Skip to main content

LUP Student Papers

LUND UNIVERSITY LIBRARIES

Decomposing Hierarchical Portfolio Performance: Tree Structure or Allocation Rule?

Bursic, Adam LU (2026) NEKH02 20261
Department of Economics
Abstract
Hierarchical portfolio allocation methods produce diversified portfolios while avoiding the instability of mean-variance optimisation. The framework depends on two separable choices: how assets are grouped and how capital is allocated across those groups. This thesis asks which of these choices matters more for out-of-sample portfolio performance. The empirical design crosses four linkage methods with four allocation kernels, evaluated on a multi-asset ETF universe from 2008 to 2026 across 3,200 resampled walk-forward paths. The results show that allocation kernels account for most differences in risk-adjusted performance, while the linkage method mainly affects turnover. Risk-aware kernels improve risk-adjusted performance and reduce... (More)
Hierarchical portfolio allocation methods produce diversified portfolios while avoiding the instability of mean-variance optimisation. The framework depends on two separable choices: how assets are grouped and how capital is allocated across those groups. This thesis asks which of these choices matters more for out-of-sample portfolio performance. The empirical design crosses four linkage methods with four allocation kernels, evaluated on a multi-asset ETF universe from 2008 to 2026 across 3,200 resampled walk-forward paths. The results show that allocation kernels account for most differences in risk-adjusted performance, while the linkage method mainly affects turnover. Risk-aware kernels improve risk-adjusted performance and reduce downside risk relative to 1/N. Practitioners should treat the allocation rule as the primary modelling choice in hierarchical portfolio construction, with linkage choice used mainly to manage implementation stability. (Less)
Please use this url to cite or link to this publication:
author
Bursic, Adam LU
supervisor
organization
course
NEKH02 20261
year
type
M2 - Bachelor Degree
subject
keywords
Asset allocation, Hierarchical Risk Parity, Diversification, Clustering, Linkage methods, Risk-based allocation, Backtesting
language
English
id
9242540
date added to LUP
2026-10-12 09:29:53
date last changed
2026-10-12 09:29:53
@misc{9242540,
  abstract     = {{Hierarchical portfolio allocation methods produce diversified portfolios while avoiding the instability of mean-variance optimisation. The framework depends on two separable choices: how assets are grouped and how capital is allocated across those groups. This thesis asks which of these choices matters more for out-of-sample portfolio performance. The empirical design crosses four linkage methods with four allocation kernels, evaluated on a multi-asset ETF universe from 2008 to 2026 across 3,200 resampled walk-forward paths. The results show that allocation kernels account for most differences in risk-adjusted performance, while the linkage method mainly affects turnover. Risk-aware kernels improve risk-adjusted performance and reduce downside risk relative to 1/N. Practitioners should treat the allocation rule as the primary modelling choice in hierarchical portfolio construction, with linkage choice used mainly to manage implementation stability.}},
  author       = {{Bursic, Adam}},
  language     = {{eng}},
  note         = {{Student Paper}},
  title        = {{Decomposing Hierarchical Portfolio Performance: Tree Structure or Allocation Rule?}},
  year         = {{2026}},
}