3
$\begingroup$

The jump diffusion model is defined as $$dS_t = \mu S_t dt + \sigma S_t dW_t + S_t d \left(\sum^{N_t}_{i=1}(V_i - 1)\right)\;\;\;\;\;\;\;(1)$$ , where ${V_i}$ is a sequence of iid non-negative random variables and it is independent of $W_t$. In the Merton's jump diffusion model , $log(V) \sim N(\mu_J, \sigma^2_J)$ and $ N_t$ is a poisson process with rate $\lambda$.

I was asked to apply Ito lemma to $(d \;logS_t)$ to obtain the following:

$$S_t = S_0 exp \left( \left(\mu - \frac{\sigma^2}{2} \right)t + \sigma W_t\right) \prod^{N_t}_{j=1}V_j \;\;\;\;\;\;\;(2)$$

I literally do not know how to solve this problem because of the term $d \left(\sum^{N_t}_{i=1}(V_i - 1)\right)$. This is how far I got to:

$$dlogS_t = \frac{1}{S_t} \left( \mu S_t dt + \sigma S_t dW_t + S_t d \left(\sum^{N_t}_{i=1}(V_i - 1)\right)\right) - \frac{1}{2S_t^2} d[S,S]_t$$

What exactly is $d[S,S]_t$ ? I know that $$d[S,S]_t = \sigma^2 S^2_t dt + ...$$ But what is that "..." ?

$\endgroup$
  • 1
    $\begingroup$ Well, apply Itô's lemma to $d(\log S)$... What did you find for $d\langle S\rangle$, already? $\endgroup$ – Did Feb 11 '15 at 8:52
  • $\begingroup$ That's the part I have trouble with. I have no idea how to solve $d<S>$ ? Applying ito lemma to $d(log S)$ is straightforward. $\endgroup$ – mynameisJEFF Feb 11 '15 at 8:56
  • $\begingroup$ Which parts of $dS$ enter into $d\langle S\rangle$? Not many... (As an aside, note that if your trouble is that you cannot compute $d\langle S\rangle$, this is what your question should mention.) $\endgroup$ – Did Feb 11 '15 at 8:57
  • $\begingroup$ Sorry. I still cannot figure out the answer. Can you write an answer please ? $\endgroup$ – mynameisJEFF Feb 11 '15 at 9:12
  • $\begingroup$ @Did, please see update of my question $\endgroup$ – mynameisJEFF Feb 11 '15 at 9:19
3
$\begingroup$

I suspect that your SDE defining the $S_t$ process is problematic in the term "$S_t\text d\Big(\sum\limits_{i=1}^{N_t}(V_i-1)\Big)$". Intuitively, this term should capture instantaneous random jumps when they occur in any given realisation of the process. That is, if a jump occurs at time $t$, then the process jumps from $S_{t-}$ to $S_{t-}V_t$, where $S_{t-}$ is the left limit of the process at time $t$. So, it makes more sense for $S_t$ to be the jump-diffusion Levy process (so right-continuous, left-limit) defined by $$ \text dS_t = \mu S_{t-}\text dt + \sigma S_{t-}\text dW_t + \text d\Big(\sum\limits_{0\leqslant r\leqslant t}S_{r-}(V_r - 1){\bf 1}_{\{N_r-N_{r-}=1\}}\Big),\tag{1} $$

with jumps determined by a Compound Poisson process independent of $\{W_t\}_{t>0}$. That is, a Poisson process $\{N_t\}_{t>0}$ determines when jumps occur, and the size of such a jump at time $t$ is determined as $S_{t-}(V_t - 1)$ (where $\log V_t\sim\mathcal N(\mu_{J},\sigma^2_{J})\ $ for i.i.d. $V_t$).

Assuming $(1)$ for your SDE and $\mathbb E[\int_{0}^{T}S_{u-}^2\text du] < \infty$, Ito's lemma for jump-diffusion processes is applicable here. Note that, for a $\ C^{1,2}\ $ function $\ f:[0,T]\times\mathbb R\to\mathbb R$, the process $f(t,S_t)$ satisfies

$$ \begin{eqnarray*} f(t,S_t) - f(0,S_0) &=& \int_{0}^{t}(\partial_uf + \mu S_{u-}\partial_sf + \frac{1}{2}\sigma^2S_{u-}^2\partial_{ss}f)\text du \\ && + \sigma \int_{0}^{t}S_{u-}\partial_sf~\text dW_u + \sum_{0\leqslant r\leqslant t}(f(r,S_r)-f(r-,S_{r-})){\bf 1}_{\{N_{r}-N_{r-}=1\}}\,. \end{eqnarray*} $$

