2
$\begingroup$

Except from the well-know and well-documented Elliptical (i.e. Gaussian, Student-t) and Archimedean (i.e. Frank, Clayton, Gumbel) copulas used to model the dependency structure between stock returns, are there any other copulas families that could be used?

In particular, I am wondering if there exists other classes of copulas that could provide at least similar performance fit than Elliptical and Archimedean copulas.

$\endgroup$
  • $\begingroup$ hi - just a question, do you find copulas actually any good at modelling correlated stock returns? My experience in areas where they are or have actually been used (principally not equities) is that being very much term - specific they can give rise to very restrictive internal conditional behaviour even if the term marginal probabilities are calibrated. For process based within asset-class stuff or (even across asset class when it makes sense) generally seen correlated brownians in some shape or form if these created enough correlation for the purpose. $\endgroup$ – Mehness Sep 25 '18 at 20:30
  • $\begingroup$ At least they offer a better alternative to measure dependence between risk factors than a simple correlation coefficient! But maybe you know something even better? Unfortunately I have never calibrated copulas between various asset classes so I cannot comment on this. In my case however, only stocks are concerned. Though, the flexibility of copulas fitting might allow to capture cross-asset dependence by 1) fitting marginals separately 2) model their joint behaviour through a "well-chosen" copula. $\endgroup$ – JejeBelfort Sep 26 '18 at 1:47
  • 1
    $\begingroup$ So it's definitely an interesting question - in terms of defining yardsticks for 'well-chosen', and fitting. Unfort have only encountered the copulae you list so maybe an expert can help, but just sth to bear in mind re any copula, if you have some measure of goodness of fit to observed history (a topic in itself), the within-term behaviour of the copula may make the within-term goodness of fit weak, so that reproducing the within-term behaviour requires a term structure of copula parameters (you'll get forward correlations etc). Anyway best of luck hope someone can help! $\endgroup$ – Mehness Sep 26 '18 at 9:24

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.