Niklas Rönnberg  presenterar sitt examensarbete

Pricing of standard and non-standard CDO-tranches using the one factor NIG copula
	
Abstract
	
The use of a one factor Gaussian copula model is market standard when
pricing CDO tranches. It is, however, widely documented in the
literature that when calibrated to CDO market data this model produces
a distinct correlation skew. The skew is evidence of the shortcomings
of the one factor Gaussian copula model. Researchers have proposed the
use of a one factor normal inverse Gaussian (NIG) copula model to
eliminate the skew. They have further described the NIG copula model
as being a good choice for the valuation of non-standard CDO tranches;
however none has shown numerical results of such procedures. The focus
of this thesis is to verify that the NIG copula model can be well
calibrated to market data using a flat default correlation level, as
well as provide numerical examples of the pricing of non-standard CDO
tranches using the NIG copula model.

The main conclusions reached are that the NIG copula model can be
calibrated to CDO market data quite well using a flat default
correlation and that most non-standard tranches can be decently
priced. However, some problems have come to light, the most severe
ones being that when valuing non-standard CDO tranches the results of
the one factor NIG copula model can be hard to interpret and that even
though the model manages to price most non-standard tranches well,
some are poorly priced. The reason to why it is hard to interpret the
results is that for some dates a certain tranche may be under-priced
whereas for other dates the same tranche may be over-priced.