Jump diffusion processes and their applications in insurance and finance

Jiwook Jang*

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review

27 Citations (Scopus)

Abstract

For insurance risks, jump processes such as homogeneous/non-homogeneous compound Poisson processes and compound Cox processes have been used to model aggregate losses. If we consider the economic assumption of a positive interest to aggregate losses, Lévy processes have proven to be useful. Also in financial modelling, it has been observed that diffusion models are not robust enough to capture the appearance of jumps in underlying asset prices and interest rates. As a result, jump diffusion processes, which are, simply speaking, combinations of compound Poisson processes with Brownian motion, have gained popularity for modelling in insurance and finance. In this paper, considering a jump diffusion process, we obtain the explicit expression of the joint Laplace transform of the distribution of a jump diffusion process and its integrated process, assuming that jump size follows the mixture of two exponential distributions, which is a special case of phase-type distributions. Based on this Laplace transform, we derive the moments of the aggregate accumulated claim amounts of insurance risk. For a financial application, we concern non-defaultable zero-coupon bond pricing. We also provide several numerical examples for the moments of aggregate accumulated claims and default-free zero-coupon bond prices.

Original languageEnglish
Pages (from-to)62-70
Number of pages9
JournalInsurance: Mathematics and Economics
Volume41
Issue number1
DOIs
Publication statusPublished - Jul 2007
Externally publishedYes

Keywords

  • Aggregate accumulated claims
  • IE51
  • IM11
  • IM30
  • Joint Laplace transform
  • Jump diffusion processes
  • Non-defaultable zero-coupon bond

Fingerprint

Dive into the research topics of 'Jump diffusion processes and their applications in insurance and finance'. Together they form a unique fingerprint.

Cite this