July 9th, 2004, 10:53 am
1) LGD is (in CDS language) 100% - recovery rate assumption. I..e what do you as the protection buyer receive when an event defaults? Usually assumed to be 60%. Or in EDS language, what is the size of the binary payment you receive when the equity first falls below 70% of the original value? I understand this is usually set at a fixed binary amount (50%?).2) EDS spread in market just means: what do I have to pay on a running basis to buy protection such that PV today = 0. This is used to calibrate hazard rate of a poisson process, and thus the probability of an event occuring at a time t.Let me turn your question around: do the mathematics of hazard rates / poisson distributions (and thus the maths of a CDS model) understand or care about the definition of an 'event' - i.e. whether an event is either a default, or a 70% drop in equity price?You can look at an EDS in 2 ways:(i) like a (binary) cds except the event is defined differently. Key parameter would be spread.(ii) like an equity option. Key parameter would be volatility.There should be a link between these two views, i.e. for a EDS market spread you should be able to back out an implied volatility and vice-versa.3) asset correlations are can be viewed as 'gaussian copula' correlations - see the Risk Metrics paper by D Li in1999. Thus if you want asset correlations use correlations that are relate to products where the market quotes are based on a gaussian copula model. in other words, use implied iBoxx / Trac-x or FTD broker quotes. People also say (via Merton) that equity correlations are a proxy to asset correlations, because the total value ofa firm is constant - changes in equity must be offset by changes in debt. Thus I don't think you would calculate asset volatilities (unobservable anyway) - you just basically say equity correlation (observable) = asset correlation. Some people pre-process equity correlations, though I don't have any details.In the absence of EDS implied correlation data, the most relevant correlations for EDS would be probably be equity correlations, although you'd need to factor in the skew, i.e. the fact that it is empirically seen that implied correlation is a function of tranche seniority. Can't see an obvious way to do this yet - the market hasn't evolved enough. It is likely that you will need to reserve against your guess of correlation.
Last edited by
Herbie on July 8th, 2004, 10:00 pm, edited 1 time in total.