3
$\begingroup$

Working through an exercise in interest rate modelling and I have the following setup: $r_t = r_0 + \delta N_t$ where $\delta > 0$ and $\lambda > 0$ is the intensity of the Poisson pricess $N_t$. I want to calculate the bond price at time $0$, which is given by the formula $$ B(0,T) = \mathbb{E}_{\mathbb{P}^*}\left( e^{-\int_0^T r_u \,du}\right) = e^{-r_0T}\mathbb{E}_{\mathbb{P}^*}\left( e^{-\delta \int_0^T N_u\,du}\right). $$ where $\mathbb{P}^*$ is the risk-neutral probability measure. I haven't been dealing with Poisson processes lately and am not sure if there is a simple transformation or use of MGFs to get the second integral/expectation using properties of the Poisson process. I have tried to find some resources online but to no avail! Any help is appreciated. Thanks!

$\endgroup$
1
  • $\begingroup$ Stochastic Fubini theorem might help? $\endgroup$
    – Quantuple
    Jul 7, 2016 at 7:44

1 Answer 1

4
$\begingroup$

Let $$Y_t = \int_0^t N_u du$$

where $(N_t)_{t \geq 0}$ figures a Poisson process with intensity $\lambda$.

Using the stochastic Fubini theorem we have that: \begin{align} Y_T &= \int_0^T N_t dt \\ &= \int_0^T \int_0^t dN_u dt \\ &\color{lightgray}{= \int_0^T \int_0^T \mathbf{1}\{u \in [0,t]\} dN_u\ dt} \\ &\color{lightgray}{= \int_0^T \int_0^T \mathbf{1}\{u \in [0,t]\} dt\ dN_u} \\ &\color{lightgray}{= \int_0^T \int_0^T \mathbf{1}\{t \in [u,T]\} dt dN_u} \\ &= \int_0^T \int_u^T dt dN_u \\ &= \int_0^T (T-u) dN_u \\ &= \sum_{i=1}^{N_T} (T-T_i) \tag{0} \end{align} where $T_i$ figures the $i^{th}$ Poisson process jump time, which is Gamma distributed with integer parameter $i$ (also known as Erlang distribution see here Proposition 15.2 for instance). The jump times $T_i$ are therefore not unconditionally i.i.d. and $Y_t$ cannot be seen as a usual compound Poisson process.

But, conditioning helps! Indeed, conditionally on the number of jumps $N_T = n$ one can show that the $n$ jump times $\{T_i\}_{i=1}^n$ are actually independently and identically distributed uniformly over $[0,T]$ (see reference here top of page 443), that is \begin{align} T_i\ \vert\ N_T=n \ \ &\sim\ \ T U_i\\ \{U_i\}_{i=1}^n \ &\sim\ \ \mathcal{U}[0,1] \ \ \text{i.i.d.} \end{align}

