# IVF and implied distribution of underlying in John Hull's book

There is a statement in John Hull's book Options, Futures and Other Derivatives 9th page 633 for the relation between implied volatility function (IVF) and implied distribution of asset in future time.

When it is used in practice the IVF model is recalibrated daily to the prices of plain vanilla options. It is a tool to price exotic options consistently with plain vanilla options. As discussed in Chapter 20 plain vanilla options define the risk-neutral probability distribution of the asset price at all future times. It follows that the IVF model gets the risk-neutral probability distribution of the asset price at all future times correct. This means that options providing payoffs at just one time (e.g., all-or-nothing and asset-or-nothing options) are priced correctly by the IVF model. However, the model does not necessarily get the joint distribution of the asset price at two or more times correct. This means that exotic options such as compound options and barrier options may be priced incorrectly.

I can not understand that, IVF guarantees the model match the market price of vanilla option for all strike $K$ and all maturity $T.$ And the implied distribution of asset in future time is totally determined by the market price: $$p(S^*,t^*;K,T) = e^{r(T - t^*)}\dfrac{\partial^2 V}{\partial K^2}.$$

Here, market value: $V,$ maturity $T,$ strike $K,$ spot price of asset: $S^*,$ current time: $t^*.$

Can anyone give me a clear explanation?

• Could you please clarify what your question is. From you last few sentences, it seems like you want to know why an arbitrage free continuum of market prices of European options implies unique marginal distributions. However, the text that you cite is mostly about two models with the same marginal distributions having different transition densities and thus leading to different prices for path-dependent options. The latter is what @dm63 's answer addresses. Commented Oct 17, 2017 at 11:24
• @LocalVolatility, Could I understand as we know $p(S^*,t^*;K_1,T_1)$ and $p(S^*,t^*;K_2,T_2),$ at time time $t^*$ but still don't know $p(K_1,T_1;K_2,T_2)?$ Commented Oct 17, 2017 at 12:01
• As explained in @dm63's answer: Let $X_1$ and $X_2$ denote two random variables of known distributions. Can you infer the distribution of $X_1/X_2$ from the available info ? No because you would require the joint pdf while you only have the marginals. This is exactly the same here, you know the distribution of the returns $X_1 = \ln(S_{T_1}/S_0)$ and $X_2 = \ln(S_{T_2}/S_0)$, but you cannot infer that of the forward return $X_2/X_1 = \ln(S_{T_2}/S_{T_1})$. Commented Oct 17, 2017 at 14:54
• For your understanding of obtainign the pdf from the option prices, see this answer.
– will
Commented Oct 17, 2017 at 16:59