Say I have an asset following arithmetic Brownian motion $$ dX(t) = \sigma dW^\bot (t) $$ with $\sigma$ constant, and I have prices of vanilla options on $X$.

I introduce a Brownian bridge $$ dY(t) = \nu dW(t) - \frac{\nu}{T} W(T) dt $$ wih $\nu$ is constant, and I also have the orthogonality condition $$ dW(t)dW^\bot(t) = 0 $$

Since $Y(T) =0$ I will have at all times $t$ $$ E_t (X_T + Y_T - K)_+ = E_t(X_T - K)_+ $$

But how do I prove this explicitly?


To be clearer, what I mean with prove explicitly is to start with the observation that since $$ \sigma dW^\bot (t) + \nu dW(t) = \sqrt{\sigma^2 + \nu^2} dZ(t) $$ The expectation can be written as $$ E_t(X_T + Y_T - K)_+ = E_t \left( \sqrt{\sigma^2 + \nu^2}Z(T) - \nu W(T) - K \right)_+ $$ Is there a way to evaluate the above expectation on the right hand side, keeping the volatility $\sqrt{\sigma^2 + \nu^2}$ and still satisfy the fact that it must equal the vanilla option price at time $t$?

If $W(T)$ were uncorrelated to $Z(T)$ then I could apply conditioning and evaluate it easily, but unless I am missing something I don't think it's possible to evaluate the above expectation without the $\nu$ dropping out again because $W(T)$ is not orthogonal to $Z(T)$, right?

Basically what I want to do is have an option, which is still delta-hedgeable with $X$ only, and which has as volatility $\sqrt{\sigma^2 + \nu^2}$, where $\nu$ is the historical volatility of $X$ and $\sigma$ is the future volatility of $X$. Vanilla options only contain the future volatility, so I thought this trick with Browian bridge might work, but not sure.

Any other ideas on how to achieve what I want with or without Brownian bridge would be welcome!

  • 1
    $\begingroup$ I am not sure I understand your question. What is it that you want to prove explicitly? $\endgroup$ Dec 1, 2019 at 22:59
  • $\begingroup$ @DaneelOlivaw thanks, I edited my question. $\endgroup$
    – user34971
    Dec 2, 2019 at 6:02
  • $\begingroup$ Haven't been able to come up with anything yet. But if you don't mind me asking, is this somehow related to SOFR rates? $\endgroup$ Dec 4, 2019 at 11:10
  • 1
    $\begingroup$ @DaneelOlivaw No, not related to SOFR. Related to volatility swaps actually. If interested see attached paper in this thread: forum.wilmott.com/viewtopic.php?f=4&t=102154 $\endgroup$
    – user34971
    Dec 4, 2019 at 15:50
  • $\begingroup$ Interesting stuff, thanks. I had never heard about timer options. I though it might be related to pricing options on SOFR-based compound rates, in which a usual problem is that the final compounded rate is not fully known until expiry. $\endgroup$ Dec 4, 2019 at 16:15

1 Answer 1


To prove that $Y(T)=0$, you integrate the SDE :

$$\int_{0}^{T}dY(t) = \int_{0}^{T}\nu dW(t) - \frac{\nu}{T} \int_{0}^{T}W(T) dt$$

$$Y(T)-Y(0)=\nu\left( W(T)-W(0)\right)- \frac{\nu}{T}W(T)\left(T-0\right)$$

Therefore $$Y(T)=Y(0)$$

You forgot to write the initial conditions but I guess $Y(0)=0.$


The dependency in $\nu$ is irrelevant in pricing that option as $Y_T=0$. The second term is just a function of $\sigma$.

If you fix the price, then you cannot fix $\sigma$( Price/Bachelier is a bijection). Somehow you decided to introduce the Brownian bridge, and what you wrote is correct. However, $Z(T)$ and $W(T)$ are correlated, $$cov( \sqrt{\sigma^2 + \nu^2}Z(T) ,\nu W(T))=\nu^2 T$$, and if you derive the distribution of $\sqrt{\sigma^2 + \nu^2}Z(T) -\nu W(T)$. the $\nu$ just disappears. At then end of the day, your model is a Bachelier model. Your motivation is still unclear to me. The option price will give you the future volatility. The historical volatility gives you the past dynamics... which is irrelevant in pricing the option. Are you trying to decompose the implied volatility into a historical vol and another component?

  • $\begingroup$ I think you typos where you interchange T and t in your first equation? $\endgroup$
    – Sanjay
    Dec 1, 2019 at 17:15
  • $\begingroup$ I interchange them ? $\endgroup$
    – Canardini
    Dec 1, 2019 at 17:20
  • $\begingroup$ @Canardini Thanks, I up voted your answer, but I didn't give it a green check yet as it was not entirely what I was asking, I was not being clear, sorry. $\endgroup$
    – user34971
    Dec 2, 2019 at 6:14
  • $\begingroup$ Check my edit, and clarify please $\endgroup$
    – Canardini
    Dec 6, 2019 at 16:01
  • $\begingroup$ See the comment I wrote to Daneel Olivaw and the link to the thread on wilmott. $\endgroup$
    – user34971
    Dec 8, 2019 at 19:35

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.