Hierarchical Portfolio Allocation in an Active Management Framework

Abstract

This master’s thesis focuses on developing and evaluating a hierarchical portfolio allocation algorithm that combines hierarchical clustering and Markowitz Modern Portfolio Theory while being adapted to an active management framework. Sixteen different constellations were constructed and evaluated on equities return data from 01/03/1990 to 10/01/2022, using three different sets of observations as input and five different performance measures. The results demonstrate that the combination of Equal Risk Contribution and Single Linkage generates the best outcomes. In general, the results also show that Tracking Error is significantly smaller when Equal Risk Contribution is used as a between-cluster allocation method. Moreover, the choice of linkage criteria is crucial for cluster size and the numerical stability of the associated sample covariance matrices. For instance, Single Linkage produces the smallest set of clusters, followed by Group-Average Linkage, Complete Linkage, and Ward’s method. In addition, the ordering of the leaves in the hierarchical structure did not have a significant effect on the results. The suggested hierarchical portfolio allocation algorithm performs consistently and is able to capture the hierarchical structure between assets during different market conditions.