The Cramér-Lundberg model with a fluctuating number of clients (Q6072261)

An insurer has a homogeneous portfolio of customers. At time zero there are \(n f_0\) customers. New customers underwrite contracts according to a Poisson process with rate \(n \lambda\). Every customer stays in the contract for random time \(\tau_i\). The arrival times and the sojourn times are independent. The sojourn times of the new customers have the same distribution. The sojourn times of the customers present at time \(0\) have the corresponding stationary (integrated tail) distribution. That is, the number of customers present in the system follow a M/G/\(\infty\) queueing system. The premium and the claims process of an individual customer follows the classical compound Poisson model with identical premium rate, claim arrival intensity, and claim size distribution. Exponential moments are assumed. The initial capital of the insurer is \(n u\). One is interested in the ruin probability within the finite horizon \(T\). A finite-dimensional large deviation principle as well as a sample-path large deviation principle is proved. In this way, the most likely path to ruin can be identified. The proofs rely on splitting the surplus process in the parts of the customers present at time \(0\) and customers underwriting later, properties of the Poisson process and large deviation techniques. In order to get the sample path properties, exponential tightness is proved. It is further discussed what effect on the ruin probability more or less customers have.