This is simply Ito's lemma for a diffusion process but, in addition, the final term adds the change in the process due to the jumps determined by the compound Poisson process. So, choosing $f(S_t):=\log S_t$ gives us

$$ \begin{eqnarray*} \log\Big(\frac{S_t}{S_0}\Big) &=& (\mu-\frac{1}{2}\sigma^2)t + \sigma W_t + \sum_{0\leqslant r\leqslant t}{\bf 1}_{\{N_{r}-N_{r-}=1\}}\log\Big(\frac{S_{r-}V_r}{S_{r-}}\Big) \\ &=&(\mu-\frac{1}{2}\sigma^2)t + \sigma W_t + \sum_{0\leqslant r\leqslant t}{\bf 1}_{\{N_{r}-N_{r-}=1\}}\log V_r\,. \end{eqnarray*} $$ Therefore, in your notation,

$$ S_t = S_0 e^{(\mu-\frac{1}{2}\sigma^2)t + \sigma W_t}\prod_{i=1}^{N_t}V_i\,. $$

$\endgroup$
  • $\begingroup$ I just want to ask in the part where you have $log(\frac{S_{r-} V_r}{S_{r-}})$, shouldn't it be $log(\frac{S_{r-} (V_r- 1)}{S_{r-}})$ ? $\endgroup$ – mynameisJEFF Mar 30 '15 at 10:54
  • $\begingroup$ Using $S_r = S_{r-}V_r$ when a jump occurs at time $r$, we have $$f(r,S_r)-f(r-,S_{r-}) = \log(S_r)-\log(S_{r-}) = \log\Big(\frac{S_r}{S_{r-}}\Big) = \log\Big(\frac{S_{r-}V_r}{S_{r-}}\Big)$$ $\endgroup$ – ki3i Mar 30 '15 at 11:53
  • $\begingroup$ Yes , I understand that part, as it is based on the ito doeblein formula for jump process. However, I don't understand why is the minus 1 missing ? $\endgroup$ – mynameisJEFF Mar 30 '15 at 11:55
  • $\begingroup$ @mynameisJEFF, If you understand that part, then why do you think there should be a "-1"? $\endgroup$ – ki3i Mar 30 '15 at 11:57
  • $\begingroup$ O. -1 is a constant so it doesn't constitute a jump. Am I right ? $\endgroup$ – mynameisJEFF Mar 30 '15 at 11:58
-1
$\begingroup$

In your case

the quadratic variation can be obtained by formally squaring the SDE $ d[S,S]_t = dS_t \cdot dS_t =\left( \mu S_t dt + \sigma S_t dW_t + S_t d \left(\sum^{N_t}_{i=1}(V_i - 1)\right) \right)^2 $ using the fact that $ (\mu S_t)^2 dt^2 = 0 $ and $ \mu S_t dt \cdot \sigma S_t dW_t =0 $ we get

$ d[S,S]_t = \sigma^2 S_t^2 dt + 2 [ d\left(\sum^{N_t}_{i=1}(V_i - 1)\right), \mu S_t dt ] + 2 [ d\left(\sum^{N_t}_{i=1}(V_i - 1)\right), dW_t ] + [ d\left(\sum^{N_t}_{i=1}(V_i - 1)\right), d\left(\sum^{N_t}_{i=1}(V_i - 1)\right) ]_t $.

Since $V_i $ are i.i.d. $ [dV_i, dt ]=0$ and $ [dV_i, dW_t ]=0$ and $ [dV_i, dV_j ]=\delta_{i\,j}$.
For the last term we also have $[dV_i , dV_i ] = dV_i$. Then finally

$ dlogS_t = \frac{1}{S_t} \left( \mu S_t dt + \sigma S_t dW_t + S_t d \left(\sum^{N_t}_{i=1}(V_i - 1)\right)\right) - \frac{1}{2S_t^2} \sigma^2 S^2_t dt = $

$ dlogS_t = \left( \mu - \frac{1}{2 } \sigma^2 \right) dt + \sigma \; dW_t + d\left(\sum^{N_t}_{i=1}(V_i - 1)\right) $.

$\endgroup$
  • 1
    $\begingroup$ $[dV_i-1, dV_i-1]=0$ ... why is that? $\endgroup$ – saz Mar 23 '15 at 7:01
  • $\begingroup$ A Poisson process has quadratic variation equal to itself. This is because it is a pure jump process with jump sizes equal to 1. If $V_t$ is the Poisson process then the only contribution to the quadratic variation $[V,V]$ comes from the jumps. $\endgroup$ – user48672 Mar 23 '15 at 22:28
  • $\begingroup$ my mistake. I overlooked. $\endgroup$ – user48672 Mar 23 '15 at 22:30
  • $\begingroup$ Is this answer wrong? $\endgroup$ – user357269 Dec 11 '16 at 12:27

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.