The Gerber-Shiu expected discounted penalty function for Lévy insurance risk processes (Q601942)

A Lévy risk model \(\{X_t\}\) without a Brownian component and with \(\mathbb{E}[X_1] > 0\) is considered, where the upward jumps are bounded by some constant \(-a \geq 0\). The goal is to calculate the Gerber--Shiu expected discounted penalty function \[ \Phi(x) = \mathbb{E}\bigl[ e^{-\delta T} 1_{T < \infty} w(X_{T-},|X_T|) \bigm| X_0 = x\bigr]\;, \] where \(\delta \geq 0\) is a discounting factor, \(w(\cdot,\cdot)\) is a bounded function and \(T = \inf\{t>0: X_t < 0\}\) is the time of ruin. First, a compound Poisson model is discussed. By the usual arguments, it is shown that \(\Phi\) solves a defective renewal equation. Thus there exists a Pollaczek--Khintchine formula expressing \(\Phi\). Approximating the Lévy-model by neglecting the jumps with absolute value smaller than \(n^{-1}\), the Pollaczek--Khintchine formula for the Lévy model is obtained. As special cases, the ruin probability, and the distribution of the surplus before and the deficit at ruin is discussed. The presentation is based on the unusual Lévy-measure \(q\) defined through \[ \Lambda(\alpha) = -\log \mathbb{E}[e^{-\alpha (x + c -X_1)}] = \int_a^{\infty} [1 - e^{-\alpha x}- \alpha x 1_{|x|\geq 1}] q(d x)\;. \] Here \(c\) is the premium rate and \(x+c t-X_t\) is the jump process. Thanks to the assumption \(E[|X_1|] < \infty\), one can in fact consider the measure \(q\). However, the density of the claim size distribution in the compound Poisson model is different from \(q(d x)/(\int_a^\infty q(d y))\). If the reader defines the Lévy-measure \(q\) through \[ \Lambda(\alpha) = \int_a^{\infty} [1-e^{-\alpha x}] q(d x) \] instead, the problem disappears. The approach only works in the case of a compound Poisson process, which already is discussed in \textit{C. Labbé} and \textit{K. P. Sendova} [Appl. Math. Comput. 215, No. 5, 1852--1867 (2009; Zbl 1181.91100)]. The expressions obtained in the paper contain the integral \(\int_a^\infty \int_t^\infty q(d y) d t\). In the case of infinitely many jumps in a finite time interval, \(\int_t^\infty q(d y) = \infty\) for \(t \leq 0\).