3
$\begingroup$

SDE 1 is S1 = S10 exp( (r1-sigma^2/2) * dt + sigma dW1 )

S2 = S20 exp( (r2-sigma2^2/2) * dt + sigma2 dW2 )

E[dW1 dW2] = rho

I want to price an option on S1 x S2 I know I need to use the SDE's to find the SDE for d(S1 S2) using Ito...but what if I use this

S = S1 x S2 using two analytical solutions of SDE? Why the drift term comes out incorrectly? Why can't I multiply two SDE solutions to get an equation for the third....?

|improve this question
$\endgroup$
  • 2
    $\begingroup$ Please use latex for formulas. $\endgroup$ – Ric Dec 3 '14 at 8:22
  • $\begingroup$ The reason it doesn't work is because you need the Ito correction term. In a nutshell: do a 2dim taylor expansion and keep the $dw^2 =dt$ terms. $\endgroup$ – Kian Dec 4 '14 at 11:56
  • $\begingroup$ You say that the drift term is incorrect - what is your result of the term. It is unclear what you are asking. Please use latex and ask a clear question. Of course it is is the same whether you multiply 2 solutions or you write down a correct (!) SDE for the product and solve it afterwards. $\endgroup$ – Ric Jan 5 '15 at 14:18
1
$\begingroup$

You can use Ito's product rule $d(X \, Y) = dX \, Y + X \, dY + dX \, dY$. In your case, you have $$ dS_{1,t} = S_{1,t} \left( r_1 dt + \sigma_1 dW_{1,t} \right) $$ and $$ dS_{2,t} = S_{2,t} \left( r_2 dt + \sigma_1 dW_{2,t} \right) $$

|improve this answer
$\endgroup$
0
$\begingroup$

By all means, if $dS_i = \mu(S^i_t)\mathrm dt + \sigma(S^i_t)\mathrm dW^i_t$ for $i=1,2$ and you know explicit formulas for $S_i$ then their product satisfies the SDE you derive using the Ito lemma for a product. Could you elaborate, on what is the issue you have with the drift term?

|improve this answer
$\endgroup$
  • 1
    $\begingroup$ The brownian motions are correlated ($\rho$), so $S\neq S_1S_2$. $\endgroup$ – emcor Dec 3 '14 at 11:11
  • 1
    $\begingroup$ @emcor: can you be more precise, $S = S_1\cdot S_2$ by definition of $S$ $\endgroup$ – Ulysses Dec 3 '14 at 13:56
  • 2
    $\begingroup$ Their joint distribution is different, therefore the solutions cannot be multiplied as individually. $\endgroup$ – emcor Dec 3 '14 at 14:11
  • $\begingroup$ Of course they can be multiplied! One just has to take care for all probabilistic aspects but $S=S_1* S_2$ always holds. one just has to carfully analyse the distribution of $S$ and/or its SDE. $\endgroup$ – Ric Jan 5 '15 at 14:19
  • $\begingroup$ The covariation term $[ X, Y]$ will handle the correlated brownians, giving you $\rho \sigma_1 \sigma_2 dt$ $\endgroup$ – Drew Jan 5 '15 at 14:51

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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