Etude des marchés d'assurance non-vie à l'aide d'équilibres de ...
Etude des marchés d'assurance non-vie à l'aide d'équilibres de ...
Etude des marchés d'assurance non-vie à l'aide d'équilibres de ...
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
tel-00703797, version 2 - 7 Jun 2012<br />
4.2. Mo<strong>de</strong>l formulation<br />
which implies an exponential <strong>de</strong>crease for ψ(u). Adopting a mixing approach, we will focus<br />
on three particular cases of special interest. We also obtain asymptotics for the tail of the<br />
resulting claim distributions and then discuss the <strong>de</strong>pen<strong>de</strong>nce structure involved.<br />
The paper is organized as follows. Section 4.2 <strong><strong>de</strong>s</strong>cribes the mixing approach for both<br />
continuous and discrete time mo<strong>de</strong>ls. Section 4.3 establishes the asymptotic rule A + B/u<br />
and some variants. Section 4.4 focuses on special features of the discrete time mo<strong>de</strong>l. Except<br />
mentioned otherwise, all numerical illustrations are done with the R statistical software (R<br />
Core Team (2012)).<br />
4.2 Mo<strong>de</strong>l formulation<br />
This section is <strong>de</strong>voted to the presentation of <strong>de</strong>pen<strong>de</strong>nt risk mo<strong>de</strong>ls, first in the continuous<br />
time framework and then in the discrete time framework. In addition to a general formula<br />
of the ruin probability un<strong>de</strong>r the mixing approach, we present two and three special cases of<br />
mixing distributions for both time scales.<br />
4.2.1 Continuous time framework<br />
In this subsection, we present the continuous time framework based on the classic Cramér-<br />
Lundberg mo<strong>de</strong>l.<br />
Surplus process<br />
The free surplus of an insurance company at time t is mo<strong>de</strong>led by<br />
Nt <br />
Ut = u + ct − Xi,<br />
where u is the initial surplus, c is the premium rate, (Xi)i are the claim amounts and (Nt)t≥0<br />
is the Poisson claim arrival process with intensity λ. We assume that the (Xi)i are i.i.d.<br />
conditionally on a latent variable Θ (distributed as X|Θ = θ, say); they are in<strong>de</strong>pen<strong>de</strong>nt of<br />
the claim arrival process. Θ can be interpreted as the heterogeneity in the claim process. In<br />
such setting, the claim sizes (X1, . . . , Xn) are <strong>de</strong>pen<strong>de</strong>nt random variables.<br />
Ruin probabilities<br />
Ruin occurs as soon as the surplus process becomes negative. Conditionally on Θ = θ, the<br />
ruin probability is thus <strong>de</strong>fined as<br />
i=1<br />
ψ(u, θ) = P (∃t > 0 : Ut < 0|U0 = u, Θ = θ).<br />
To <strong>de</strong>termine such a probability, a standard method consists in looking at the state of the<br />
surplus after the first claim arrival. This leads to an integro-differential equation that can<br />
be solved by using Laplace-Stieltjes transforms, see, e.g., Asmussen and Albrecher (2010). In<br />
the case of exponentially distributed claims (Xi)i ∼ E(θ), we have the well-known following<br />
formula<br />
<br />
λ<br />
λ<br />
ψ(u, θ) = min e−u(θ− c<br />
θc ) <br />
, 1 ,<br />
175