I am trying to replicate the JPM CDSW calculator in Excel according to the attached article. I believe I understand the intuition behind the formulae. However, I am a little confused by the meaning of "par spread" and why the curve in the example appears to slope downwards (one would expect the curve to slope upwards typically).I can replicate the "Approx SN" in Excel if I treat each period in isolation (i.e. I can replicate the figures for period 3 if I only have two periods 0 and 3). However, if I use the summation formula and have intermediate periods between 0 and 3, I get something completely different.I would be grateful if someone could point me in the right direction. I can supply an example spreadsheet if necessary.Many thanksJon

Without seeing your calcs, my only guess is you're not taking the differences correctly, can you put the post the calcs for the first three periods?

I've managed to get it working now (see attached). I am calculating default probabilities from the par spread using "goal seek". In the attached example they deviate quite a lot from Bloomberg at the 7yr and 10yr point. Do you think this is because goal seek isn't very good at hitting the required target or because Bloomberg uses more intermediate points on the curve that it doesn't show? Is there a better way than goal seek? I guess it's possible to write the JPM formula in terms of PND but have yet to try it. Sorry it's in a rubbish format but I think you can cut an paste it into excel easily to view it better.Period Yield DF PND (obtained by goal seek) Annuity Default Accrual Contingent Approx S Default probability Difference between this and BB0.50.02133 0.989502636 0.991299653 0.49044681 0.002152254 0.00516541 0.010486032 0.008700347 0.0002003471 0.02164 0.97881837 0.983436722 0.971749775 0.004076349 0.009783238 0.010025596 0.016563278 -0.0001367222 0.02328 0.955016837 0.967432064 2.386110494 0.013870963 0.024119479 0.010049861 0.032567936 -0.0005320643 0.02481 0.929115994 0.951900995 4.732733519 0.027314648 0.047726226 0.010026417 0.048099005 -0.0013009954 0.02627 0.901474539 0.936890165 9.425623227 0.054180154 0.094913481 0.010012178 0.063109835 -0.0020901655 0.02766 0.872474488 0.922450322 18.8114782 0.107893566 0.189266872 0.010003867 0.077549678 -0.0032503227 0.03023 0.811821697 0.8778425 38.2434452 0.245701532 0.392702865 0.010202951 0.1221575 0.010957510 0.033251 0.721010613 0.71 76.59733983 0.636713803 0.836287305 0.010827961 0.29 0.135

I would put in all the quarterly points and check your discount curve, but don't expect to match BBG exactly

I agree with Wibble. You have to generate the curve for all quarterly points to match it more closely because the quoted spreads that you are using are usually quarterly CDS's. Therefore the farther you go the more error you encounter. Also you need to be sure that you are using the same discount curve as BBG or else your numbers will always be slightly different.