clustering algorithm effectively identifies homogeneous stock clusters.

Application to multiple markets reveals distinct correlation structures.

Abstract

Modeling cross-sectional correlations between thousands of stocks, across countries and industries, can be challenging. In this paper, we demonstrate the advantages of using Hierarchical Principal Component Analysis (HPCA) over the classic PCA. We also introduce a statistical clustering algorithm for identifying of homogeneous clusters of stocks, or "synthetic sectors". We apply these methods to study cross-sectional correlations in the US, Europe, China, and Emerging Markets.