I have recently been tasked to work on fair value derivation for futures on equity indices (non-US). I know that the FVD function in Bloomberg can have a huge discrepancy from markets: where cheap is actually expensive and vice versa. How do I determine the accuracy of my model without waiting for expiry?
The main difficulties I have encountered are:
Interest rates for cost of carry (some of the countries have closed FX markets), so what is the closest approximation I should look at?
Some people have mentioned NDFs but I need help on curve construction, best to ask for quotes from NDF traders?
Dividends - big difference from my calculations vs what Bloomberg's BDVD function shows (divisors differ, confirmed ex-dates and FVD missing out on dividends). How do I determine who is right or wrong?
Accuracy of fair value - given that it is only an estimate, how do people trade the calendar spreads? (Might be OT, forgive me)
And if possible, what has your experience been when deriving fair values?
All pretty good questions given this is your first question on this site. Answer is you do not know with absolute certainty whose dividend model is the right one. But generally you get a pretty good read given you talk to broker dealers regularly. Anything that is estimated can by definition not be accurate so you need to build you own model with the given information at hand, which includes what other market participants are seeing. How wide the spread in estimates is also is of importance and should be taken into account by your model.
I would definitely take into account market dividend rates. Even if you just infrequently trade with certain sell-side desks they generally share their dividend data. If that is not the case but you have access to an in-house vol desk then you should talk to them because they most certainly get the dividend data from whichever counter party they trade with. In fact dividend data are often shared during the price agreement process when trading single stock or index options.