As a finance professional, I often encounter investors who struggle with portfolio construction. Traditional methods like mean-variance optimization or equal weighting have flaws—they either rely on unstable estimates or ignore risk dynamics. Hierarchical Risk Parity (HRP) offers a compelling alternative. It blends hierarchical clustering with risk parity to build diversified portfolios that adapt to market structure. In this article, I dissect HRP, its mathematical foundations, and why it outperforms conventional approaches.
Table of Contents
The Problem with Traditional Asset Allocation
Most investors use one of three methods:
- Equal Weighting: Assigns the same weight to each asset. Simple but ignores correlations and volatilities.
- Mean-Variance Optimization (MVO): Maximizes returns for a given risk level. However, it is sensitive to input errors—small changes in expected returns or covariance matrices lead to drastic weight shifts.
- Risk Parity: Allocates based on risk contribution. More stable than MVO but treats all assets as equally connected, which is rarely true.
HRP improves on these by considering the hierarchical relationships between assets.
Understanding Hierarchical Risk Parity
HRP, introduced by Marcos López de Prado in 2016, combines hierarchical clustering with risk parity. The goal is to group similar assets and allocate risk across clusters rather than individual securities. This reduces concentration risk and enhances diversification.
Step 1: Hierarchical Clustering
Assets are grouped based on their correlation structure. The process involves:
- Computing the Correlation Matrix: Measures how assets move relative to each other.
- Calculating the Distance Matrix: Transforms correlations into distances using:
d_{i,j} = \sqrt{0.5 \times (1 - \rho_{i,j})}
where \rho_{i,j} is the correlation between assets i and j. - Applying Hierarchical Clustering: Uses algorithms like Ward’s method to form a dendrogram (tree structure).
Step 2: Quasi-Diagonalization
The covariance matrix is reordered to place closely related assets near each other. This step simplifies risk allocation by treating clusters as single entities.
Step 3: Recursive Bisection
Risk is distributed top-down:
- Split the Portfolio: The dendrogram is bisected into two sub-clusters.
- Allocate Risk: Assign inverse volatility weights to each cluster.
- Repeat: Apply the same logic within each sub-cluster until individual assets are reached.
The final weight for asset i is:
w_i = \frac{1/\sigma_i}{\sum_{j=1}^N 1/\sigma_j}
where \sigma_i is the asset’s volatility.
Why HRP Outperforms
1. Robustness to Estimation Errors
MVO fails with noisy inputs. HRP, by relying on clustering, is less sensitive to small changes in correlations.
2. Better Diversification
Assets within the same cluster share risk. HRP spreads exposure across clusters, avoiding over-concentration.
3. Adaptive to Market Structure
During crises, correlations spike. HRP dynamically adjusts by merging clusters, preventing unintended bets.
Empirical Evidence
A 2020 study compared HRP, MVO, and equal weighting across US equities, bonds, and commodities. Results showed:
| Metric | HRP | MVO | Equal Weight |
|---|---|---|---|
| Sharpe Ratio | 0.85 | 0.62 | 0.58 |
| Max Drawdown | -18% | -25% | -22% |
| Turnover | Low | High | Medium |
HRP delivered superior risk-adjusted returns with lower turnover.
Practical Implementation
Example: A US Investor’s Portfolio
Suppose we have three assets:
- SPY (S&P 500 ETF)
- TLT (Long-Term Treasury ETF)
- GLD (Gold ETF)
Step 1: Compute Correlations
Historical data shows:
- SPY-TLT: -0.4
- SPY-GLD: 0.1
- TLT-GLD: -0.2
Step 2: Form Clusters
Using hierarchical clustering, SPY and GLD might cluster separately from TLT.
Step 3: Allocate Risk
If SPY and GLD form one cluster, and TLT another, HRP assigns:
- 50% risk to SPY-GLD (split further within)
- 50% risk to TLT
Final weights might look like:
- SPY: 40%
- GLD: 10%
- TLT: 50%
Limitations
- Computational Complexity: Requires clustering algorithms, which may deter some investors.
- Assumes Stationarity: Correlations change over time; periodic rebalancing is necessary.
- Not a Silver Bullet: Works best in diversified portfolios—less effective for concentrated bets.
Conclusion
Hierarchical Risk Parity is a sophisticated yet practical approach to asset allocation. By leveraging clustering and recursive risk balancing, it mitigates the pitfalls of traditional methods. For US investors navigating volatile markets, HRP provides a robust framework that adapts to structural shifts while maintaining diversification. While not perfect, its empirical track record makes it a compelling choice for long-term portfolios.
