From what I understand, an FVA is a swap on the future implied volatility of at-the-money, which is guaranteed by a front/straddle start option. In terms of sensitivity, this is similar to that of forward/var start swaps, as you currently have no gamma and are exposed to forward flight. However, it is different from the fact that you are exposed to Standard Vega distortions of vanilla options and MTMs due to distortions, given that the spot moves away from the initial trading date. Looking at FX in particular, but I think it`s a general question. any good reference would be appreciated. FVAs are not mentioned in Derman`s paper («More than you ever wanted to Know about volatility swaps») A start-up volatility swap on arrival is really an exchange on realized future volatility. In another thread, I wrote that Rolloos & Arslan wrote an interesting paper on price reconciliation without a Model spot Starting Volswap. FVA has nothing to do with Volswaps. This is a volatility agreement and you agree to a contract to buy/sell a vanilla forward starting option with black Scholes parameters (except for the spot price) that have been set today. This option is used to commit to implied volatility in advance and usually resembles trading a longer option and cutting your gamma exposure with another option, whose expiration date matches the departure date in advance, constantly rebalancing yourself, so that you are gamma-flat.

In a very recent (quite compressed) working paper, I saw that Rolloos also deduced a model-free pricing approach for Forward Starting Volswaps: I think the underlying idea is that the future ATM IV is a proxy for expected future volatility. However, ATM IV, Spot or Future, is not a good proxy for expected volatility when there is a significant correlation between the underlying and volatility. The mathematics in this last work looks correct — but I have not yet seen numerical tests of the result without a model. Someone tested the latest result of Rolloos, any comment/ideas about it?. . . .