Can anyone assist in calculation of PnL attribution (valuation method) using Bloomberg SWPM for options?

By replacing fundamental factors (curve, settlement, IV) unexplained is enormous. I suppose that it could arise from IV interrelation with other risk factors, but unfortunately no clues how to solve it. Any suggestions?!

Do you have a specific numerical example? Otherwise, no one knows what you are talking about. If you bump vol by 1% and the expected vega PnL is not the same as the actual PnL, then maybe the attribution is based on a derivative method, not actual bumping of the parameters. If the problem arises as a result of bumping multiple parameters, then it’s more complex and depends on how Bloomberg handles multiple parameter changes in PnL attribution.

I agree that you should start with BB Help, as they probably have the calculation methods documented.

About BB help - they more help on technical issues. I’m looking for more professional advice. For calculations I’m using valuation method, not Greeks therefore vega approach doesn’t help - I’m just replacing parameters used in valuation. Even without direct linkage with Bloomberg, it would be nice to hear some professional thoughts on how IV interrelation with other parameters could be solved… Theory is quite clear, however practical side is quite foggy…

How about your BB consultant? They should at least be able to direct you to any documentation on it. There are smart people at BB who can help you, you just have to find the right ones.

quant.stackexchange.com would probably be better for this type of question than AF, but they might tell you the same things we’re telling you (to get bb to explain it). At a minimum you would have to follow ohai’s advice and post sufficient details of what you’re doing.

Well, Ance, I can comment on what I know about PnL attribution, and this might or might not be helpful to your situation. Bloomberg most likely uses a sensitivities-based attribution. So, let’s say you are modeling a vanilla European option. You can easily get greeks for delta, gamma, vega, etc. - either analytically or through reval method. The naive greeks based PnL attribution is just to take these sensitivities, multiply them by the market data change (example: 1% of spot), and say that is the attribution. Unfortunately, this method does not take into account cross greeks sensitivity and thus, you will end up with big unexplained PnL. It might also fail to take into account non-linear sensitivities. For instance, does a 2% vol bump produce 2x the vega PnL as a 1% vol bump? Probably not.

A good PnL attribution has small or zero unexplained attribution. One way to reduce unexplained PnL is to calculate the Nth order cross sensitivities between all the greeks and incorporate that into the PnL explain. However, this is resource intensive and not many people want to deal with it.

Another method to reduce unexplained PnL is a step by step calculation. For instance, first you bump spot and get your delta/gamma PnL. Then, with this new spot, bump interest rates for the rho attribution. Then, with the new spot and the new rates, bump vol for vega PnL. Repeat this process until all explanatory factors are accounted for. This method produces low unexplained PnL, but the order of parameters does affect the actual attribution results. For small bumps, the noise will be small though.