A note on the net profit condition for discrete and classical risk models (Q904327)

From MaRDI portal
scientific article
Language Label Description Also known as
English
A note on the net profit condition for discrete and classical risk models
scientific article

    Statements

    A note on the net profit condition for discrete and classical risk models (English)
    0 references
    0 references
    0 references
    13 January 2016
    0 references
    From the theory of random walks one knows that a random walk \(\sum_{k=1}^n Y_k\) crosses any large level if \(E[Y_k] \geq 0\), and \(P[\sup_n \sum_{k=1}^n Y_k \leq 0] > 0\) if \(E[Y_k] < 0\). Considering the discrete risk model \(U(n) = \sum_{k=1}^n (Z_k-1)\), this implies that ruin occurs almost surely if \(E[Z_k] \geq 1\). Here \(\{Z_k\}\) are iid random variables with values in \(\{0,1,2,\ldots\}\). Define the ruin probability \(\psi(u) = P[\sup_n U(n) > u]\). In this paper, the aim is to prove \(\psi(u) = 1\) if \(E[Z] \geq 1\) by simple arguments. One further aim is to get \(\psi(0)\) in the case \(E[Z] < 1\), Define \(\varphi(u) = 1 - \psi(u)\). As in the literature on the discrete model, one starts with \[ \varphi(u) = \sum_{k=0}^{u+1} P[Z = k] \varphi(u + 1 - k)\,, \] adds these equations for \(u=0\) to \(n\) and interchanges the sums. Since \(\lim_{u \to \infty} \varphi(u)\) exists, one can take the limit on both sides yielding an expression for \(\varphi(0)\). In the literature, one uses here that \(\lim_{u \to \infty} \varphi(u) = 1\). In the present paper, the only possible limit in the case \(E[Z] > 1\) is zero. In the case \(E[Z] = 1\), one gets that either \(P[Z = 1] = 1\) (in which case \(\varphi(u) = 1\) for all \(u\)) or \(\varphi(0) = 0\). It is then verified that \(\varphi(u) = 0\) for all \(u\). Finally, using the strong law of large number, \(\lim_{u \to \infty} \varphi(u) = 1\) is shown in the case \(E[Z] < 1\). In a last section, the classical compound Poisson risk model is considered. That is, \(U(t) = \sum_{k=1}^{N_t} Z_k - t\), where \(\{N_t\}\) is a Poisson process (with rate \(\lambda\), say), and \(\{Z_k\}\) is an iid sequence of positive random variables independent of \(\{N_t\}\). Considering the model at times \(n/M\) for some integer valued \(M \geq 1\) and discretising the claim size distribution, the model can be compared to the discrete model. One obtains \(\psi(u) = P[\sup_t U(t) > u]\) is equal to one if \(\lambda E[Z] \geq 1\) and \(\psi(0) = \lambda E[Z]\), if \(\lambda E[Z] < 1\).
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    net profit condition
    0 references
    Silverman-Toeplitz theorem
    0 references
    ruin probability
    0 references
    discrete-time risk model
    0 references
    classical risk model
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references