This is not exactl modelling question, but would like to be directed to the right references.If we are pricing a long-dated option, say a 10yr structured product, but we can only observe short-term options on the markets, such as exchange traded options on the same underlying. What is the most practical way to come up with the long vols/skews. The purpose if not only for pricing but also for validating desk's marks.

I smell an actuary.

a friend told me he became a Quant because he was not qualified as an actuary. he made WSJ front page later on for one of his contributions to the finance world. ironically, that contribution rooted from his work as a actuary.anyway, i 'llstill appreciate for any sharing about my original question.

Reminds me a bit of Rutherford who said the only science was Physics, as all other sciences can be reduced to this. Not he was awarded a nobel prize in Chemistry for the scattering of an alpha particle on gold leaf in retribution.

Merriam-webster explains:physicist: (archaic ) a person skilled in natural science!So, no physicist is willing to share toughts about long vol methodology? such as find a proxy security etc.

QuoteOriginally posted by: GWB ...say a 10yr structured product...The purpose if not only for pricing but also for validating desk's marks...What this means is:1) that you have a bunch of this crap on the books already2) you don't know how to price it, you aren't clear on what volatility is, or how to calculate it3) you have hedged it, but you don't understand your hedge instruments either, and are trying to understand your OTC dealer marks I recommend you go to the Wilmott Bookshop and order Rebonato's book "Volatility and Correlation - the Perfect Hedger and the Fox."

PaperCut, thanks for your reply.But your comment sounds too textbook (sorry) and it might be most useful for a FE student's interview purpose.Two point from you feedback are particular interesting:- The hedge. I am not sure the desk hedges the long vega positions. Does not the desk usually delta hedge the position and only manage the vega (so it's under the risk limit)? but again you have to calculate the risks based on a bunch of crap...- The hedgeing counterparty. If there is any counterparty (ideally, a couple of them) is willing to provide a hedge, the market is then observable. i will use the average of the c/p's vol to mark my book. There won't be the original question.Any one from the practical world?

BTW, PaperCut I will mention Rebonato's book to the desk to see whether they buy it.

GWB, with what do you hedge your 10y vega, out of curiosity? BTW PaperCut I didn't know about the 2ed of the Rebonato book. I have his IR book and just bought the fox one on amazon.ca.

Wouldn't your modeling approach depend on the nature of the underlying? I would argue that you need to think about the potential 10-year histories of the underlying and ask whether those histories are priced into the short-term options. For many categories of underlying -- especially equities -- I would argue that highly divergent paths (extreme growth, bankrupcty, structural makeover, industry shifts) may be invisible to the short-term options. On the other hand, interest-rate-linked underlying instruments would be a bit tamer (barring the potential for bankruptcy).Have you done any backtesting? That is have you picked a representative selection of similar underlying instruments from 10 years ago, looked at the option price structure from 10 years ago, and then analyzed the relationship between 10-year-old option implied volatility and the actual 10-years of underlying data?

Traden4Alpha, thanks for your sharing.yes, I agree with you that understanding nature of the underlying is critical. The products that are covering are broadly ranged, i.e. from energy to IR/CR to equity. That makes the job a little bit harder.Taking the equity class as an example, among other things we have studied newer securities (which have neither much historical prices and nor listed options) using proxy securities. for indicies, we have tried to manufacturing vols using individual vols of components and historical corr, etc, etc, etc.For emerging credit derivatives, we have used something like worst-case-in-region etc approach.However, I understand this is an open topic and would love to pick brains from more seasoned wilmottors.