Integro-Differential Equations for a Jump-Diffusion Risk Process with Dependence between Claim Sizes and Claim Intervals ()
1. Introduction
In the risk process that is perturbed by diffusion, the surplus process
of an insurance portfolio is given by
(1)
where
is the initial surplus,
is the positive constant premium income
rate,
is the aggregate claims process, in which
![](//html.scirp.org/file/8-1720732x7.png)
is the claim number process (denoting the number of claims up to time t), and the interarrival times
is a sequence of positive random variables.
is a sequence of nonnegative independent identically distributed (i.i.d.) random variables with distribution function
and density function
,
is a standard Brownian motion that is independent of the aggregate claims process
,
is a positive constant.
2. Improved Risk Model
In this paper, it is assumed that the claim occurrence process to be of the following type: If a claim
is larger than a random variable
, then the time until the next claim
is exponentially distributed with rate
, otherwise it is exponentially distributed with rate
. The quantities
are assumed to be i.i.d. random variables with distribution function
. Assuming that
,
which is the net profit condition.
In the daily operation of insurance company, in addition to the premium income and claim to the operation of spending has a great influence on the outside, and there is also a factor that interest rates should not be neglected. As in [15] , this paper assume that the risk model Equation (1) is invested in a stochastic interest process which is assumed to be a geometric Brownian motion
, where r and σ2 are positive constants, and
is a standard Brownian motion independent of
. Let
denote the surplus of the insurer at time t under this investment assumption. Thus,
(2)
Denote T to be the ruin time (the first time that the surplus becomes negative), i.e.,
![]()
and
if
.
This article is interested in the expected discounted penalty (Gerber-Shiu) function:
, (3)
where
is the indicator function,
is the force of interest and
is a nonnegative function of
and satisfies
.
Furthermore, let
be the time when the first claim occurs, and random variable
being exponentially distributed with rate
. Assuming that
,
.
For
, define
, (4)
such that
, (5)
then,
.
3. Integro-Differential Equations for ![]()
In this section, a system of integro-differential equations with initial value conditions satisfied by the Gerber-Shiu function
is derived.
Define
, and
(6)
Lemma 3.1 Let
for
. For
define the hitting time
. Then, for
, it can be concluded that
(7)
Proof
is a reflecting diffusion with generator
,
acting on functions satisfying the reflecting boundary condition
.
If
![]()
and
for t > 0, then, according to Itô’s formula
is a local mar-
tingale. Using the separation variable technique, we find that
![]()
is a solution, where
,
is a solution of
.
Here
.
Using the initial condition
for
, we get
, consequently
.
Applying the Optional Stopping Theorem, it follows that
,
and thus
.
This ends the proof of Lemma 3.1.
Similarly, the following lemma can also be obtained.
Lemma 3.2 Let
for
. For
define the hitting time
. Then, for
, it can be concluded that
(8)
Theorem 3.1 Assuming that
is second order continuously differentiable functions in u, then
satisfies the following integro-differential equation
, (9)
, (10)
with the initial value conditions
,
.
Proof Let
be the time when the first claim occurs which exponentially distributed with rate
. Consider the risk process
defined by Equation (2) in an infinitesimal time interval
. There are three possible cases in
as follows.
1) There are no claims in
with probability
, thus
;
2) There is exactly one claim in
with probability
. According to different of the claim amount, there are three possible cases in this case as follows.
a) The amount of the claim
, i.e., ruin does not occur, and thus
;
b) The amount of the claim
, i.e., ruin occurs due to the claim;
c) The amount of the claim
, i.e., ruin occurs due to oscillation (observe that the probability that this case occurs is zero).
3) There is more than one claim in
with probability
.
Thus, considering the three cases above and noting that
is a strong Markov process, we have
(11)
By Taylor expansion, we have
, thus Equation (11) becomes
(12)
Then, by Itô’s formula we have
. (13)
Therefore, by dividing t on both sides of Equation (12), letting
, using Equation (13), we obtain Equation (9), and similarly we can obtain Equation (10).
The condition
follows from the oscillating nature of the sample paths of
. Now, we prove
.
For all
, let
,
. Then, by the strong property of
, it can be concluded that
![]()
According to Lemma 3.1, it can be concluded that
,
![]()
Thus,
, and correspondingly
. Similar results can be derived for
.
And for all
, let
,
, according to Lemma 3.2 we obtain
, thus
.
This ends the proof of Theorem 3.1.
4. Differential Equations for ![]()
Let
and
in Equation (3), correspondingly the expected discounted penalty function
turns into the ultimate ruin probability
.
Obviously,
,
and
.
Suppose that
,
,
and
is exponentially distributed with rate
. Then, we get the following theorem.
Theorem 4.1 Assuming that
is second order continuously differentiable functions in u, then
satisfies the following integro-differential equation
(14)
(15)
with the initial value conditions
![]()
Proof According to Equation (9), it can be concluded that
![]()
By taking the derivative with respect to u on both sides of the above formula, and after some careful calculations, we obtain Equation (14). And similarly we can prove that Equation (15) holds. This ends the proof of Theorem 4.1.
5. Conclusion
In this paper, we consider a jump-diffusion risk process compounded by a geometric Brownian motion with dependence between claim sizes and claim intervals. We derive the integro-differential equations for the Gerber-Shiu functions and the ultimate ruin probability by using the martingale measure. Further studies are needed for the numerical solution of Equations (9), (10), (14) and (15). The results derived in this paper can be generalized to similar dependence ruin models.
Acknowledgements
This research was supported by the National Natural Science Foundation of China (No. 11601036), the Natural Science Foundation of Shandong (No. ZR2014GQ005) and the Natural Science Foundation of Binzhou University (No. 2016Y14).