0
$\begingroup$

I am trying to price a path dependent option which uses two underlyings (a stock index and an interest rate index). I am using Hull White model for interest rate modelling and local vol for stock index modelling. \begin{eqnarray} dr(t) &=& (\theta(t) - kr(t))dt + \sigma_r(t)dw_r(t)\\ \end{eqnarray}

\begin{eqnarray} ds(t) &=& r'(t)sdt + \sigma_s(t,s)sdw_s(t)\\ \end{eqnarray} Further, assume that $<dw_s, dw_r> = \rho dt$ and that $r'$ is deterministic. I need the samples for $s$ and $r$(and stochastic discount factors) only at certain times $t_1, t_2...t_n$. In case we did not have correlated processes, I could have calculated mean and variance of $r$ and discount factors analytically (and then creating the corresponding samples). Can we still solve this by calculating the mean and variance for $r$ and discount factors analytically and simulate $s_t$ numerically?

$\endgroup$

1 Answer 1

0
$\begingroup$

Yes, in the case of path-dependent option with two underlying assets, you can approach the problem by calculating the mean and variance for the interest rate $r(t)$ and corresponding discount factor analytically, while simulating the stock price $S(t)$ numerically. This will still work with correlated process since we have the correlation coefficient between the Brownian motions.

You can solve the Hull-White SDE analytically to find the mean and variance of the interest rate at future times $t_1,t_2,...,t_n$, then simulate the interest rate path.

For the local volatility model for the stock price, $r'(t)$ is deterministic, and $\sigma_s(t,s)$ is the local volatility model which is dependent on time and the stock price, you can typically simulate this process numerically. You can look into Euler–Maruyama method.

Then when simulating the paths for $S(t)$ and $r(t)$ just make sure you account for the correlation. Using the simulated paths for $r(t)$, you can calculate the discount factor for the respective time $t_1,t_2,...,t_n$.

Now that you have the analytical expression for the interest rate and the discount factor, and the numerical simulation setup for the stock price, you can generate the paths and compute the option's payoff at each point in time. Then you just discount these payoffs back to the present value.

$\endgroup$
2
  • $\begingroup$ Thanks but it seems like this is equivalent to not using the analytic solution of Hull White model at all. We are basically doing the simulation from scratch. I was hoping that having an analytical solution for one equation will help speed up the solution/reduce discretization/reduce simulation error... $\endgroup$
    – Madhuresh
    Commented Dec 9, 2023 at 3:50
  • $\begingroup$ @Madhuresh If you read this answer carefully (just the first sentence) you should realize that it does not seem we are "not using the analytic solution of Hull White model at all." $\endgroup$
    – Kurt G.
    Commented Dec 9, 2023 at 12:16

Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.