Tail behavior of Poisson shot noise processes under heavy-tailed shocks and actuarial applications (Q370897)

Let \(X_1\), \(X_2,\dots\) be nonnegative identically distributed and exchangeable random variables such that \(\operatorname{P}(X_1>x)\) is regularly varying at \(\infty\) with a negative index and \(\lim_{x\to\infty}\operatorname{P}(X_i>x, X_j>x)/\operatorname{P}(X_1>x)=0\) for \(i,j\geq 1\), \(i\neq j\). Further, let \(\tau_1\), \(\tau_2,\dots\) be the arrival times of a (non-homogeneous) Poisson process with positive intensity \(\lambda(t)\), and \(h(t,s)\) and \(g(s)\) nonnegative measurable functions. The authors provide sufficient conditions such that \[ \operatorname{P}\bigg(\sum_{k\geq 1}X_kh(t,\tau_k)1_{\{\tau_k\leq t\}}>x\bigg)\sim \int_0^t\operatorname{P}\big(h(t,s)X_1>x\big)\lambda(s)ds, \;\;x\to\infty, \] holds for each fixed \(t>0\) and sufficient conditions such that a uniform version of this relation holds. Also, they prove that under certain additional assumptions \[ \operatorname{P}\bigg(\sum_{k\geq 1}g(\tau_k)X_k>x\bigg)\sim \int_0^\infty\operatorname{P}\big(g(s)X_1>x\big)\lambda(s)ds, \;\;x\to\infty. \] Applications of the obtained results are given to the finite and infinite time ruin probabilities of a continuous time risk model and an approximation on the preium of the stop-loss insurance coverage.