October 31st, 2002, 6:37 pm
If you are interested in empirical rather than analytical copulae, search the Web for Nick Webber from Warwick. However, choosing a copula-based model the core question is whether you want to use use copula to approximate the distribution of times to default (like Mashall and Naldi) or the distribution of some (probably, very abstract) processes associated with the states of the firms under consideration at a particular moment (like James and Webber). In both cases calibration techniques are either weakly developed or rely on historical default data, which are very limited. (The use of the data on defaults occured 10 years ago is also rather questionable in modellyng defaults, say, on a year interval.)Regards,Rector