Quant Memo

Correlation Clustering

Grouping assets or signals by correlation structure to manage diversification and risk.

Definition

Correlation clustering (or hierarchical clustering on correlation/distance) groups assets that move together. Useful for portfolio construction (diversify across clusters) and risk (avoid over-concentration in one cluster).

Why it matters

  • Equal weight across 10 assets that are highly correlated gives little diversification.
  • Clusters can correspond to sectors, regions, or risk factors.

Common mistakes

  • Using correlation in trending regimes (correlation is not stable).
  • Treating cluster count as fixed; it can change over time.
  • Ignoring that correlation is not a distance (use ( 1 - \rho ) or similar for clustering).

Typical use

Define clusters, then equal weight within cluster and across clusters; or use for constraint (max weight per cluster).

Linked strategies

ShareTwitterLinkedIn