Hierarchical Risk Parity: Cluster-Based Allocation for Better Diversification

Hierarchical Risk Parity (HRP) is a portfolio construction method developed by Marcos López de Prado that uses machine learning clustering techniques to group assets by similarity before allocating capital. Unlike minimum variance optimization, which directly solves for the weight vector that minimizes portfolio variance, HRP first organizes assets into a tree structure based on their correlations, then allocates capital top-down through the tree, equalizing risk at each branching point.

How It Works

The HRP algorithm proceeds in three stages. First, hierarchical clustering groups assets based on their return correlation structure — assets that behave similarly are placed in the same cluster. Second, the cluster tree is reordered to minimize the distance between adjacent assets, creating a natural sequence from most similar to most different. Third, capital is allocated recursively through the tree: at each split, capital is divided between the two sub-clusters in inverse proportion to their aggregate volatility, ensuring each cluster contributes equal risk.

Why HRP Over Traditional Optimization

Traditional mean-variance and minimum variance optimization are notoriously sensitive to estimation errors in the covariance matrix. Small changes in estimated correlations can produce large changes in optimal weights, leading to unstable portfolios that require frequent rebalancing. HRP avoids this instability by not inverting the covariance matrix — the step that amplifies estimation errors in traditional optimization. Instead, it uses the correlation structure only for grouping, which is much more robust to measurement noise.

López de Prado demonstrated that HRP produces portfolios with better out-of-sample risk-adjusted returns than both minimum variance and equal-weight approaches, particularly in environments with many assets and shifting correlation structures.

Strategies That Use HRP

HRP is available as an allocation method option in the platform's Builder tool, where users can construct custom strategies using hierarchical risk parity weighting. It is not currently used as the default method in any of the published research strategies, which predate its widespread adoption. However, HRP's stability advantages make it a natural complement to momentum-based selection in custom strategy construction.

Practical Advantages

HRP's primary practical advantage is stability — portfolio weights change gradually even when correlation estimates fluctuate, producing lower turnover and more predictable behavior than optimization-based alternatives. This stability makes it particularly well-suited for portfolios with many positions where correlation estimation is noisiest and optimization instability is most severe.