Supermodular ordering and stochastic annuities (Q1302132)

From MaRDI portal
scientific article
Language Label Description Also known as
English
Supermodular ordering and stochastic annuities
scientific article

    Statements

    Supermodular ordering and stochastic annuities (English)
    0 references
    0 references
    0 references
    16 August 2000
    0 references
    The authors propose a method how to deal appropriately from an actuarial point of view with a risk \(V\) whose distribution function is unavailable. Their idea is to replace \(V\) by a `more risky' quantity \(W\) with known distribution where `more risky' means that \(V\) is dominated by \(W\) in stop-loss order, i.e., \(E(V-d)^+ \leq E(W-d)^+\) for all \(d\). The authors give an explicit description of such a dominating risk \(W\) for a primary risk \(V\) of the additive form \(V=\varphi_1(X_1)+\ldots+\varphi_n(X_n)\) where \(\varphi_i:\mathbb{R}\to\mathbb{R^+}\) is decreasing and \(X=(X_1,\ldots,X_n)\) is a random variable with known marginals \(P[X_i \leq x]=F_i(x) (x \in \mathbb{R})\). More precisely, they show that \(W\) may be chosen as \(W:=\varphi_1(Y_1)+\ldots+\varphi_n(Y_n)\) where \(Y=(Y_1,\ldots,Y_n):= (F^{-1}_1(U),\ldots,F^{-1}_n(U))\) with \(U\) uniformly distributed on \([0,1]\). To prove this result, the authors proceed by noting first that risks \(V\) and \(W\) of the above additive form are in stop-loss order if \(X\) is smaller than \(Y\) in the supermodular order; for a definition of this order see, e.g., \textit{A. Müller} [Insur. Math. Econ. 21, No. 3, 219-223 (1997; Zbl 0894.90022)]. By a result due to \textit{A. H. Tchen} [Ann. Probab. 8, 814-827 (1980; Zbl 0459.62010)], the largest random vector \(Y\) with given marginals \(F_i(.)\) is the comonotonic variable \((F^{-1}_1(U), \ldots, F^{-1}_n(U))\). The combination of these results then yields the assertion. The method is illustrated by considering different types of annuities as case studies. They start with \(V_1= \sum_{i=1}^n \alpha_i \exp(-\delta i X_i)\) for nonnegative constants \(\alpha_i\), \(\delta\), and \(X_i \sim N(0,1)\). Passing from this to the continuous time limit \(V_2 = \int_0^t \alpha(s) \exp(-\delta s - \sigma X(s))\) where \(X\) is, e.g., a standard Brownian motion, shows how one may treat continuously composed annuities. Finally, the authors consider the case where \(\alpha(.)\) is discounted using the CIR-model for the short rate. For the infinite horizon case \(t=+\infty\), this allows them to derive upper bounds for the moments of a perpetuity in the CIR-model. The comparison of these bounds with the bounds obtained by \textit{F. Delbaen} [Math. Finance 3, No. 2, 125-134 (1993; Zbl 0884.90023)] closes the paper.
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    stop-loss order
    0 references
    supermodular order
    0 references
    stochastic annuities
    0 references
    0 references
    0 references