What distinguishes Clayton, Gumbel, and Frank copulas and when do I use each?

Question

AcadiFi · Accepted Answer

Archimedean copulas are built from a single generator function phi(u), giving C(u1,u2) = phi^{-1}( phi(u1) + phi(u2) ). The family is tractable and each member captures a different tail asymmetry. ```mermaid flowchart TD A[Archimedean
Copulas] --> B[Clayton] A --> C[Gumbel] A --> D[Frank] B --> B1[Lower tail
dependence] C --> C1[Upper tail
dependence] D --> D1[No tail
dependence] ``` Clayton has lower-tail dependence. Use it when joint crashes matter more than joint booms — equity factors, credit spreads, or Oakspire's commodity-equity pair where both collapse in recessions. The parameter theta greater than 0 controls strength; lambda_L = 2^(-1/theta). Gumbel has upper-tail dependence. It fits insurance losses and catastrophic bond defaults where extreme joint highs dominate. It is also useful for commodity pairs like natural gas and heating oil during demand shocks. Frank has no tail dependence but allows strong central dependence and both positive and negative association. It's common in biometrics and consumer data. For Oakspire's problem: if you observe historical joint exceedances below the 5th percentile far exceeding those above the 95th percentile, Clayton is the right pick. An AcadiFi Level III tip: you can rotate these copulas by 90, 180, or 270 degrees to flip the tail behavior, which expands the modeling toolkit without leaving the Archimedean family.

What distinguishes Clayton, Gumbel, and Frank copulas and when do I use each?

Master Level III with our CFA Course

Related Questions

Practice Questions