We then have: \begin{align} \mathbb{E} \left[e^{-\delta Y_T} \vert \mathcal{F}_0 \right] &= \mathbb{E} \left[ \left. \exp\left(-\delta \sum_{i=1}^{N_T} (T-T_i)\right) \right\vert \mathcal{F}_0 \right] \tag{1} \\ &= \mathbb{E} \left[ \left. \mathbb{E} \left[ \left. \exp\left(-\delta \sum_{i=1}^{N_T} (T-T_i)\right) \right\vert \mathcal{F}_T \right] \right\vert \mathcal{F}_0 \right] \tag{2} \\ &= \mathbb{E} \left[ \left. \mathbb{E} \left[ \left. \exp\left(-\delta T \sum_{i=1}^{N_T} (1-U_i) \right) \right\vert \mathcal{F}_T \right] \right\vert \mathcal{F}_0 \right] \tag{3} \\ &= \sum_{n=0}^\infty \mathbb{E} \left[ \exp\left(-\delta T \sum_{i=1}^{n} \tilde{U}_i \right) \right] \mathbb{P}^*\left( N_T = n \right) \tag{4}\\ &= e^{-\lambda T} \sum_{n=0}^\infty \frac{(\lambda T)^n}{n!} \mathbb{E} \left[ \exp\left(-\delta T \sum_{i=1}^{n} \tilde{U}_i \right) \right] \tag{5} \\ &= e^{-\lambda T} \sum_{n=0}^\infty \frac{(\lambda T)^n}{n!} \prod_{i=1}^n \mathbb{E} \left[ \exp\left(-\delta T \tilde{U}_i \right) \right] \tag{6} \\ &= e^{-\lambda T} \sum_{n=0}^\infty \frac{(\lambda T)^n}{n!} \left(\mathbb{E} \left[ \exp\left(-\delta T \tilde{U}_1 \right) \right]\right)^n \tag{7} \\ &= \exp(-\lambda T) \exp\left(\lambda T \mathbb{E} \left[ \exp\left(-\delta T \tilde{U}_1 \right) \right] \right) \tag{8} \\ &= \exp\left(\lambda T \left(\mathbb{E} \left[ \exp\left(-\delta T \tilde{U}_1 \right) \right] - 1 \right) \right) \tag{9} \end{align} where $\tilde{U}_i = 1-U_i$ are also $\mathcal{U}[0,1]$ i.i.d. by construction and we used:

  1. Def. of $Y_T \vert \mathcal{F}_0 = \int_0^T N_u du = \sum_{i=1}^{N_T} (T-T_i)$ see $(0)$
  2. Tower property of conditional expectation
  3. Conditionally on $N_T$ the Poisson jump times are i.i.d. $\mathcal{U}[0,T]$
  4. Expectation over $N_T \vert \mathcal{F}_0$
  5. Def. of Poisson Process $(N_t)_{t \geq 0}$
  6. Properties of $\exp(.)$ + $\{\tilde{U}_i\}_{i=1}^n$ independent
  7. $\{\tilde{U}_i\}_{i=1}^n$ identically distributed
  8. Series expansion of $\exp(.)$
  9. Properties of $\exp(.)$

Now all you need to do is compute $Z = \mathbb{E}[\exp\left(-\delta T U \right)]$ with $U \sim \mathcal{U}[0,1]$:

\begin{align} Z &= \mathbb{E}[\exp\left(-\delta T U \right)] \\ &= \int_{0}^1 \exp\left(-\delta T u \right) \underbrace{p^*(u)}_{\mathbf{1}\{u \in [0,1]\}} du \\ &= \left[ \frac{\exp\left(-\delta T u \right)}{-\delta T} \right]_0^1 \\ &= \frac{1 - e^{-\delta T}}{\delta T} \end{align}

You will then be able to write: $$ e^{-r_0T}\mathbb{E}^{\mathbb{P}^*}\left[ e^{-\delta \int_0^T N_u\,du} \vert \mathcal{F}_0 \right] = \exp\left(-r_0T + \lambda T\left(\frac{1 - e^{-\delta T}}{\delta T}-1\right)\right) $$

$\endgroup$
4
  • $\begingroup$ Wow thanks so much for that long type up! I suspected stochastic Fubini had to be used but wasn't sure of the details. Conditioning for Poisson (discrete) / Exponential (continuous) always seem to be a good approach for these type of things don't they. Thanks again!! $\endgroup$
    – user89635
    Jul 7, 2016 at 12:01
  • $\begingroup$ You're welcome and yes indeed :) I am glad that it helped. I suspect there might be a less hand wavy way to do it though. $\endgroup$
    – Quantuple
    Jul 7, 2016 at 12:09
  • $\begingroup$ Just a quick one - the last expression does not need the expectation but all good apart from that. Cheers $\endgroup$
    – user89635
    Jul 8, 2016 at 12:07
  • $\begingroup$ @user89635 You are absolutely right, thanks for proofreading. $\endgroup$
    – Quantuple
    Jul 8, 2016 at 12:17

Your Answer

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

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