wilmott.com

Hi, I have a question and believe a lot of you can help me. When we price a CDS, we can find spread from MarkIT, then transfer the spread curve to default probability curve. Then use the default probability to get the value. Does anybody know how to transfer CDS spreads to default probability? Thanks a lot

protection leg = premium leg to derive default intenstiy through which the defaul probability can be calculated straightforwd.

Thank you. I think I figure it out.We can get default probability curve from spread curve. But where does that spreads come from? I mean, if spreads can not be derived from MarkIT or BLG, can we find a website that includes default probability for AAA company's bonds, default probability for AA company's bonds, .......Thanks

It is vaguely possible that CDS were traded prior to 2001, when markit where formed, and that some assumptions about the difference in price between corporate and govvie bonds being due to credit gave people a clue about default probabilities

I think I saw a relevant topic including a website where u could retrieve default probabilities for AAA. If you cannot find it, google it.

generally default probabilities are attached to ratings by rating agencies.. but once could also do a historical rating based study and see what teh mapping would be at an industry level maybe if sufficient companies go rated in that industry ...

QuoteOriginally posted by: xfl883Hi, I have a question and believe a lot of you can help me. When we price a CDS, we can find spread from MarkIT, then transfer the spread curve to default probability curve. Then use the default probability to get the value. Does anybody know how to transfer CDS spreads to default probability? Thanks a lotIt's described in Hull. You need to assume the recovery rate but the actual pricing (for later times than the CDS start) will not depend on it, as you're first dividing by 1-R and then multiplying by it (or the other way round, I can't remember and do not feel like checking ).But to do it correctly, you'll also need a yield curve.

QuoteOriginally posted by: pcggenerally default probabilities are attached to ratings by rating agencies.. but once could also do a historical rating based study and see what teh mapping would be at an industry level maybe if sufficient companies go rated in that industry ...I think the OP is asking about risk-neutral probabilities.

I somehow always thought the probabilities we discuss in relation to CDS are real world only ever !!! Can the enlightened shed some light ?

I would imagine that by "transfer the spread curve to default probability", you mean that you are looking to compute the survival probability numbers given the spread values. The latest round of Bloomberg pricing logic uses an equilibrium-based pull used in setting the equations, that compute the protection leg and fee leg, equal to other.. And then using a root-finding algorithm(newton-rhapson, or the like, etc..) to determine a piece-wise constant hazard rate per given "term".. It is from these computed pwc hazards that you produce your sp's..And the keyword "term" here refers to the intervals in which your spreads are being reported.. Markit Partners generally dishes-out these spreads in the following terms: [1y, 2y, 3y, 4y, 5y, 7y, 10y]. Your model should then compute one term's worth of protection given quarterly fee payments on this protection(this should be modular, quarterly.. as this is currently a market convention, but a variant could be used going forward..). The equations that make up the the protection and fee legs encompass a similar "unknown" variant, being the hazard rate. So, your NR method should solve for this value. Then, from these "found" hazard rates, you compute your sp's.. and dp's are nothing but the reciprocal of the sp's..By using this model, you are implying default given the "market's view" on a given underlying..I find this to be the best (closest to bloomberg pricing) way to produce your dp's..

recently rating is not a great indicator of the default probability that market assigns to an entity, look at where AAA monolines are trading.for a quick and dirty estimation of implied default probs from cds spread never forget the useful equation:yearly_def_prob = cds_spread / (1 - recovery_assumption)

Or implement the Bloomberg CDSW calculation as described:http://www.wilmott.com/messageview.cfm? ... adid=63960