Hi all,I have an issue concerning implied vol for non fungible assets like the majority of commodities (let alone precious that in a trader perspective behave like FX).Let's say you have the first 2 NatGas futures maturities t1 and t2 on which you can observe an implied (spot) vol of 50% and 30%. The difference in vol% it's huge due to the strong seasonal pattern of the market (the example could be real). How would you compute fwd vol between t1 and t2? Actually it seems to me that the method used for fungible assets would be misleading since I don't expect the sum of vol in the two time slices to equal the t2 spot vol. This become very relevant when pricing path dependent options. One example could be the asian option fron t0 to t2 (some practitioners' approach i saw around consist in just averaging the two vols but i don't have it clear)If you can help with ideas or suggestions you're welcomehave a nice day

just wanted to point you to this link...not quite related to fwd vol for commodities but it's a start:http://www.wilmott.com/messageview.cfm? ... TARTPAGE=1

yes actually I surfed the forum before posting but i didn't find anything strictly related. There are some threads that just mention the issue (in the forum search under "fungibility").Thank you anyway.

Your two NG futures are definitely fungible. Fungibility is not the issue here. The issue is vol regime and as you say, seasonality.Your assumption is that both your futures vols are from the same barrel and that they have the same distribution of local variance over time... which in fact is not the case.To point you in the right direction, let me take a simple example using your two futures and with two periods.- for the first period [To;T1], the futures NG1 has a vanilla vol of 50%,- for the second period [To;T2], the futures NG2 has a vanilla vol of 30%,What the market is saying on this is two things:- the local/instantaneous correlation of NG1 vs NG2 during [To;T1] is low, and in all cases not 1, otherwise we would probably in a case of negative forward vol,- the forward vol of NG2 is higher than it current cumulative/quoted vol. Again, same reason.Think of the vol of NG2 during the [To;T1] period as priced with a scalar versus the quote NG1 vol, and of the vol of NG2 during the [T1;T2] period as priced with a scalar versus NG1 vol during [To;T1] and you will be almost there. In other words, we have something like this:volNG2[To;T2] = Integral {[To;T1] a1 x volNG1 dt } + Integral {[T1;T2] a2 x volNG1 dt }. Gold is easy. The curve has no or little convexity. All point of the forward curve are very similar in terms of instantaneous volatilities. NG is probably the commodity at the extreme opposite of the spectrum (when we want to have fun in interviews to see how candidates really undestand their models, we also take Power). So with gold you have dF/F = s dW(1) type moves (FX like), but here with NG you have a second or even a third noise source. To picture this, you can use the simple model dF/F = s dW(1) + s(t) exp(-k.(T-t)) dW(2). Gold would have k=0.06 while NG would probably give you results with k = [2.8;3.5] (depending on the level of s and cheating a little...).To price path dependent options, you need a forward curve model to calculate your serial vols (vols on long dated forwards for short dated tenors).Good luck.Jezza

Jezza thank you very much for the accurate answer.Can you just clarify some points further?a) The equation: ...volNG2[To;T2] = Integral {[To;T1] a1 x volNG1 dt } + Integral {[T1;T2] a2 x volNG1 dt}... doesn't have multiple solutions? I have to find a1 and a2 but I know only VolNG1[T0;T1]b) in the function ...dF/F = s dW(1) + s(t) exp(-k.(T-t)) dW(2.... what is the difference between s and s(t)? Which model/paper do you refer to? Is it a derivation of forward price model with sthocastic convenience yield?c) To make an example. In pricing with a montecarlo an arithmetic average asian option between T0 and T1, written on the spot price of gas (not the futures) with the vol structure seen above, how would you model the vol?thks

I am a Jr. quant working on the same problem (Forward implied volatilities in commodities) and have revived this thread to limit redundancy. I have considered various approaches, such as the method proposed by Pilipovic (2007) to create a forward volatility matrix or Rebonato?s (1999/2002) instantaneous volatility parametrization, however, although these approaches create the desired terminal de-correlation between contracts I am reluctant as I don?t see how I can hedge the imposed volatility structure. To be a tractable model, I believe that I need to price the exotic I am interested in with a vol that can be replicated and don?t see how to do this with these methods.I am now considering a slightly more ad-hoc method to relate all the contracts in the strip with each other but have never taken this type of approach before and could use any advice users are willing to share. Certainly I have the correlation matrix but I imagine that I will need to define the shape or relation of each contract in a more meaningful way and currently I am not sure how to approach it. Any thoughts would be greatly appreciated. Thanks!!