Uniform asymptotic estimates for ruin probabilities of a multidimensional risk model with càdlàg returns and multivariate heavy tailed claims
Dimitrios G. Konstantinides, Charalampos D. Passalidis · 2024
Read Uniform asymptotic estimates for ruin probabilities of a multidimensional risk model with càdlàg returns and multivariate heavy tailed claims by Dimitrios G. Konstantinides, Charalampos D. Passalidis (2024) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2412-13073.txt.
Introduction
Dimitrios Konstantinides and Charalampos Passalidis estimate ruin risk for an insurer holding many lines at once. Their work appeared in twenty twenty-four. Claims arrive in vectors across the lines, returns on reserves jump over time, and the worst claims are heavy-tailed.
The model renews claims through one shared counting process. Claim vectors share a heavy-tailed law, depend weakly on each other, and mix dependence inside each vector. The authors derive uniform estimates for the chance that discounted claims enter rare danger sets within a set horizon. Ruin probabilities follow directly. Under a stricter tail class, the estimates hold across the whole time horizon at once. The results serve pricing and capital planning under extreme joint losses.
Catalog record
| CREATOR | Dimitrios G. Konstantinides, Charalampos D. Passalidis |
|---|---|
| YEAR | 2024 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 47 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2412-13073
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Uniform asymptotic estimates for ruin probabilities of a multidimensional risk model with càdlàg returns and multivariate heavy tailed claims Dimitrios G. Konstantinides, Charalampos D. Passalidis Abstract . We study a multidimensional renewal risk model, with common counting process and càdlàg returns. Considering that the claim vectors have common distribution from some multivariate distribution class with heavy tail, are mutually weakly dependent, and each one has arbitrarily dependent components, we obtain uniformly asymptotic estimations for the probability of entrance of discounted aggregate claims into a some rare sets, over a finite time horizon. Direct consequence of the claim behavior is the estimation of the ruin probability of the model in some ruin sets. Further, restricting the distribution class of the claim vectors in the multivariate regular variation, the estimations still hold uniformly over the whole time horizon. Keywords: Multi-dimensional risk model, Heavy-tailed random vectors, Interdependence, Uniformity, Ruin probability. Mathematics Subject Classification : Primary 62P05; Secondary 60G70. 1. Introduction We consider an insurer, who operates d -business lines simultaneously, with d\in{\mathbb{N}} , that follow a common, renewal, claim counting process. Because of the intense competition, the modern insurance companies invest the surplus either in risk free or in risky assets. In our model, we assume that the price process of the investment portfolio, is described by the stochastic process \{e^{\xi(t)}\,,\;t\geq 0\} , where \{\xi(t)\,,\;t\geq 0\} , represents a càdlàg stochastic process, with \xi(0)=0 , which is general enough, found in the literature of uni-variate or multivariate risk models, since it contains the constant interest rate, the zero interest rate or any Lévy process, see for example [ 12 ] , [ 30 ] , [ 31 ] , [ 14 ] , [ 17 ] , etc. Furthermore, we accept that the insurer receives premiums, whose density is depicted by the vector {\bf c}(t)=(c_{1}(t),\,\ldots,\,c_{d}(t)) , with t\geq 0 , and c_{i}(t) is the density function of premiums in i -th line of business, for i=1,\,\ldots,\,d . In what follows, the density of premiums is bounded, namely 0\leq c_{i}(t)\leq M_{i} , for some constants M_{i}\geq 0 and for any t\geq 0 and i=1,\,\ldots,\,d . Also we consider the total initial capital of the insurer is equal to x and it is allocated in the d lines of business according to weights l_{1},\,\ldots,\,l_{d}>0 , such that l_{1}+\cdots+l_{d}=1 . Finally, the i -th claim vector {\bf X}^{(i)}=(X_{1}^{(i)},\,\ldots,\,X_{d}^{(i)}) , can allow zero components, but not all of them, and it arrives at time moment T_{i} , with i\in{\mathbb{N}} and conventionally we put T_{0}=0 . The \{T_{i}\,,\;i\in{\mathbb{N}}\} constitute a renewal counting process, symbolically \{N(t)\,,\;t\geq 0\} , with finite renewal function \lambda(t)=\mathbf{E}[N(t)]=\sum_{i=1}^{\infty}\mathbf{P}[T_{i}\leq t] . The discount surplus process of the insurer, can be written through the relations \displaystyle\left(\begin{array}[]{c}U_{1}(t)\\ \vdots\\ U_{d}(t)\end{array}\right)=x\,\left(\begin{array}[]{c}l_{1}\\ \vdots\\ l_{d}\end{array}\right)+\left(\begin{array}[]{c}\int_{0-}^{t}e^{-\xi(s)}\,c_{1}(s)\,ds\\ \vdots\\ \int_{0-}^{t}e^{-\xi(s)}\,c_{d}(s)\,ds\end{array}\right)-\left(\begin{array}[]{c}\sum_{i=1}^{N(t)}X_{1}^{(i)}\,e^{-\xi(T_{i})}\\ \vdots\\ \sum_{i=1}^{N(t)}X_{d}^{(i)}\,e^{-\xi(T_{i})}\end{array}\right)\,, for any t\geq 0 . It seems that the dependence among the variables of model ( 1 ) has real impact on insurance practice. For this reason, the risk model ( 1 ) attracted the interest of many researchers, and especially in case d=2 . For bi-variate versions of ( 1 ) we refer to [ 38 ] , [ 16 ] , [ 27 ] , [ 20 ] , etc. For more dimensions, there only few references, see for example [ 15 ] , [ 19 ] , [ 25 ] , [ 26 ] , [ 39 ] , [ 4 ] , where is some of them the counting process is not necessarily common in the d lines of business. In [ 3 ] we find an extension of the time-dependent risk model from [ 25 ] , using the \{\xi(t)\,,\;t\geq 0\} , as a càdlàg process, instead of Lévy process. Although, all these multivariate models provide a wide framework of dependence modeling, under the presence of heavy tails, there are three assumptions on the claims that can be regarded as restrictive. Namely (1) The random vectors \{{\bf X}^{(i)}\,,\;i\in{\mathbb{N}}\} , which represent claims, belong to the class of multivariate regular variation, symbolically MRV . In spite of the importance of the MRV class, in the frame of the the heavy-tailed distribution classes, it excludes several important distributions from the actuarial practice. (2) Each claim-vector {\bf X}^{(i)} , has asymptotically dependent components. In this case, from one side it is restrictive to dependence, and from the other side it implies, in combination of the MRV , that the components of {\bf X}^{(i)} follow distributions with tails that are strongly equivalent each other, which is clearly restrictive, especially for large values of d . (3) The sequence of \{{\bf X}^{(i)}\,,\;i\in{\mathbb{N}}\} , contains independent and identically distributed terms. In this work we attempt to face the points (1) - (3) . Most of the previous papers are focused on the points (1) and (2) . Of course, in order to avoid the points (1) - (3) we sacrifice the generality of the set family over the whole positive quadrant, except the point of origin, to some set from family \mathscr{R} , defined in Section 2 below. Concretely, we consider that the random vectors \{{\bf X}^{(i)}\,,\;i\in{\mathbb{N}}\} are identically distributed, with multivariate distribution from class (\mathcal{D}\cap\mathcal{L})_{A} , larger that MRV , and we permit arbitrary dependence among the components of each vector {\bf X}^{(i)} . Further, we assume for the family of vectors \{{\bf X}^{(i)}\,,\;i\in{\mathbb{N}}\} is subject to a weak dependence structure, which contains the independence as special case. This way, we obtain the interdependence among the claims, which means that each vector has dependent components (and here even arbitrary), and at the same time any two vectors are still under some dependence structure. The restrictions in sets from the family \mathscr{R} , found in the seminal paper [ 34 ] , can help to face the points (1) and (2) , by defining the corresponding multivariate subexponentiality. Indeed, it was examined a multivariate renewal risk model, without interest rate, with independent, multivariate, subexponential claims. Next, in [ 22 ] the asymptotic behavior of the total discounted claims, in a Poisson risk model with multivariate subexponential claims was examined, under similar conditions, with the present ones, for the the price process of the investment portfolio. In last paper the time horizon is finite. However, in the two last works the point (3) was not faced and furthermore, the asymptotic estimations are not uniform. In this paper, we restrict the multivariate claim distribution class from the multivariate subexponentiality \mathcal{S}_{A} , into the multivariate dominatedly varying, long tailed (\mathcal{D}\cap\mathcal{L})_{A} , in order to face point (3) , while the same time we establish uniformity in the asymptotic expressions. The structure of the paper is as follows. In Section 2, we provide necessary preliminary statements for the distribution classes of multivariate heavy tailed distributions. In Section 3 we provide the assumptions and the main result in finite time horizon, in which we obtain the asymptotic estimation of the behavior of the total, discounted claims of model ( 1 ), on a rare set A . As direct consequence of it, we can approximate the ruin probability on each rare set. In Section 4 , we generalize the results of Section 3 , in infinite time horizon. For this, we are forced to restrict ourselves in MRV