3
$\begingroup$

Currently I am working on studying stochastic integrals of the form: $$Z_\infty = \int_0^\infty e^{-f(t)}\mathop{d}S_t$$

where $S_t$ is a Compound-Poisson process with Exponentially-distributed increments. We know that the integral converges so long as $f$ is continuous on $[0,\infty)$ and $\lim_{t\to\infty}f(t) \to \infty$ holds true. $Z_\infty$ effectively models a perpetuity where $f$ is our discount rate function, and $S_t$ is the cashflow aggregation process.

As far as I am aware, in the case where $f$ is linear, $Z_\infty$ is Gamma distributed. I am interested in knowing if there are further results for when the discounting is a power-law rate; $f(t) = \lambda t^\alpha $.

$\endgroup$
3

1 Answer 1

1
$\begingroup$

This is more of an extended comment trying to study this integral.

In the similar spirit as here we study the Laplace transform. As explained here we have the Lévy-Khintchine formula for compound Poisson processes for a process whose increments have distribution $\nu$:

$$E(exp(r\int_{0}^{T}f(t)dY_{t}))=exp(\lambda\int_{0}^{T}\int_{-\infty}^{\infty}(e^{yrf(t)}-1) \nu(dy) dt )$$

and so in the above case of $\nu(y)=e^{-\mu y}1_{y\geq 0}$ and $f(t)=exp(-t^{\alpha})$ we have $$exp(\lambda\int_{0}^{T}\int_{0}^{\infty}(e^{yr exp(-t^{\alpha})}-1) e^{-\mu y} dt ).$$

I tried in Mathematica to possibly get exact formulas for the above for general $a>1$ and $T=\infty$ but didn't get anything back. It might require (keyhole) contour methods to study. You could also try to directly invert its Laplace transform but that seems tricky.

$\endgroup$
3
  • $\begingroup$ Thank you for your answer. My knowledge of contour integration is fairly basic, but is worth looking further into. I was wondering if you know how one may numerically solve this problem and generate an approximation to the pdf that way? I'm not interested in simulating from the original process but would like for instance an inversion method or a potential series approximation. Numerical analysis is not my strong suite and so any references or code would be handy $\endgroup$
    – jam jelly
    Commented Mar 13, 2023 at 11:17
  • $\begingroup$ I see in this post mathoverflow.net/questions/373665/laplace-transform-inversion, they talk about in the comments about some numerical papers/methods for a complicated LT like yours. $\endgroup$ Commented Mar 13, 2023 at 17:20
  • $\begingroup$ i see in this article "Efficient computation of first passage times in Kou’s jump-diffusion model" they include some code on "Inversion of the Laplace transform". But it seems to be an entire area dedicated to this issue with many different techniques. Even there they suggest using contour methods as part of the numerics. $\endgroup$ Commented Mar 13, 2023 at 17:20

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .