Next Article in Journal
Firm’s Credit Risk in the Presence of Market Structural Breaks
Previous Article in Journal
Towards a Topological Representation of Risks and Their Measures
 
 
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Asymptotic Ruin Probability of a Bidimensional Risk Model Based on Entrance Processes with Constant Interest Rate

College of Mathematics and Statistics, Northwest Normal University, Lanzhou 730070, China
*
Author to whom correspondence should be addressed.
Risks 2018, 6(4), 135; https://doi.org/10.3390/risks6040135
Submission received: 2 October 2018 / Revised: 5 November 2018 / Accepted: 15 November 2018 / Published: 20 November 2018

Abstract

:
In this paper, the risk model with constant interest based on an entrance process is investigated. Under the assumptions that the entrance process is a renewal process and the claims sizes satisfy a certain dependence structure, which belong to the different heavy-tailed distribution classes, the finite-time asymptotic estimate of the bidimensional risk model with constant interest force is obtained. Particularly, when inter-arrival times also satisfy a certain dependence structure, these formulas still hold.

1. Introduction

In this paper, we investigate a bidimensional risk model based on entrance processes, in which an insurance company operates two kinds of business. Suppose that the initial insurance fund for i-th class is x i and S j ( i ) is entry time of the j-th policy with 0 < S 1 ( i ) < S 2 ( i ) < and S j ( i ) = k = 1 j θ k ( i ) , θ 1 ( i ) = S 1 ( i ) , i = 1 , 2 . The corresponding renewal process, up to the time t, is
N i ( t ) = j = 1 I { S j ( i ) t } .
where I { · } is the indicator function.
Denote renewal function by m i ( t ) = E ( N i ( t ) ) = j = 1 P ( S j ( i ) t ) , t 0 and suppose the m i ( t ) < for all 0 < t < and m i ( 0 ) = 0 , i = 1 , 2 . Let the validity time of the j-th policy be { C j ( i ) , j = 1 , 2 , } with probability P ( C j ( i ) = α ) = p , = 1 , 2 , , K , and α 1 < α 2 < < α K , where they are independent and identically distributed. The premium is f i ( C j ( i ) ) and f i ( · ) is a strictly increasing function. D j ( i ) is claim time of the j-th policy and independent and identically distributed function H i ( · ) . Y j ( i ) is the j-th claim size and identically distributed function F i ( · ) . Suppose that D j ( i ) , Y j ( i ) and C j ( i ) have the same distributions with random variables D i , Y i and C i , respectively.
Assume that an insurance company invests risk free market with force of interest δ > 0 , then up to time t, the surplus process of the insurance company is written as:
R 1 ( t ) R 2 ( t ) = x 1 e δ t x 2 e δ t + j = 1 N 1 ( t ) f 1 ( C j ( 1 ) ) e δ ( t S j ( 1 ) ) j = 1 N 2 ( t ) f 2 ( C j ( 2 ) ) e δ ( t S j ( 2 ) ) j = 1 N 1 ( t ) Y j ( 1 ) e δ ( t S j ( 1 ) D j ( 1 ) ) I { S j ( 1 ) + D j ( 1 ) t , D j ( 1 ) C j ( 1 ) } j = 1 N 2 ( t ) Y j ( 2 ) e δ ( t S j ( 2 ) D j ( 2 ) ) I { S j ( 2 ) + D j ( 2 ) t , D j ( 2 ) C j ( 2 ) } .
Now the following two types of ruin times for a bidimensional risk model based on entrance processes are considered. we define the first time when both R 1 ( t ) and R 2 ( t ) become negative by
τ max ( x 1 , x 2 ) = inf { t : max { R 1 ( t ) , R 2 ( t ) } < 0 | R i ( 0 ) = x i , i = 1 , 2 } ,
the first time when both R 1 ( t ) or R 2 ( t ) become negative by
τ min ( x 1 , x 2 ) = inf { t : min { R 1 ( t ) , R 2 ( t ) } < 0 | R i ( 0 ) = x i , i = 1 , 2 } .
Then we define the corresponding ruin probabilities with the finite time t > 0 respectively by
ψ max ( x 1 , x 2 , T ) = P ( τ max ( x 1 , x 2 ) T ) = P ( i = 1 2 { R i ( s ) < 0 } for some 0 s T ) ,
and
ψ min ( x 1 , x 2 , T ) = P ( τ min ( x 1 , x 2 ) T ) = P ( i = 1 2 { R i ( s ) < 0 } for some 0 s T ) .
We know that Li et al. (2005) put forward into a new model ( LIG model ) based on an entrance process and discussed asymptotic normality of the risk process. Furthermore, Some scholars got some conclusions through the study of the LIG model. Li and Kong (2007) discussed the weak convergence properties of the model. Xiao et al. (2008) studied some limit properties of the model under constant interest. Xiao and Tang (2009) studied the infinite ruin probability with constant interest within Poisson process and class R α . Xiao et al. (2013) discussed the ruin probability of LIG model. It is clear that the above literatures are improved and investigated for one-dimensional risk model based on entrance processes. Recently, people have been interested in two-dimensional risk model, see, for example, Chan et al. (2003); Li et al. (2007); Zhang and Wang (2012) and so on. It is well known that these literatures are investigated in the classical model, and the risk model based on entrance processes is more important and actual. Therefore, on the basis of above literatures, we consider a bidimensional risk model based on entrance processes.
Because Theorem 1 of Xiao and Tang (2009) is obtained under the Poisson process and the regular variation class, we know that this is far from the actual. Hence, in this paper, we consider the LIG model and obtain the finite-time ruin probability under the class L D with constant interest when claim sizes satisfy a certain dependence under the renewal process. The conclusion also extend the above Theorem 1 and Theorem 3.1 of Xiao et al. (2013). At the same time, it indicates that tail characteristics of claim distribution determine the ruin probability of insurance company, which is of great significance to the safe operation and the risk assessment of insurance company.
This paper is organized as follows: The second Section introduces the preliminary knowledge. The third Section presents the main results of this paper. The fourth Section gives some lemmas. Finally, the fifth Section gives the proofs of main Theorems.

2. Some Preliminaries

Firstly, we give some markers. All limit relationships of this paper are for x unless stated otherwise. For the two positive function f ( · ) and g ( · ) , if lim   sup f ( x ) g ( x ) < , write f ( x ) = O ( g ( x ) ) ; if lim f ( x ) g ( x ) = 0 , write f ( x ) = o ( g ( x ) ) ; if lim   sup f ( x ) g ( x ) 1 , write f ( x ) g ( x ) ; if lim inf f ( x ) g ( x ) 1 , write f ( x ) g ( x ) ; if lim f ( x ) g ( x ) = 1 , write f ( x ) g ( x ) ; if f ( x ) = O ( g ( x ) ) , g ( x ) = O ( f ( x ) ) , write f ( x ) g ( x ) .
Here are some important concepts of heavy-tailed distributions.
Definition 1.
Say a distribution F belongs to the class L , if F satisfies for any y > 0 (or equivalent for y = 1 )
lim x F ¯ ( x + y ) F ¯ ( x ) = 1 .
Say a distribution F belongs to the class D , if F satisfies for any 0 < y < 1 (or equivalent for y = 1 2 )
lim   sup x F ¯ ( x y ) F ¯ ( x ) < .
Say a distribution F belongs to the class C , if F satisfies
lim y 1 lim   sup x F ¯ ( x y ) F ¯ ( x ) = 1 .
where their relationship is as follows:
C L D L .
For more properties and applications of the heavy-tailed distribution, we can refer to Bingham et al. (1987) and Embrechts et al. (1997).
There are an important relationship between heavy-tailed distribution and Matuszewska index of the distribution, which is defined by
J F + = lim x log F ¯ ( x ) log x , J F = lim x log F ¯ ( x ) log x ,
where
F ¯ ( y ) = lim inf x F ¯ ( x y ) F ¯ ( x ) , F ¯ ( y ) = lim   sup x F ¯ ( x y ) F ¯ ( x ) .
Furthermore, other indices of the distribution F can be defined by
L F = lim y 1 F ¯ ( y ) .
As for any y > 0 , there is F ¯ ( y ) = 1 / F ¯ ( y ) , hence
L F = lim y 1 F ¯ ( y ) = 1 / lim y 1 F ¯ ( y ) .
Particularly, if F C , then L F = 1 .
For more properties and applications of the heavy tailed distribution, we can refer to Tang and Tsitsiashvili (2003) and Yang and Wang (2010).
Here we introduce some concepts and properties of dependence.
Definition 2.
If there exists the finite real sequence { g U ( n ) , n 1 } for x i ( , ) , 1 i n such that
P ( i = 1 n { ξ i > x i } ) g U ( n ) i = 1 n P ( ξ i > x i ) ,
then we say random variable sequence { ξ n , n 1 } are widely upper orthant dependent ( W U O D ) .
If there exists the finite real sequence { g L ( n ) , n 1 } for x i ( , ) , 1 i n such that
P ( i = 1 n { ξ i x i } ) g L ( n ) i = 1 n P ( ξ i x i ) ,
then we say random variable sequence { ξ n , n 1 } are widely lower orthant dependent ( W L O D ) .
Furthermore, if { ξ n , n 1 } satisfy W U O D and W L O D at the same time, then we say random variables { ξ n , n 1 } are widely orthant dependent ( W O D ) .
For more detailed information, we can refer to Wang et al. (2013); Ghosh (1981) and Block et al. (1982).
Definition 3.
If real valued random variables X i , i 1 with distribution functions F i , i 1 satisfy for any i j
lim x P ( | X i | X j > x | X i X j > x ) = 0 ,
or, equivalently
lim x P ( X i > x , X j > x ) + P ( X i < x , X j > x ) F ¯ i ( x ) + F ¯ j ( x ) = 0 ,
then we say random variable variables X i , i 1 are pairwise quasi-asymptotically independent ( P Q A I ) .
If real valued random variables X i , i 1 with distribution functions F i , i 1 satisfy for any i j
lim x i x j P ( | X i | > x | X j > x ) = 0 ,
or, equivalently
lim x i x j P ( X i > x , X j > x ) + P ( X i < x , X j > x ) F ¯ j ( x ) = 0 ,
then we say random variable variables X i , i 1 are pairwise strong quasi-asymptotically independent ( P S Q A I ).
Remark 1.
If random variables X i , i 1 are P S Q A I , then they are P Q A I .
For more detailed information, we can refer to Li (2013) and Liu et al. (2012).
The first lemma comes from Theorem 2.1 of Li (2013).
Lemma 1.
Assume that { X j , 1 j n } are n real-valued random variables with functions of distribution F j , 1 j n . Then
P ( j = 1 n c j X j > x ) j = 1 n P ( c j X j > x ) .
holds if either (i) { X j } are P S Q A I and F j L D for 1 j n and ( c 1 , , c n ) [ a , b ] n , or (ii) { X j } are P Q A I and F j C for 1 j n and ( c 1 , , c n ) [ a , b ] n .
The following lemma comes from Proposition 2.2.1 of Bingham et al. (1987).
Lemma 2.
Let F D . For any 0 < α < α < β < β , there exist positive constants A i and B i , i = 1 , 2 satisfying the following inequality
F ¯ ( y ) F ¯ ( x ) A 1 ( x / y ) α ,
for any x y B 1 , and the inequality
F ¯ ( y ) F ¯ ( x ) A 2 ( x / y ) β .
for any x y B 2 .
The following lemma comes from Theorem 3.3 of Cline and Samorodnitsky (1994), Lemma 3.4 of Liu and Wang (2016) and Lemma 3.5 of Tang and Tsitsiashvili (2003).
Lemma 3.
Let X be a random variable with distribution F and Y be a random variable independent of X. Suppose that H is the distribution of X Y . If E Y p < for any 0 < J F + β < β and some p > J F + , then there exist the following conclusions:
(i) 
If F D , t h e n F ¯ ( x ) H ¯ ( x ) a n d x β = o ( F ¯ ( x ) ) ;
(ii) 
If F C , H C .
The following lemma can be proved in Appendix A.
Lemma 4.
( 1 ) Under the conditions of Theorem 1 ( or Theorem 2 ) , for all T T 0 , then we have
P ( j = 1 Y j ( i ) e δ ( S j ( i ) + D j ( i ) ) I { S j ( i ) + D j ( i ) T , D j ( i ) C j ( i ) } I { S j ( i ) T } > x i ) j = 1 P ( Y j ( i ) e δ ( S j ( i ) + D j ( i ) ) I { S j ( i ) + D j ( i ) T , D j ( i ) C j ( i ) } I { S j ( i ) T } > x i ) .
( 2 ) Suppose the conditions of Theorem 2 are true and the inter-arrival times { θ j i , j 1 } are W L O D random variables satisfying ( 16 ) for some ϵ i 0 > 0 , depending on F i and G i , i = 1 , 2 . Then, for all T T 0 , the relation ( 6 ) still holds uniformly.

3. Main Results

In this paper, we make the following assumptions:
A 1 Assume that the i-th class of random variables { C j ( i ) , j 1 } , { D j ( i ) , j 1 } , { Y j ( i ) , j 1 } , { S j ( i ) , j 1 } , i = 1 , 2 , are independent mutually.
A 2 Assume that E ( N i ( t ) ) p + 1 < for any fixed t > 0 and some p > J F i + , i = 1 , 2 .
Theorem 1.
Consider the bidimensional risk model ( 1 ) under the assumptions A 1 A 2 . Let p 0 = P ( S 1 ( i ) < T 0 ) > 0 for some 0 < T 0 < . Assume that claim sizes, { Y j ( i ) , j 1 } be P S Q A I random variables with common distribution F i L D , i = 1 , 2 such that J F i > 0 , respectively. Then for T T 0 , we have
ψ max ( x 1 , x 2 , T ) i = 1 2 ( = 1 K p 0 T 0 α ( T s ) F ¯ i ( e δ ( s + y ) x i ) d H i ( y ) d m i ( s ) ) .
and
ψ min ( x 1 , x 2 , T ) i = 1 2 = 1 K p 0 T 0 α ( T s ) F ¯ i ( e δ ( s + y ) x i ) d H i ( y ) d m i ( s ) .
where a b = min { a , b } .
Proof. 
(i) Firstly, we deal with the relation ( 7 ) . Write U i ( t ) = j = 1 N i ( t ) Y j ( i ) e δ ( S j ( i ) + D j ( i ) ) I { S j ( i ) + D j ( i ) t , D j ( i ) C j ( i ) } . Due to ( 1 ) and ( 2 ) , we know that for all T T 0 ,
ψ max ( x 1 , x 2 , T ) = P ( i = 1 2 { U i ( s ) j = 1 N i ( t ) f i ( C j ( i ) ) e δ S j ( i ) > x i } for some 0 s T ) .
By ( 9 ) , we know for all T T 0 ,
P ( i = 1 2 { U i ( T ) j = 1 N i ( T ) f i ( C j ( i ) ) e δ S j ( i ) > x i } ) ψ max ( x 1 , x 2 , T ) P ( i = 1 2 { U i ( T ) > x i } ) .
Because Lemma 4 ( 1 ) and random variables { C j ( i ) , j 1 } , { D j ( i ) , j 1 } , { Y j ( i ) , j 1 } , { S j ( i ) , j 1 } , i = 1 , 2 , are independent mutually, it is clear that
ψ max ( x 1 , x 2 , T ) i = 1 2 P ( U i ( T ) > x i ) i = 1 2 P ( j = 1 Y j ( i ) e δ ( S j ( i ) + D j ( i ) ) I { S j ( i ) + D j ( i ) T , D j ( i ) C j ( i ) } I { S j ( i ) T } > x i ) i = 1 2 ( = 1 K p 0 T 0 α ( T s ) F ¯ i ( e δ ( s + y ) x i ) d H i ( y ) d m i ( s ) ) .
holds uniformly for all T T 0 .
Because f i ( C j ( i ) ) , j = 1 , , K are bounded and { Y j ( i ) , j 1 } are P S Q A I random variables with distributions F i L D , i = 1 , 2 , respectively, it is easy to prove that { Y j ( i ) + f i ( C j ( i ) ) , j 1 } are P S Q A I random variables, whose distributions belong to the class L D , i = 1 , 2 , respectively. Hence, by Lemma 4 ( 1 ) , we know
ψ max ( x 1 , x 2 , T ) P ( i = 1 2 { U i ( T ) j = 1 N i ( T ) f i ( C j ( i ) ) e δ S j ( i ) > x i } ) = i = 1 2 P ( { U i ( T ) j = 1 N i ( T ) f i ( C j ( i ) ) e δ S j ( i ) > x i } ) = i = 1 2 P ( j = 1 N i ( t ) ( Y j ( i ) e δ ( S j ( i ) + D j ( i ) ) I { S j ( i ) + D j ( i ) t , D j ( i ) C j ( i ) } f i ( C j ( i ) ) e δ S j ( i ) ) > x i ) i = 1 2 P ( j = 1 Y j ( i ) e δ ( S j ( i ) + D j ( i ) ) I { S j ( i ) + D j ( i ) t , D j ( i ) C j ( i ) } I { S j ( i ) t } > x i ) i = 1 2 ( = 1 K p 0 T 0 α ( T s ) F ¯ i ( e δ ( s + y ) x i ) d H i ( y ) d m i ( s ) ) .
Combining ( 11 ) with ( 12 ) , we obtain that ( 7 ) holds uniformly for all T T 0 .
Next, we handle ( 8 ) . By ( 3 ) , we know that
ψ min ( x 1 , x 2 , T ) = P ( τ 1 ( x 1 T ) ) + P ( τ 2 ( x 2 T ) ) ψ max ( x 1 , x 2 , T ) .
where ( τ i ( x i ) = inf { t : R i ( t ) < 0 | R i ( 0 ) = x i } , i = 1 , 2 . Then by ( 7 ) with its one-dimensional case, it is clear that
P ( τ i ( x i ) T ) = 1 K p 0 T 0 α ( T s ) F ¯ i ( e δ ( s + y ) x i ) d H i ( y ) d m i ( s ) , i = 1 , 2 .
holds uniformly for all T T 0 .
Again by ( 7 ) and F i L D D , i = 1 , 2 , we have that
lim   sup x 1 x 2 sup T Λ [ T 0 , ) ψ max ( x 1 , x 2 , T ) i = 1 2 = 1 K p 0 T 0 α ( T s ) F ¯ i ( e δ ( s + y ) x i ) d H i ( y ) d m i s lim   sup x 1 x 2 sup T Λ [ T 0 , ) F ¯ 1 ( x 1 ) m 1 ( T ) F ¯ 2 ( x 2 ) m 2 ( T ) ( = 1 K p H i ( α ( T s ) ) ) 2 F ¯ 1 ( x 1 e δ T ) m 1 ( T 0 ) ( = 1 K p H i ( α ( T 0 s ) ) ) = 0 .
Hence, by ( 13 ) ( 15 ) , we prove that ( 8 ) holds uniformly for all T T 0 . □
Theorem 2.
Consider the bidimensional risk model ( 1 ) under the assumption A 1 A 2 . Let p 0 = P ( S 1 ( i ) < T 0 ) > 0 for some 0 < T 0 < . Assume that claim sizes, { Y i , i 1 } be P Q A I random variables with common distribution F i C , i = 1 , 2 such that J F i > 0 , respectively. Then the relations ( 7 ) and ( 8 ) hold uniformly for T T 0 .
Proof. 
Similarly, when f i ( C j ( i ) ) , j = 1 , , K are bounded and { Y j ( i ) , j 1 } are P Q A I random variables with distributions F i C , i = 1 , 2 , respectively, it is easy to prove that { Y j ( i ) + f i ( C j ( i ) ) , j 1 } are P Q A I random variables, whose distributions belong to the class C , i = 1 , 2 , respectively. Hence, applying the same method of proof of Theorem 1, we know that ( 7 ) and ( 8 ) still hold uniformly for all T T 0 . □
Theorem 3.
Under the conditions of Theorem 2, suppose that entry inter-arrival times, { θ k ( i ) , k 1 } are W L O D random variables with common distribution G i satisfying
lim n g L i ( n ) e ϵ i 0 n = 0 .
holds for some ϵ i 0 > 0 , depending on F i and G i , i = 1 , 2 . Then, the relations ( 7 ) and ( 8 ) still hold uniformly for all T T 0 .
Proof. 
By the Lemma 4 ( 2 ) , Theorem 2 and similar proof of Theorem 1, the relations ( 7 ) and ( 8 ) still hold uniformly for all T T 0 . □
Corollary 1.
Consider one-dimensional risk model satisfying the same conditions as those in Theorem 1, and denote ruin time by τ ( x ) = inf { t : R ( t ) < 0 | R ( 0 ) = x } , where
R ( t ) = x e δ t + j = 1 N ( t ) f ( C j ) e δ ( t S j ) j = 1 N ( t ) Y j e δ ( t S j D j ) I { S j + D j t , D j C j } ,
then, we have
ψ ( x , T ) = P ( τ ( x ) T ) = 1 K p 0 T 0 α ( T s ) F ¯ ( e δ ( s + y ) x ) d H ( y ) d m ( s ) .
It is easy to prove Corollary 1 from the proof of relation ( 7 ) of Theorem 1.
Remark 2.
Corollary 1 is a partial extension for the results of Theorem 3.1 of Xiao et al. (2013), Theorem 1 of Xiao and Tang (2009), and Theorem 3.1 of Xiao and Xie (2018).

4. Conclusions

In summary, this paper studies the two-dimensional independent risk model based on entrance processes with constant interest rate. Under the assumptions that the entry process of policies of two kinds of business of insurance companies have different renewal processes, the claims sizes of two kinds of business are independent of each other, and the claims sizes of the same kind of business are pairwise strong quasi-asymptotically independent, which belong to the class L D , the maximum finite-time ruin probability and the minimum finite-time ruin probability are obtained, respectively. If intervals of entry time of the policy satisfy the wide lower quadrant dependence, The finite-time maximum ruin probability and the finite-time minimum ruin probability are also obtained.

Author Contributions

L.X. completed the article preliminarily, and the Professor H.X. perfected it.

Funding

This research was funded by Natural Scientific Funds of China (71261023).

Conflicts of Interest

The authors declare no conflicts of interest.

Appendix A. Proof of Lemma 4

Proof. 
( 1 ) Firstly, we deal with the upper bound of ( 6 ) . Along with the method of proof of Theorem 2.1 of Hao and Tang (2008), for any positive integer N, we have
P ( j = 1 Y j ( i ) e δ ( S j ( i ) + D j ( i ) ) I { S j ( i ) + D j ( i ) T , D j ( i ) C j ( i ) } I { S j ( i ) T } > x i ) = P ( j = 1 N i ( T ) Y j ( i ) e δ ( S j ( i ) + D j ( i ) ) I { S j ( i ) + D j ( i ) T , D j ( i ) C j ( i ) } > x i ) = ( n = 1 N + n = N + 1 ) P ( j = 1 n Y j ( i ) e δ ( S j ( i ) + D j ( i ) ) I { S j ( i ) + D j ( i ) T , D j ( i ) C j ( i ) } > x i , N i ( T ) = n ) = I 1 ( x i , t , N ) + I 2 ( x i , t , N ) .
For I 2 ( x i , t , N ) . Because of the Lemma 2, there exist positive constants β , A 1 and B 1 for any h > 0 and x i n B 1 . Hence, we have
I 2 ( x i , t , N ) n = N + 1 P ( j = 1 n Y j ( i ) e δ ( S 1 ( i ) + D j ( i ) ) I { S 1 ( i ) + D j ( i ) T , D j ( i ) C j ( i ) } > x i , S n ( i ) T S n + 1 ( i ) ) = n = N + 1 0 T P ( j = 1 n Y j ( i ) e δ ( s + D j ( i ) ) I { s + D j ( i ) T , D j ( i ) C j ( i ) } > x i , S n ( i ) S 1 ( i ) T s S n + 1 ( i ) S 1 ( i ) ) · P ( S 1 ( i ) d s ) n = N + 1 0 T P ( j = 1 n + 1 Y j ( i ) e δ ( s + D j ( i ) ) I { s + D j ( i ) T , D j ( i ) C j ( i ) } > x i ) · P ( N i ( T s ) = n ) P ( S 1 ( i ) d s ) = N n x i / B 1 0 T P ( j = 1 n + 1 Y j ( i ) e δ ( s + D j ( i ) ) I { s + D j ( i ) T , D j ( i ) C j ( i ) } > x i ) P ( N i ( T s ) = n ) d m i ( s ) + n > x i / B 1 0 T P ( j = 1 n + 1 Y j ( i ) e δ ( s + D j ( i ) ) I { s + D j ( i ) T , D j ( i ) C j ( i ) } > x i ) P ( N i ( T s ) = n ) d m i ( s ) N n x i / B 1 0 T ( n + 1 ) P ( Y j ( i ) e δ ( s + D j ( i ) ) I { s + D j ( i ) T , D j ( i ) C j ( i ) } > x i / ( n + 1 ) ) P ( N i ( T s ) = n ) · d m i ( s ) + n > x i / B 1 0 T P ( N i ( T s ) = n ) d m i ( s ) N n x i / B 1 0 T C 1 P ( Y i e δ ( s + D i ) I { s + D i T , D i C i } > x i ) ( n + 1 ) β + 1 P ( N i ( T s ) = n ) d m i ( s ) + 0 T P ( N i ( T s ) > x i / B 1 ) d m i ( s ) A 1 0 T P ( Y i e δ ( s + D i ) I { s + D i T , D i C i } > x i ) E ( N i ( T s ) + 1 ) β + 1 I { N N 0 ( T s ) x i / B 1 } d m i ( s ) + P ( N i ( T s ) > x i / B 1 ) m i ( T ) A 1 E ( N i ( T s ) + 1 ) β + 1 I { N N 0 ( T s ) x i / B 1 } 0 T P ( Y i e δ ( s + D i ) I { s + D i T , D i C i } > x i ) d m i ( s ) + e h ( x i / B 1 ) E ( e h N i ( T ) ) m i ( T ) .
If N , then we have
E ( N i ( T s ) + 1 ) β + 1 I { N N i ( T s ) x i / B 1 } 0 ,
Because of the Lemma 3.2 of Hao and Tang (2008), we know
e h ( x i / B 1 ) E ( e h N i ( T ) ) 0 .
Hence, for x i > 0 , we have
lim N sup t ( 0 , T ] I 2 ( x i , t , N ) 0 T P ( Y i e δ ( s + D i ) I { s + D i T , D i C i } > x i ) d m i ( s ) = lim N sup t ( 0 , T ] I 2 ( x i , t , N ) = 1 K p 0 T 0 α ( T s ) F ¯ ( e δ ( s + y ) x i ) d H i ( y ) d m i ( s ) = 0 .
For I 1 ( x i , t , N ) . Because of the Lemma 1, for t ( 0 , T ] , we have
I 1 ( x i , t , N ) n = 1 N i = 1 n P ( Y j ( i ) e δ ( S j ( i ) + D j ( i ) ) I { S j ( i ) + D j ( i ) T , D j ( i ) C j ( i ) } > x i , N i ( T ) = n ) ( n = 1 j = 1 n n = N + 1 j = 1 n ) P ( Y j ( i ) e δ ( S j ( i ) + D j ( i ) ) I { S j ( i ) + D j ( i ) T , D j ( i ) C j ( i ) } > x i , N i ( T ) = n ) = J 1 ( x i , t , N ) + J 2 ( x i , t , N ) .
For J 1 ( x i , t , N ) , we obtain
J 1 ( x i , t , N ) = j = 1 P ( Y j ( i ) e δ ( S j ( i ) + D j ( i ) ) I { S j ( i ) + D j ( i ) T , D j ( i ) C j ( i ) } > x i , N i ( T ) j ) = 0 T P ( Y i e δ ( s + D i ) I { s + D i T , D i C i } > x i ) d m i ( s ) = = 1 K p 0 T 0 α ( T s ) F ¯ ( e δ ( s + y ) x i ) d H i ( y ) d m i ( s ) .
For J 2 ( x i , t , N ) . Applying the similar method of dealing with I 2 ( x i , t , N ) , we have
J 2 ( x i , t , N ) n = N + 1 j = 1 n P ( Y j ( i ) e δ ( S 1 ( i ) + D j ( i ) ) I { S 1 ( i ) + D j ( i ) T , D j ( i ) C j ( i ) } > x i , N i ( T ) = n ) n = N j = 1 n + 1 P ( Y j ( i ) e δ ( s + D j ( i ) ) I { s + D j ( i ) T , D j ( i ) C j ( i ) } > x i ) · P ( N i ( T s ) = n ) d m i ( s ) 0 T P ( Y i e δ ( s + D i ) I { s + D i T , D i C i } > x i ) d m i ( s ) · n = N ( n + 1 ) P ( N i ( T ) n ) .
Hence, for x i > 0 ,
lim N sup t ( 0 , T ] J 2 ( x i , t , N ) 0 T P ( Y i e δ ( s + D i ) I { s + D i T , D i C i } > x i ) d m i ( s ) = lim N sup t ( 0 , T ] J 2 ( x i , t , N ) = 1 K p 0 T 0 α ( T s ) F ¯ i ( e δ ( s + y ) x i ) d H i ( y ) d m i ( s ) = 0 .
By the relations (A1)–(A5), we obtain
P ( j = 1 Y j ( i ) e δ ( S j ( i ) + D j ( i ) ) I { S j ( i ) + D j ( i ) T , D j ( i ) C j ( i ) } I { S j ( i ) T } > x i ) = 1 K p 0 T 0 α ( T s ) F ¯ i ( e δ ( s + y ) x i ) d H i ( y ) d m i ( s ) .
Next, we cope with the lower bound of ( 6 ) .
For m = 0 , 1 , 2 , , we write Δ m = j = 1 Y j ( i ) e δ ( S j ( i ) + D j ( i ) ) I { S j ( i ) + D j ( i ) T , D j ( i ) C j ( i ) } I { S j ( i ) T } . According to the similar method of Tang and Tsitsiashvili (2004), for all integer m such that j = m + 1 j 2 < 1 , we have
P ( Δ m > x i ) P ( j = m + 1 Y j ( i ) e δ ( S j ( i ) + D j ( i ) ) I { S j ( i ) + D j ( i ) T , D j ( i ) C j ( i ) } I { S j ( i ) T } > j = m + 1 x i n 2 ) P ( j = m + 1 ( Y j ( i ) e δ ( S j ( i ) + D j ( i ) ) I { S j ( i ) + D j ( i ) T , D j ( i ) C j ( i ) } I { S j ( i ) T } > x i n 2 ) ) j = m + 1 P ( Y j ( i ) e δ ( S j ( i ) + D j ( i ) ) > x i n 2 ) .
Because of the Lemma 2, there exist 0 < α < J F J F + < , A i and B i , i = 1 , 2 satisfying the relations ( 4 ) and ( 5 ) . The events are written as A 1 ( j , x i ) = ( j 2 e δ ( S j ( i ) + D j ( i ) ) B 2 / x i ) , A 2 ( j , x i ) = ( B 2 / x i < j 2 e δ ( S j ( i ) + D j ( i ) ) 1 ) and A 3 ( j , x i ) = ( j 2 e δ ( S j ( i ) + D j ( i ) ) > 1 ) . Hence, the relation (A7) is written as
P ( Δ m > x i ) j = m + 1 P ( Y j ( i ) e δ ( S j ( i ) + D j ( i ) ) > x i n 2 ) = k = 1 3 j = m + 1 P ( Y j ( i ) e δ ( S j ( i ) + D j ( i ) ) > x i n 2 , A k ( j , x i ) ) = k = 1 3 I k ( m , x i ) .
By Chebyshev’s inequality and the Lemma 3 ( i ) , we have
I 1 ( m , x i ) j = m + 1 P ( A 1 ( j , x i ) ) ( x i B 2 ) β j = 1 j 2 β ( E ( e δ β S 1 ( i ) ) ) j = o ( F ¯ ( x i ) ) .
Because of the relations ( 4 ) and ( 5 ) , for all x i max { B 1 , B 2 } , we can obtain
I 2 ( m , x i ) A 2 F ¯ ( x i ) j = m + 1 E ( j 2 β e δ β S j ( i ) I A 2 ( j , x i ) ) A 2 F ¯ ( x i ) j = m + 1 j 2 β ( E ( e δ β S 1 ( i ) ) ) j .
and
I 3 ( m , x i ) F ¯ ( x i ) A 1 j = m + 1 E ( j 2 α e δ α S j ( i ) I A 3 ( j , x i ) ) F ¯ ( x i ) A 1 j = m + 1 j 2 α ( E ( e δ α S 1 ( i ) ) ) j .
Because of F D and p 0 > 0 , there exists a positive number M > 0 satisfying for T T 0 ,
Y 1 ( i ) e δ ( S 1 ( i ) + D 1 ( i ) ) I { S 1 ( i ) + D 1 ( i ) T , D 1 ( i ) C 1 ( i ) } I { S 1 ( i ) T } P ( Y 1 ( i ) e δ ( S 1 ( i ) + D 1 ( i ) ) I { S 1 ( i ) + D 1 ( i ) T , D 1 ( i ) C 1 ( i ) } I { S 1 ( i ) T } , S 1 ( i ) T 0 ) p 0 P ( Y 1 ( i ) e δ ( T 0 + D 1 ( i ) ) I { S 1 ( i ) + D 1 ( i ) T , D 1 ( i ) C 1 ( i ) } I { S 1 ( i ) T } ) M F ¯ ( x i ) .
By the relations (A8)–(A10), we have
lim m lim   sup x i P ( Δ m > x i ) F ¯ ( x i ) = lim m lim   sup x i j = m + 1 P ( Y j ( i ) e δ ( S j ( i ) + D j ( i ) ) > x i n 2 ) F ¯ ( x i ) = 0 .
For ε > 0 , by the relations (A11) and (A12), there exist integer n 0 > 0 and enough large number x i . Then we have
P ( j = n 0 + 1 Y j ( i ) e δ ( S j ( i ) + D j ( i ) ) I { S j ( i ) + D j ( i ) T , D j ( i ) C j ( i ) } I { S j ( i ) T } > x i ) j = n 0 + 1 P ( Y j ( i ) e δ ( S j ( i ) + D j ( i ) ) > x i n 2 ) ε F ¯ ( x i ) ε P ( Y 1 ( i ) e δ ( S 1 ( i ) + D 1 ( i ) ) I { S 1 ( i ) + D 1 ( i ) T , D 1 ( i ) C 1 ( i ) } I { S 1 ( i ) T } ) .
Let the above number n 0 be fixed. Because of the Lemma 1, the relation (A13), we know
P ( j = 1 Y j ( i ) e δ ( S j ( i ) + D j ( i ) ) I { S j ( i ) + D j ( i ) T , D j ( i ) C j ( i ) } I { S j ( i ) T } > x i ) P ( j = 1 n 0 Y j ( i ) e δ ( S j ( i ) + D j ( i ) ) I { S j ( i ) + D j ( i ) T , D j ( i ) C j ( i ) } I { S j ( i ) T } > x i ) ( j = 1 j = n 0 + 1 ) P ( Y j ( i ) e δ ( S j ( i ) + D j ( i ) ) I { S j ( i ) + D j ( i ) T , D j ( i ) C j ( i ) } I { S j ( i ) T } > x i ) ( 1 ε ) j = 1 P ( Y j ( i ) e δ ( S j ( i ) + D j ( i ) ) I { S j ( i ) + D j ( i ) T , D j ( i ) C j ( i ) } I { S j ( i ) T } > x i ) .
Combining (A6) with (A14) and arbitrariness of ε , we prove the Lemma 4 ( 1 ) .
( 2 ) Because the inter-arrival times { θ j ( i ) , j 1 } are W L O D random variables satisfying ( 9 ) for some ϵ i 0 > 0 , depending on F i and G i , i = 1 , 2 , by ( 11 ) and ( 12 ) of Block et al. (1982), it is clear that
E e δ β S n ( i ) ( E e δ β S 1 ( i ) ) n .
holds for any β > 0 and n 1 . Hence, we know the relations (A8)–(A10) still hold. Along with the similar proof of Lemma 4 ( 1 ) , it is easy to prove Lemma 4 ( 2 ) . □

