Skip to content
All library documents

Choosing Copulas for Symmetric and Asymmetric Return Tail Dependence

Article Quant Q&A · Author: develarist
Read the original

Summary

The document frames a model-selection question about dependence between empirical asset returns. It contrasts the Student t copula, which permits symmetric tail dependence, with the Clayton copula, which captures asymmetric tail dependence, and asks which better represents observed returns. It also asks whether the two approaches can be combined when neither is clearly preferable.

No dataset, fitted model, comparison metric, or empirical conclusion is supplied, so the post does not establish that either copula is generally superior or describe a specific blending procedure. The practical lesson is that the choice should reflect the dependence features present in the data, including whether co-movements in extreme outcomes differ across tails. Any comparison or hybrid model would need validation against the target assets and use case; the question alone offers no evidence that a compromise improves fit or forecasts.

Key ideas

  • The Student t and Clayton copulas represent different patterns of tail dependence.
  • The t copula allows symmetric tail dependence, while the Clayton copula allows asymmetry.
  • Copula choice should be assessed against empirical dependence in the assets being modeled.
  • The post provides no data or comparison results to establish which model performs better.
  • Combining the two is raised as a question, but no construction method is given.

Tags

StatisticsMulti-assetRisk management
Full text
# Does the $t$-copula or Clayton copula capture the dependence structure of empirical returns better?


# Does the $t$-copula or Clayton copula capture the dependence structure of empirical returns better?












Which copula captures the dependence structure of empirical asset returns better? the $t$-copula, which has symmetric tail dependence, or the Clayton copula, which has asymmetric tail dependence, and why? I've seen conflicting advice.

If they're both competitive to one another, how to mix or generate a compromise between the two?

Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.