Extensions of Breiman's theorem of product of dependent random variables with applications to ruin theory (Q2417991)

Let \(X\) and \(\Theta\) denote two (non-negative) random variables, which are defined on the same probability space \((\Omega, \mathcal{F}, \mathbb{P})\). Under the assumption of stochastic independence of \(X\) and \(\Theta\), [\textit{L. Breiman}, Teor. Veroyatn. Primen. 10, 351--360 (1965; Zbl 0147.37004)], [\textit{D. B. H. Cline} and \textit{G. Samorodnitsky}, Stochastic Processes Appl. 49, No. 1, 75--98 (1994; Zbl 0799.60015)], as well as [\textit{D. Denisov} and \textit{B. Zwart}, J. Appl. Probab. 44, No. 4, 1031--1046 (2007; Zbl 1141.60041)] proved that \[ \lim_{x \to \infty} \frac{\mathbb{P}(X \Theta > x)}{\mathbb{P}(X > x)} = \mathbb{E}\left[\Theta^\alpha\right], \] provided that \(\mathbb{E}\left[\Theta^{\alpha+\varepsilon}\right] < \infty\) for some \(\varepsilon > 0\) or \(\varepsilon \geq 0\), respectively, and that the distribution of \(X\) has a regularly varying tail of index \(-\alpha\), where \(\alpha > 0\). This result is referred to as Breiman's theorem in the literature. In the present article, the authors extend Breiman's theorem to the case that the dependency structure of \((X, \Theta)\) can be described by an absolutely continuous copula function fulfilling certain assumptions. Specific copula families such as the family of Farlie-Gumbel-Morgenstern copulae are investigated in detail. Furthermore, the authors discuss applications to risk theory, and they present some numerical results from this field.