References

  1. Bingham, Nicholas H., Charles M. Goldie, and Jef L. Teugels. 1987. Regular Variation. Cambridge: Cambridge University Press. [Google Scholar]
  2. Block, Henry W., Thomas H. Savits, and Moshe Shaked. 1982. Some concepts of negative dependence. The Annals of Probability 10: 765–72. [Google Scholar] [CrossRef]
  3. Chan, Wai-Sum, Hailiang Yang, and Lianzeng Zhang. 2003. Some results on ruin probabilities in a two-dimensional risk model. Insurance: Mathematics and Economics 32: 345–58. [Google Scholar] [CrossRef]
  4. Cline, Daren B. H., and Gennady Samorodnitsky. 1994. Subexponentiality of the product of independent random variables. Stochastic Processes and Their Applications 49: 75–98. [Google Scholar] [CrossRef]
  5. Embrechts, Paul, Claudia Kllüppelberg, and Thomas Mikosch. 1997. Modelling Extremal Events for Insurance and Finance. Berlin: Springer. [Google Scholar]
  6. Ghosh, Malay. 1981. Multivariate negative dependence. Communications in Statistics-Theory and Methods 10: 307–37. [Google Scholar] [CrossRef]
  7. Hao, Xuemiao, and Qihe Tang. 2008. A uniform asymptotic estimate for discounted aggregate claims with subexponential tails. Insurance: Mathematics and Economics 43: 116–20. [Google Scholar] [CrossRef]
  8. Li, Jinzhu. 2013. On pairwise quasi-asymptotically independent random variables and their applications. Statistics and Probability Letters 83: 2081–87. [Google Scholar] [CrossRef]
  9. Li, Zehui, and Xinbing Kong. 2007. A new risk model based on policy entrance process and its weak convergence properties. Applied Stochastic Models in Business and Industry 23: 235–46. [Google Scholar] [CrossRef]
  10. Li, Zehui, Jinxia Zhu, and Feng Chen. 2005. Study of a risk model based on the entrance processes. Statistic and Probability Letters 72: 1–10. [Google Scholar] [CrossRef]
  11. Li, Junhai, Zaiming Liu, and Qihe Tang. 2007. On the ruin probabilities of a bidimensional perturbed risk model. Insurance: Mathematics and Economics 41: 185–95. [Google Scholar] [CrossRef]
  12. Liu, Rongfei, and Dingcheng Wang. 2016. The ruin probabilities of a discrete-time risk model with dependent insurance and financial risks. Journal of Mathematical Analysis and Applications 444: 80–94. [Google Scholar] [CrossRef]
  13. Liu, Xijun, Qingwu Gao, and Yuebao Wang. 2012. A note on a dependent risk model with constant interest rate. Statistics and Probability Letters 82: 707–12. [Google Scholar] [CrossRef]
  14. Tang, Qihe, and Gurami Tsitsiashvili. 2003. Precise estimates for the ruin probability in finite horizon in a discrete-time model with heavy-tailed insurance and financial risks. Stochastic Processes and their Applications 108: 299–325. [Google Scholar] [CrossRef]
  15. Tang, Qihe, and Gurami Tsitsiashvili. 2004. Finite-and infinite-time ruin probabilities in the presence of stochastic returns on investments. Advances in Applied Probability 36: 1278–99. [Google Scholar] [CrossRef]
  16. Wang, Kaiyong, Yuebao Wang, and Qingwu Gao. 2013. Uniform asymptotics for the finite-time ruin probability of a dependent risk model with a constant interest rate. Methodology and Computing in Applied Probability 15: 109–24. [Google Scholar] [CrossRef]
  17. Xiao, Hongmin, and Jiashan Tang. 2009. Asymptotic Ruin Probabilities of an Entrance Processes Based Risk Model with Interest Force and Regularly Varying Claims. Chinese Journal of Engineering Mathematics 26: 1126–32. [Google Scholar]
  18. Xiao, Hongmin, and Lin Xie. 2018. Asymptotic ruin probabilities of a dependent renewal risk model based on entrance processes with constant interest rate. Communications in Statistics-Theory and Methods 47: 5396–417. [Google Scholar]
  19. Xiao, Hongmin, Zehui Li, and Weiwei Liu. 2008. The limit behavior of a risk model based on entrance processes. Computers and Mathematics with Applications 56: 1434–40. [Google Scholar] [CrossRef]
  20. Xiao, Hong Min, Ze Hui Li, and Wei Wei Liu. 2013. The finite time ruin probability of a new risk model based on entrance process. Communications in Statistics-Theory and Methods 42: 336–45. [Google Scholar] [CrossRef]
  21. Yang, Yang, and Yuebao Wang. 2010. Asymptotics for ruin probability of some negatively dependent risk models with a constant interest rate and dominatedly-varying-tailed claims. Statistics and Probability Letters 80: 143–54. [Google Scholar] [CrossRef]
  22. Zhang, Yuanyuan, and Wensheng Wang. 2012. Ruin probabilities of a bidimensional risk model with investment. Statistics and Probability Letters 82: 130–38. [Google Scholar] [CrossRef]

Share and Cite

MDPI and ACS Style

Xiao, H.; Xie, L. Asymptotic Ruin Probability of a Bidimensional Risk Model Based on Entrance Processes with Constant Interest Rate. Risks 2018, 6, 135. https://doi.org/10.3390/risks6040135

AMA Style

Xiao H, Xie L. Asymptotic Ruin Probability of a Bidimensional Risk Model Based on Entrance Processes with Constant Interest Rate. Risks. 2018; 6(4):135. https://doi.org/10.3390/risks6040135

Chicago/Turabian Style

Xiao, Hongmin, and Lin Xie. 2018. "Asymptotic Ruin Probability of a Bidimensional Risk Model Based on Entrance Processes with Constant Interest Rate" Risks 6, no. 4: 135. https://doi.org/10.3390/risks6040135

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop