Professional Documents
Culture Documents
Economic Modelling: Wenguang Yu
Economic Modelling: Wenguang Yu
Economic Modelling
journal homepage: www.elsevier.com/locate/ecmod
Some results on absolute ruin in the perturbed insurance risk model with
investment and debit interests
Wenguang Yu ⁎
School of Insurance, Shandong University of Finance and Economics, Jinan 250014, China
a r t i c l e i n f o a b s t r a c t
Article history: In this paper, we consider a perturbed compound Poisson risk model with investment and debit interests.
Accepted 21 December 2012 Dividends are paid to the shareholders according to a threshold dividend strategy. An alternative assumption
is that when the surplus is negative, a debit interest is applied and when the surplus is above a certain
MSC:
positive level, the insurer could earn investment interest. Integro-differential equations with boundary condi-
91B30
tions satisfied by the moment-generating function, the nth moment of the present value of all dividends until
absolute ruin and the Gerber–Shiu expected discounted penalty function are obtained. Then, we present the
Keywords: explicit expressions for the zero discounted nth moment of the present value of all dividends until absolute
Absolute ruin
ruin in the case of exponential claims. Finally, numerical example is also given to illustrate the effect of the related
Threshold dividend strategy
parameters on the first moment of the present value of all dividends until absolute ruin.
Gerber–Shiu expected discounted penalty
function © 2013 Elsevier B.V. All rights reserved.
Moment-generating function
Confluent hypergeometric function
1. Introduction asymptotic results for a more practical case with a higher borrowing
rate. Yang et al. (2008) investigate the absolute ruin problems in a
The classical risk model perturbed by a diffusion was first intro- multi-layer compound Poisson model with constant interest force. Yuan
duced by Gerber (1970) and has been further studied by many and Hu (2008) study the absolute ruin in the compound Poisson risk
authors during the last few years. See, for example, Tsai and Willmot model with nonnegative interest and a constant dividend barrier. Wang
(2002), Cai and Yang (2005), Wan (2007), Gao and Yin (2008), Wang et al. (2010) consider the dividend payments in a compound Poisson
and Wu (2008), Albrecher and Thonhausera (2008), Gao and Liu risk model with credit and debit interests under absolute ruin. Zhang et
(2010), Zhang et al. (2010), Liu and Liu (2011), Zhang (2012) and the al. (2011) study an absolute ruin model where claims arrive according
references therein. Generally, when the surplus is below zero, we say to a Markovian arrival process. Li and Liu (2012) investigate a regulated
that ruin occurs in the classical risk theory. But in reality, when the risk process, which is modeled by interest and linear dividend barrier.
surplus falls below zero or the insurer is on deficit, the insurer could Under absolute ruin, the expected discounted dividends are derived
borrow an amount of money equal to the deficits at a debit interest by PDMP method. Explicit solutions and numerical results are obtained
rate to continue his/her business. Meanwhile, the insurer will repay for exponential claims. Bai and Song (2012) consider the probability of
the debts continuously from his/her premium income. Thus, the surplus random time absolute ruin in the renewal risk model with constant pre-
of the insurer is driven under the debit interest rate when the surplus is mium rate and constant force of interest. However, there is no work that
negative. The negative surplus may return to a positive level if debts are deals with perturbed compound Poisson risk model with investment
reasonable. However, it is clear that when the negative surplus reaches and threshold strategy under absolute ruin. This motivates us to
some certain level, the surplus is no longer able to recover, and conse- investigate such a risk model in this work. Inspired by the work of Gao
quently absolute ruin occurs at this moment. and Liu (2010), we will extend their results to absolute ruin risk model.
Absolute ruin probability is an important risk measure and has been The rest of the paper is organized as follows. In Section 2, we describe
frequently considered in recent research works. Dassios and Embrechts the risk model. In Section 3, integro-differential equations with boundary
(1989) derive an explicit expression for the probability of absolute ruin conditions satisfied by the moment-generating function and the nth mo-
in the case of exponentially distributed individual claim amounts using ment of the present value of all dividends until absolute ruin are derived.
a martingale approach. Cai (2007) considers the Gerber–Shiu discounted In Section 4, we give the integro-differential equations satisfied by the
penalty function under absolute ruin. Zhu and Yang (2008) obtain Gerber–Shiu expected discounted penalty function. As applications, in
Section 5, we present explicit expressions for the nth moment of the
⁎ Tel.: +86 531 8859 6135. present value of all dividends until absolute ruin for exponential claims
E-mail address: yuwg@mail.sdu.edu.cn. when the interest force is zero. Finally, in Section 6, we use numerical
0264-9993/$ – see front matter © 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.econmod.2012.12.020
626 W. Yu / Economic Modelling 31 (2013) 625–634
example to illustrate the impact of the related parameters on the first An alternative expression for Du,b is
moment of the present value of all dividends until absolute ruin.
T b −αt
Du;b ¼ ε∫ e IðU b ðt Þ > bÞdt: ð2:4Þ
0
2. The risk model
It is obvious that 0 b Du,b ≤ ε/α.
In the actuarial literature, the surplus process of an insurance com- In the sequel we will be interested in the moment-generating function
pany is often modeled by the following perturbed compound Poisson
h i
risk process yD
Mðu; y; bÞ ¼ E e u;b ; ð2:5Þ
N ðt Þ
X
U ðt Þ ¼ u þ ct− X n þ σBðt Þ; t ≥0; ð2:1Þ and the nth moment function
n¼1
h i
where u≥0 is the initial surplus, c >0 is the gross premium rate, {N(t), V n ðu; bÞ ¼ E Dnu;b ; n∈N; ð2:6Þ
t≥0} is a Poisson process with jump intensity λ >0 denoting the number
of claims up to time t, and {Xn; n≥1}, representing the sizes of claims with V0(u;b) = 1, and the Gerber–Shiu expected discounted penalty
and independent of {N(t), t≥0}, is a sequence of independent and iden- function
tically distributed nonnegative random variables with a common distri- h
−αT b
bution function F(x) which satisfies F(0)=0 and has a positive mean Φðu; bÞ ¼ E e ωðU b ðT b −Þ; jU b ðT b ÞjÞIðT b b∞ÞjU b ð0Þ ¼ u; ð2:7Þ
∞
μ ¼ ∫ F ðxÞdx > 0. Here, F ðxÞ ¼ 1−F ðxÞ is the survival function of F(x).
0
where, Ub(Tb−) is the surplus prior to absolute ruin and |Ub(Tb)| is the
The net profit condition is given by c>λE[Xi]. σ>0 is a constant,
deficit at absolute ruin. The penalty function ω(x1, x2) is an arbitrary non-
representing the diffusion volatility parameter. {B(t), t≥0} is a standard
negative measurable function defined on (−c1/β, +∞)×(c1/β, +∞).
Brownian motion with B(0)=0. In addition, {N(t), t≥0}, {Xn; n≥1} and
Throughout this paper we assume that M(u, y;b), Vn(u;b) and Φ(u;b)
{B(t), t≥0} are mutually independent.
are sufficiently smooth functions in u and y in their respective domains.
Recently, Gao and Liu (2010) consider the compound Poisson risk
model perturbed by diffusion with constant interest and a threshold
3. Integro-differential equations for M(u, y;b) and Vn(u;b)
dividend strategy. Under the threshold dividend strategy, whenever the
surplus is above b, dividends are paid continuously at a constant rate ε.
Clearly, the moment-generating function M(u, y;b) behaves differ-
However when the surplus is below the level b, no dividends are paid.
ently, depending on whether its initial surplus u is below zero or
Then, integro-differential equations with certain boundary conditions
above the barrier level b. Hence, we write
for the moment-generation function and the nth moment of the present
value of all dividends until ruin are derived. Motivated by the work of 8
< M 2 ðu; y; bÞ; u > b;
Gao and Liu (2010), we consider the following extension of model Mðu; y; bÞ ¼ M 1 ðu; y; bÞ; 0 ≤ u ≤ b; ð3:1Þ
:
(2.1) which is enriched by investment, debit and threshold strategy. M 3 ðu; y; bÞ; −c1 =β b u ≤ 0:
We denote the aggregate dividends paid in the time interval [0, t]
by D(t) and the modified surplus by Ub(t) = U(t) − D(t), which is the For notational convenience, let
insurer's surplus at time t and Ub(0) = u. We assume that the insurer
pays dividends according to the following strategy governed by h1 ðu; t Þ ¼ ueβt þ c1 eβt −1 =β; h2 ðu; t Þ ¼ ueγt þ c2 eγt −1 =γ:
threshold parameters b > 0 and dividend rate ε > 0. Whenever the
modified surplus Ub(t) is below the level b, no dividends are paid Theorem 3.1. For 0 ≤ u ≤ b,
and the constant premium income rate is c1 > 0. However, when the
2 2
modified surplus Ub(t) is above b, dividends are paid continuously σ ∂ M1 ðu; y; bÞ ∂M1 ðu; y; bÞ ∂M1 ðu; y; bÞ
þ c1 ¼ λM1 ðu; y; bÞ þ αy
at a constant rate ε (0 b ε b c1), meanwhile the insurer could earn 2 ∂u2 ∂u ∂y
interest at investment rate γ > 0. The insurer could borrow an amount " #
u
c
uþ β1
of money equal to the deficit at a debit interest force β when the −λ ∫ M1 ðu−x; y; bÞdF ðxÞ þ ∫ M3 ðu−x; y; bÞdF ðxÞ þ F ðu þ c1 =β Þ ;
0 u
surplus is negative. Meanwhile, the insurer will repay the debts
continuously from her/his premium income. But when the surplus is ð3:2Þ
below −c1/β, the insurer cannot repay all her/his debts for her/his
premium income, then the insurer is no longer allowed to run her/his and, for u > b,
business. Then the modified surplus process Ub(t) is given by
2 2
8 σ ∂ M2 ðu; y; bÞ ∂M2 ðu; y; bÞ ∂M2 ðu; y; bÞ
< ðc2 þ γU b ðt ÞÞdt−dSðt Þ þ σdBðt Þ; U b ðt Þ > b; þ ðγu þ c2 Þ ¼ λM2 ðu; y; bÞ þ αy
2 ∂u2 ∂u ∂y
dU b ðt Þ ¼ c1 dt−dSðt Þ þ σdBðt Þ; 0 ≤ U b ðt Þ ≤ b; ð2:2Þ
: h u−b
ðc1 þ βU b ðt ÞÞdt−dSðt Þ þ σdBðt Þ; −c1 =β ≤ U b ðt Þ ≤ 0; u
−λ ∫ M2 ðu−x; y; bÞdF ðxÞ þ ∫ M1 ðu−x; y; bÞdF ðxÞ
0 u−b
where, c2 = c1 − ε is the net premium rate after dividend payments and
c
uþ 1 i
S(t) = ∑ nN(t)
=1Xn. þ∫
β
M3 ðu−x; y; bÞdF ðxÞ þ F ðu þ c1 =βÞ ; ð3:3Þ
u
We denote the absolute ruin time of the modified surplus process
Ub(t) by Tb, which is defined by
and, for − c1/β ≤ u ≤ 0,
T b ¼ inf ft ≥ 0 : U b ðt Þ ≤ −c1 =βg;
and Tb = ∞ if Ub(t) > −c1/β, for all t ≥ 0. Let α > 0 be the force of
interest valuation, then the present value of all dividends until Tb is σ 2 ∂2 M 3 ðu; y; bÞ ∂M 3 ðu; y; bÞ ∂M 3 ðu; y; bÞ
þ ðβu þ c1 Þ ¼ λM3 ðu; y; bÞ þ αy
defined by 2 ∂u2 ∂u ∂y
c
uþ β1
Du;b ¼ ∫ e
T b −αt
dDðt Þ: ð2:3Þ −λ ∫ M 3 ðu−x; y; bÞdF ðxÞ þ F ðu þ c1 =βÞ ; ð3:4Þ
0
0
W. Yu / Economic Modelling 31 (2013) 625–634 627
Proof. time t0. Thus by conditioning on the time and the amount of the
first claim, we obtain, when −c1/βb u b 0,
h i
(1) When 0 ≤ u ≤ b, consider a small time interval (0, t], where M 3 ðu; y; bÞ ¼ ð1−λt ÞE M 3 h1 ðu; t Þ þ σBðt Þ; ye−αt ; b
t(t > 0) is sufficiently small so that the surplus process will
h h ðu;t ÞþσBðtÞþc1
not reach b. In view of the strong Markov property of the sur- þ λt⋅E ∫01
−αt
M3 h1 ðu; t Þ þ σBðt Þ−x; ye ; b dF ðxÞ
β
h i ð3:10Þ
−αt
Mðu; y; bÞ ¼ E M U b ðt Þ; ye ; b þ oðt Þ: ð3:5Þ
By Taylor's expansion, we have
h i ∂M3 ðu; y; bÞ
By conditioning on the time and amount of the first claim and −αt
E M 3 h1 ðu; t Þ þ σBðt Þ; ye ; b ¼ M3 ðu; y; bÞ þ ðβu þ c1 Þt
∂u
whether the claim causes absolute ruin, we obtain
h i σ 2 t ∂2 M3 ðu; y; bÞ ∂M3 ðu; y; bÞ
−αt
þ −αyt þ oðt Þ:
M1 ðu; y; bÞ ¼ ð1−λt ÞE M1 u þ c1 t þ σBðt Þ; ye ; b 2 ∂u2 ∂y
h uþc tþσBðtÞ
ð3:11Þ
−αt
þ λt·E ∫
1
M1 u þ c1 t þ σBðt Þ−x; ye ; b dF ðxÞ
0
uþc1 tþσBðt Þþ
c1
þ∫
β −αt
M3 u þ c1 t þ σBðt Þ−x; ye ; b dF ðxÞ
Substituting Eq. (3.11) into Eq. (3.10), and then dividing both
uþc1 tþσBðt Þ
sides of Eq. (3.10) by t and letting t → 0, we get Eq. (3.4).
þF ðu þ c1 t þ σBðt Þ þ c1 =β Þ þ oðt Þ: ð3:6Þ
Theorem 3.2. M1(u, y;b), M2(u, y;b) and M3(u, y;b) satisfy
2
By Taylor's expansion, E[B(t)] = 0 and E[B (t)] = t, then we lim M2 ðu; y; bÞ ¼ e
εy=α
; ð3:12Þ
have u→∞
(4) For −c1/βb ub 0, letting τ0 be the time that the surplus reach Theorem 3.4. Vn1(u;b), Vn2(u;b) and Vn3(u;b) satisfy the following
0 for the first time from ub 0 and using the Markov property of boundary conditions
the surplus process {Ub(t), t≥0}, we have
′ ðu; bÞu¼b ¼ nV n−1;1 ðb; bÞ; n∈Nþ ;
V n1 ð3:26Þ
h i h i
u yDu;b u yDu;b
M 3 ðu; y; bÞ ¼ E I ðτ0 bT b Þe þ E Iðτ 0 ≥T b Þe
V n3 ð−c1 =β; bÞ ¼ 0; n∈Nþ ; ð3:27Þ
u T b −τ 0 −αt
¼ E Iðτ 0 bT b Þ exp y∫ e dDðt þ τ 0 Þ þ P ðτ0 ≥T b Þ
0
V n1 ð0þ; bÞ ¼ V n3 ð0−; bÞ; n∈Nþ ; ð3:28Þ
T b −αt
¼ Eu Iðτ 0 bT b Þ exp ye−ατ0 ∫ e dDðt Þ þ P ðτ 0 ≥T b Þ
τ0
X
∞
y
n
σ2
M i ðu; y; bÞ ¼ 1 þ V ni ðu; bÞ; i ¼ 1; 2; 3 ; ð3:22Þ Φ ″ ðu; bÞ þ c1 Φ ′1 ðu; bÞ ¼ ðλ þ α ÞΦ1 ðu; bÞ
n! 2 1" #
n¼1 u uþ
c1 ð4:2Þ
−λ ∫ Φ1 ðu−x; bÞdF ðxÞ þ ∫
β
Φ3 ðu−x; bÞdF ðxÞ þ AðuÞ ;
0 u
and equating the coefficients of yn(n ∈N) in Eqs. (3.2)–(3.4), we
have the following integro-differential equations.
and, for u > b,
Theorem 3.3. For 0 ≤ u ≤ b, σ2
Φ ″ ðu; bÞ þ ðγu þ c2 ÞΦ ′2 ðu; bÞ ¼ ðλ þ α ÞΦ2 ðu; bÞ
2 2h
σ2 −λ ∫
u−b u
Φ2 ðu−x; bÞdF ðxÞ þ ∫ Φ1 ðu−x; bÞdF ðxÞ
′ ðu; bÞ ¼ ðλ þ nα ÞV n1 ðu; bÞ
V ″ ðu; bÞ þ c1 V n1
2 n1 " # ð3:23Þ c1
0 u−b
i
u uþ
c1 uþ
þ∫
β
−λ ∫ V n1 ðu−x; bÞdF ðxÞ þ ∫
β
V n3 ðu−x; bÞdF ðxÞ ; Φ3 ðu−x; bÞdF ðxÞ þ AðuÞ ;
0 u u
ð4:3Þ
and, for u > b,
and, for − c1/β ≤ u ≤ 0,
2
σ
V ″ ðu; bÞ þ ðγu þ c2 ÞV n2
′ ðu; bÞ ¼ ðλ þ nα ÞV n2 ðu; bÞ σ2
2 n2" # Φ ″ ðu; bÞ þ ðβu þ c1 ÞΦ 3′ ðu; bÞ ¼ ðλ þ α ÞΦ3 ðu; bÞ
c
uþ 1 2 3" #
−λ ∫
u−b
V n2 ðu−x; bÞdF ðxÞ þ ∫
u
V n1 ðu−x; bÞdF ðxÞ þ ∫
β
V n3 ðu−x; bÞdF ðxÞ ; uþ
c1 ð4:4Þ
−λ ∫
0 u−b u β
Φ3 ðu−x; bÞdF ðxÞ þ AðuÞ ;
0
ð3:24Þ
and, for − c1/β ≤ u ≤ 0,
with boundary conditions
2
σ
c
uþ 1
′ ðu; bÞ ¼ ðλ þ nα ÞV n3 ðu; bÞ−λ∫ β V n3 ðu−x; bÞdF ðxÞ:
V ″ ðu; bÞ þ ðβu þ c1 ÞV n3
2 n3 0 lim Φ2 ðu; bÞ ¼ Φ2 ðuÞ; ð4:5Þ
b→∞
ð3:25Þ
Proof. The method is similar to Theorem 3.1. Φ1 ð0þ; bÞ ¼ Φ3 ð0−; bÞ; ð4:6Þ
W. Yu / Economic Modelling 31 (2013) 625–634 629
2 2
σ σ Substituting Eq. (4.15) into Eq. (4.14) yields Eq. (4.10).
Φ ″ ð0þ; bÞ þ c1 Φ1′ ð0þ; bÞ ¼ Φ ″ ð0−; bÞ þ c1 Φ 3′ ð0−; bÞ; ð4:9Þ
2 1 2 3 Similar to the proof of Eq. (4.10), we can obtain Eq. (4.11) and
Eq. (4.12).
∞
where AðuÞ ¼ ∫ ωðu; x−uÞdF ðxÞ, Ф2(u) is the Gerber–Shiu expected
uþcβ1 Remark. It is well known, L1(u), L2(u) and L3(u) are absolutely integra-
discounted penalty function. ble, and l1(u, b), l2(u, b) and l3(u, b) are all continuous. Then Ф1(u;b),
Ф2(u;b) and Ф3(u;b) can be approximated recursively by Picard's
Proof. The method is similar to Theorem 3.1. sequence, i.e.
u X
∞
u
Φ2 ðu; bÞ ¼ ∫ l2 ðu; vÞΦ2 ðv; bÞdv þ L2 ðuÞ; u > b; ð4:11Þ Φ2 ðu; bÞ ¼ L2 ðuÞ þ ∫b l2n ðu; yÞL2 ðyÞdy; u > b;
b
n¼1
u
Φ3 ðu; bÞ ¼ ∫ c l ðu; vÞΦ3 ðv; bÞdv þ L3 ðuÞ; −c1 =β ≤ u ≤ 0; ð4:12Þ u
− 1 3
β where, l21(u, y)=l2(u, y), l2n(u,y)=∫ l2(u,x)l2,n−1(x,y)dx, n=2, 3, …
y
where X
∞
u
Φ3 ðu; bÞ ¼ L3 ðuÞ þ ∫−c l3n ðu; yÞL3 ðyÞdy; −c1 =β ≤ u ≤ 0;
1 =β
2 h u i n¼1
l1 ðu; vÞ ¼ 2 −c1 þ ðλ þ α þ γ Þðu−vÞ−λ∫ F ðy−vÞdy ;
σ" !
v
#
u
2 2
where, l31(u, y)=l3(u, y), l3n(u,y)=∫y l3(u,x)l3,n−1(x,y)dx, n=2, 3, ....
2 σ uσ u
L1 ðuÞ ¼ 2 þ c1 u Φ1 ð0; bÞ þ Φ1′ ð0; bÞ−λ∫ W 1 ðyÞdy ; Therefore, at least in principle, if we can find Ф1(0;b), Ф1′(0;b),
σ 2 2 0
Ф2(b;b), Ф2′(b;b), Ф3(−c1/β;b) and Ф3′(−c1/β;b), we can get the
0 y
W 1 ðyÞ ¼ ∫ c Φ ðv; bÞðF ðy−vÞ−F ð−vÞÞdv þ ∫ AðvÞdv exact form of the solutions for Ф1(u;b), Ф2(u;b) and Ф3(u;b), and
− β1 3 0
can approximate Ф1(u;b), Ф2(u;b) and Ф3(u;b), recursively.
2 h u i
l2 ðu; vÞ ¼ −ðγb þ c2 Þ þ ðλ þ α Þðu−vÞ−λ∫ F ðy−vÞdy ;
σ2 v Example 4.1. (The Laplace transform of absolute ruin time) We
discuss a particular case of Ф(u, b) by setting ω(x, y) = 1 for all
2 u
L2 ðuÞ ¼ 1− 2 ðu−bÞðγb þ c2 Þ Φ2 ðb; bÞ þ ðu−bÞΦ′2 ðb; bÞ−λ∫ x > −c1/β, y > c1/β. Denote
σ b
h i
−αT b
W 2 ðyÞdy; φðu; bÞ ¼ E e U b ð0Þ ¼ u ;
b
W 2 ðyÞ ¼ ∫ Φ1 ðv; bÞðF ðy−vÞ−F ðb−vÞÞdv
0 which is the expected present value of a payment 1 at the time of
0 y ruin, and at the same time, the Laplace transform of absolute ruin
þ∫ c Φ ðv; bÞðF ðy−vÞ−F ðb−vÞÞdv þ ∫ AðvÞdv;
− β1 3 b time Tb. It is easy to see that the sample path of the φ(u;b) behaves
2 h u i differently with different initial surplus. Hence, for notational conve-
l3 ðu; vÞ ¼ 2 −ðc1 þ βvÞ þ ðλ þ β þ α Þðu−vÞ−λ∫ F ðy−vÞdy ;
σ v nience, we set
2λ u v 8
L3 ðuÞ ¼ Φ3 ð−c1 =β; bÞ þ Φ ′3 ð−c1 =β; bÞðu þ c1 =βÞ− 2 ∫ c1 ∫ c1 AðyÞdydv: < φ2 ðu; bÞ; u > b;
σ − − β β
φðu; bÞ ¼ φ1 ðu; bÞ; 0 ≤ u ≤ b; ð4:16Þ
:
φ3 ðu; bÞ; −c1 =β b u ≤ 0:
Proof. In Eq. (4.2), substituting u with v and integrating Eq. (4.2)
over (0, u) yields By similar derivation to Theorem 3.1, φ1(u;b), φ2(u;b) and φ3(u;b)
satisfy the following second-order integro-differential equations:
σ
2
σ
2 For 0 ≤ u ≤ b,
Φ ′ ðu; bÞ þ c1 Φ1 ðu; bÞ ¼ Φ ′ ð0; bÞ þ c1 Φ1 ð0; bÞ
2 1 2 1
σ2
u
ð4:13Þ φ ″ ðu; bÞ þ c1 φ 1′ ðu; bÞ ¼ ðλ þ α Þφ1 ðu; bÞ
þ∫0 Φ1 ðv; bÞðλ þ α−λF ðu−vÞÞdv 2 1" #
u uþ
c1
u
−λ ∫ φ1 ðu−x; bÞdF ðxÞ þ ∫
β
0
−λ∫−c1 Φ3 ðv; bÞðF ðu−vÞ−F ð−vÞÞdv−λ∫ AðvÞdv: φ3 ðu−x; bÞdF ðxÞ þ F ðu þ c1 =βÞ ;
β 0 0 u
ð4:17Þ
In Eq. (4.13), substituting u with y and integrating Eq. (4.13) over
(0, u) yields and, for u > b,
" #
σ2 σ2 σ2 σ2
φ ″ ðu; bÞ þ ðγu þ c2 Þφ 2′ ðu; bÞ ¼ ðλ þ α Þφ2 ðu; bÞ
u
Φ1 ðu; bÞ ¼ Φ1 ð0; bÞ−c1 ∫ Φ1 ðy; bÞdy þ Φ ′1 ð0; bÞ þ c1 Φ1 ð0; bÞ u
2 2 0 2 2 2h
u−b u
u h y i u −λ ∫ φ2 ðu−x; bÞdF ðxÞ þ ∫ φ2 ðu−x; bÞdF ðxÞ ð4:18Þ
0 u−b
þ∫ ∫ Φ1 ðv; bÞðλ þ α−λF ðy−vÞÞdv dy−λ∫ W 1 ðyÞdy:
0 0 0 uþ
c1 i
þ∫
β
2
σ σ2 σ2
φ ″ ðu; bÞ þ ðβu þ c1 Þφ ′3 ðu; bÞ ¼ ðλ þ α Þφ3 ðu; bÞ φ ″ ðb−; bÞ þ c1 φ 1′ ðb−; bÞ ¼ φ ″ ðbþ; bÞ þ ðγb þ c2 Þφ ′2 ðbþ; bÞ; ð4:23Þ
2 3" # 2 1 2 2
uþ
c1 ð4:19Þ
−λ ∫
β
φ3 ðu−x; bÞdF ðxÞ þ F ðu þ c1 =βÞ ;
0 σ2 σ2
φ ″1 ð0þ; bÞ þ c1 φ ′1 ð0þ; bÞ ¼ φ ″ ð0−; bÞ þ c1 φ ′3 ð0−; bÞ; ð4:24Þ
2 2 3
with boundary conditions
φ3 ð−c1 =β; bÞ ¼ 1; ð4:25Þ
lim φ2 ðu; bÞ ¼ 0; ð4:20Þ
u→∞
″ 2α
φ1 ð0þ; bÞ ¼ φ3 ð0−; bÞ; ð4:21Þ φ 3 ð−c1 =β þ 0; bÞ ¼ : ð4:26Þ
σ2
In this section, we derive the explicit expressions of Vn(u, b) in the case of α = 0. We assume that F(x) = 1 − e −x/μ, x > 0, μ > 0, namely, the
claim size distribution F is an exponential distribution with mean μ.
In view of Eqs. (3.23)–(3.25) and the exponential density function, we obtain
σ2 λ u u 0 x
′ ðu; bÞ ¼ λV n1 ðu; bÞ− e− ∫ V n1 ðx; bÞe dx þ ∫ c V n3 ðx; bÞe dx ;
x
″ ðu; bÞ þ c1 Vn1
V n1 μ μ μ
ð5:1Þ
2 μ 0 − 1
for 0 ≤ u ≤ b, and
2
σ λ u u b x 0
′ ðu; bÞ ¼ λV n2 ðu; bÞ− e− ∫ V n2 ðx; bÞe dx þ ∫ V n1 ðx; bÞe dx þ ∫ c V n3 ðx; bÞe dx ;
x xx
″ ðu; bÞ þ ðγu þ c2 ÞV n2
V n2 μ μ μ μ
ð5:2Þ
2 μ b 0 − β
1
σ2 λ u u
′ ðu; bÞ ¼ λV n3 ðu; bÞ− e−μ ∫ c1 V n3 ðx; bÞeμ dx:
x
V ″ ðu; bÞ þ ðβu þ c1 ÞV n3 ð5:3Þ
2 n3 μ −β
d 1
Applying the operator þ on Eqs. (5.1)–(5.3), respectively, and then rearranging, we have
du μ
for 0 ≤ u ≤ b, and
!
σ 2 000 σ2 ″ γu þ c2 ′
V ðu; bÞ þ γu þ c2 þ V n2 ðu; bÞ þ γ−λ þ Vn2 ðu; bÞ ¼ 0; ð5:5Þ
2 n2 2μ μ
where ξn1, ξn2 and ξn3 are arbitrary constants, δ1 and δ2 are the two real roots of the following equation
2
δ þ ρδ þ η ¼ 0; ð5:8Þ
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
−ρ þ ρ2 −4ρη −ρ− ρ2 −4ρη ð5:9Þ
δ1 ¼ ; δ2 ¼ :
2 2
W. Yu / Economic Modelling 31 (2013) 625–634 631
For Eq. (5.5), let u = z − Δ1, Δ1 = c2 /γ − σ 2 / (2μγ) and Vn2′(u, b) = e−z/μm(z). For Eq. (5.6), let u = z − Δ2, Δ2 = c1 / β− σ2 / (2μβ) and Vn3′(u, b) =
e −z/βm(z), then Eqs. (5.5) and (5.6) are changed to
σ2
m ″ðzÞ þ γzm ′ðzÞ þ ðγ−λÞmðzÞ ¼ 0; ð5:10Þ
2
σ2
m ″ðzÞ þ βzm ′ðzÞ þ ðβ−λÞmðzÞ ¼ 0: ð5:11Þ
2
Further, let z 2 = −σ 2x / γ in Eq. (5.10), z 2 = −σ 2x / β in Eq. (5.11) and m(z) = H(x). Then Eqs. (5.10) and (5.11) are converted to the following
Kummer's confluent hypergeometric equations
1 1 λ
xH ″ðxÞ þ −x H ′ðxÞ− − H ðxÞ ¼ 0; ð5:12Þ
2 2 2γ
1 1 λ
xH ″ðxÞ þ −x H ′ðxÞ− − HðxÞ ¼ 0: ð5:13Þ
2 2 2β
( ) !
u γðu þ Δ1 Þ2 λ 1 γ ðu þ Δ1 Þ2
g 1 ðuÞ ¼ exp − − U ; ; ;
μ σ2 2γ 2 σ2
( ) !
2 2
u γðu þ Δ1 Þ 1 λ 3 γ ðu þ Δ1 Þ
g 2 ðuÞ ¼ ðu þ Δ1 Þ exp − − D þ ; ; ;
μ σ2 2 2γ 2 σ2
( ) !
2 2
u βðu þ Δ2 Þ λ 1 βðu þ Δ2 Þ
q1 ðuÞ ¼ exp − − U ; ; ;
μ σ2 2β 2 σ2
( ) !
2 2
u βðu þ Δ2 Þ 1 λ 3 βðu þ Δ2 Þ
q2 ðuÞ ¼ ðu þ Δ2 Þ exp − − D þ ; ; ;
μ σ2 2 2β 2 σ2
and D(l, m;x) and U(l, m;x) are the confluent hypergeometric functions of the first and second kinds, respectively.
Hence, by Eqs. (3.27) and (3.33), we have
∞ ε n ∞ ∞ ε n ε n
′ ðy; bÞdy þ
V n2 ðu; bÞ ¼ −∫ V n2 ¼ −kn1 ∫ g 1 ðyÞdy−kn2 ∫ g 2 ðyÞdy þ ¼ −kn1 G1 ðuÞ−kn2 G2 ðuÞ þ ; u≥b;
u α u u α α
ð5:16Þ
u u u
V n3 ðu; bÞ ¼ ∫ ′ ðy; bÞdy ¼ k3 ∫
V n3 c q ðyÞdy þ k4 ∫ c1 q2 ðyÞdy ¼ kn3 Q 1 ðuÞ þ kn4 Q 2 ðuÞ; −c1 =β ≤ u ≤ 0; ð5:17Þ
− β1 1
c
− β1 − β
∞ u
where Gi(u) = ∫u gi(y)dy, i = 1, 2, Q j ðuÞ ¼ ∫ q ðyÞdy, j = 1, 2.
−cβ1 j
Now, we determine the coefficients ξn1, ξn2, ξn3, kn1, kn2, kn3, kn4 by the boundary conditions Eqs. (3.26)–(3.32), Eq. (5.1) and V01(b, b) = 1.
Using the formulas of
d l d
Dðl; m; xÞ ¼ Dðl þ 1; m þ 1; xÞ and U ðl; m; xÞ ¼ −lDðl þ 1; m þ 1; xÞ;
dx m dx
632 W. Yu / Economic Modelling 31 (2013) 625–634
8
>
>
> ξn1 δ1 eδ1 b þ ξn2 δ2 eδ2 b ¼ ρn ;
>
>
>
> kn3 Q 1″ð−c1 =βÞ þ kn4 Q ″2 ð−c1 =βÞ ¼ 0;
>
>
>
>
>
> ξn1 þ ξn2 þ ξn3 ¼ kn3 Q 1 ð0Þ þ kn4 Q 2 ð0Þ;
>
>
>
>
> ξn1 eδ1 b þ ξn2 eδ2 b þ ξn3 ¼ −kn1 G1 ðbÞ−kn1 G2 ðbÞ þ τ n ;
>
>
>
>
> σ2
>
> 2 2 σ2
>
>
< 2 ξ n1 δ1 þ ξ n2 δ2 þ c 1 ðξ n1 δ1 þ ξn2 δ2 Þ ¼ k Q ″ ð0Þ þ kn4 Q ″2 ð0Þ
2 n3 1
þ c1 kn3 Q 1′ ð0Þ þ kn4 Q 2′ ð0Þ ; ð5:18Þ
>
> 2 σ2
>
> σ
>
> 2 δ b 2 δ b δ b
ξn1 δ1 e 1 þ ξn2 δ2 e 2 þ c1 ξn1 δ1 e 1 þ ξn2 δ2 e 2 ¼
δ b
−kn1 G ″1 ðbÞ−kn2 G ″2 ðbÞ
>
>
>
> 2
2
>
> þ ðγb þ c2 Þ −kn1 G1′ ðbÞ−kn2 G ′2 ðbÞ ;
>
>
>
>
>σ
>
2
>
>
2 2
ξn1 δ1 þ ξn2 δ2 þ c1 ðξn1 δ1 þ ξn2 δ2 Þ ¼ λ½ξn1 þ ξn2 þ ξn3
>
>
>
> 2
>
> λ 0 λ 0
>
x x
: −kn3 ∫−c1 Q 1 ðxÞe dx−kn4 ∫−c Q 2 ðxÞe dx:
μ
1
μ
μ β μ β
Let
δ b δ b
a1 ¼ δ1 e 1 ; a2 ¼ δ2 e 2 ; a3 ¼ Q ″1 ð−c1 =βÞ; a4 ¼ Q ″2 ð−c1 =βÞ; a5 ¼ Q 1 ð0Þ; a6 ¼ Q 2 ð0Þ;
2 2 2 2
δ1 σ δ σ
a7 ¼ eδ1 b ; a8 ¼ eδ2 b ; a9 ¼ G1 ðbÞ; a10 ¼ G2 ðbÞ; a11 ¼ þ c1 δ1 ; a12 ¼ 2 þ c1 δ2 ;
2 2
2 2
σ σ 1 2 2 δ1 b δ b
a13 ¼ ″ ′
Q ð0Þ þ c1 Q 1 ð0Þ; a14 ¼ ″ ′
Q ð0Þ þ c1 Q 2 ð0Þ; a15 ¼ σ δ1 e þ c1 δ1 e 1 ;
2 1 2 2 2
2
1 2 2 δb δ b σ
a16 ¼ σ δ2 e 2 þ c1 δ2 e 2 ; a17 ¼ G ″ ðbÞ þ ðγb þ c2 ÞG1′ ðbÞ;
2 2 1
σ2 1 2 2 1 2 2
a18 ¼ G ″ ðbÞ þ ðγb þ c2 ÞG2′ ðbÞ; a19 ¼ σ δ1 þ c1 δ1 −λ; a20 ¼ σ δ2 þ c1 δ2 −λ;
2 2 2 2
λ 0 x λ 0 x
a21 ¼ λ; a22 ¼ ∫c Q 1 ðxÞe dx; a23 ¼ ∫ c Q 2 ðxÞe dx;
μ μ
μ β
1
μ − 1
β
a1 a a ρ
A1 ¼ 1−a7 − ð1−a8 Þ; A2 ¼ ða5 þ a9 Þ; A3 ¼ 3 6 ; A4 ¼ −τn − n ð1−a8 Þ;
a2 a4 a2
a1 a3
A5 ¼ a19 þ a21 − ða20 þ a21 Þ; A6 ¼ a5 a21 ; A7 ¼ a22 − ða23 −a6 a21 Þ;
a2 a4
ρn a1 a21 a3 a14 a
A8 ¼ − ða20 þ a21 Þ; A9 ¼ a11 − ; A10 ¼ −a13 ; A11 ¼ − 12 ρn ;
a2 a2 a4 a2
a1 a16 a16 n
A12 ¼ a15 − ; A13 ¼ − ρ ; ρ ¼ nV n−1;1 ðb; bÞ; τ n ¼ ðε=α Þ :
a2 a2 n n
a18 ½A2 ðA8 A9 −A5 A11 Þ−A6 ðA4 A9 −A1 A11 Þ−a10 ½A6 ðA9 A13 −A11 A12 Þ þ a17 ðA8 A9 −A5 A11 Þ
kn3 ¼
a18 ½A2 ðA7 A9 −A5 A10 Þ−A6 ðA3 A9 −A1 A10 Þ−a10 ½A17 ðA7 A9 −A5 A10 Þ−A6 A10 A12
1 A2 ðA8 A9 −A5 A11 Þ−A6 ðA4 A9 −A1 A11 Þ
kn2 ¼
a10 A 6 A9
1 A2 ðA7 A9 −A5 A10 Þ−A6 ðA3 A9 −A1 A10 Þ
− kn3 ;
a10 A6 A9
a3
kn4 ¼ − k ;
a4 n3
A8 A9 −A5 A11 A7 A9 −A5 A10
kn1 ¼ − þ kn3 ;
A6 A9 A6 A9
A11 A10
ξn1 ¼ − k ;
A9 A9 n3
ρ a A A
ξn3 ¼ n − 1 11 − 10 kn3 ;
a2 a2 A9 A9
A A −A5 A10 A A −A5 A11 a A A
ξn3 ¼ a5 7 9 kn3 − 8 9 −a6 3 kn3 − 11 − 10 kn3 ;
A6 A9 A6 A9 a4 A9 A9
ρ a A A
− n þ 1 11 − 10 kn3 :
a2 a2 A9 A9
W. Yu / Economic Modelling 31 (2013) 625–634 633
Note that V0,1(b;b) = 1, setting n = 1 in Eq. (5.18), then ρ1 = 1 and τ1 = ε/α, so ξ11, ξ12, ξ13, k11, k12, k13, and k14 are determined and we obtain
the explicit expressions of V11(u;b), V12(u;b) and V13(u;b). Let n = 2 and in view of the V11(b;b), we have ρ2 = 2 V11(b;b) and τ 2 = ε 2 / α 2.
Substituting ρ2 and τ2 into Eq. (5.18), then ξ21, ξ22, ξ23, k21, k22, k23, and k24 are determined and we get the explicit expressions of V21(u;b),
V22(u;b) and V23(u;b). Repeating the above steps, we can obtain the explicit expressions of Vn1(u;b), Vn2(u;b) and Vn3(u;b).
References
Albrecher, H., Thonhausera, S., 2008. Optimal dividend strategies for a risk process
under force of interest. Insurance: Mathematics and Economics 43 (1), 134–149.
Bai, X.D., Song, L.X., 2012. Asymptotic behavior of random time absolute ruin probability
with D∩L tailed and conditionally independent claim sizes. Statistics & Probability
Letters 82 (9), 1718–1726.
Cai, J., 2007. On the time value of absolute ruin with debit interest. Advances in Applied
Probability 39 (2), 343–359.
Cai, J., Yang, H.L., 2005. Ruin in the perturbed compound Poisson risk process under
interest force. Advances in Applied Probability 37 (3), 819–835.
Dassios, A., Embrechts, P., 1989. Martingales and insurance risk. Stochastic Models 5,
149–166.
Gao, S., Liu, Z.M., 2010. The perturbed compound Poisson risk model with constant in-
terest and a threshold dividend strategy. Journal of Computational and Applied
Mathematics 233 (9), 2181–2188.
Gao, H.L., Yin, C.C., 2008. The perturbed Sparre Andersen model with a threshold divi-
dend strategy. Journal of Computational and Applied Mathematics 220 (1–2),
394–408.
Gerber, H.U., 1970. An extension of the renewal equation and its application in the
collective theory of risk. Skandinavisk Aktuarietidskrift 205–210.
Li, M.M., Liu, Z.M., 2012. Regulated absolute ruin problem with interest structure and
linear dividend barrier. Economic Modelling 29 (5), 1786–1792.
Liu, D.H., Liu, Z.M., 2011. The perturbed compound Poisson risk model with linear dividend
barrier. Journal of Computational and Applied Mathematics 235 (8), 2357–2363.
Fig. 4. The effect of γ on V1(u;b).
Slater, L.J., 1960. Confluent Hypergeometric Functions. Cambridge University Press.
Tsai, C.C.L., Willmot, G.E., 2002. On the moments of the surplus process perturbed by
diffusion. Insurance: Mathematics and Economics 31 (3), 327–350.
Wan, N., 2007. Dividend payments with a threshold strategy in the compound Poisson
risk model perturbed by diffusion. Insurance: Mathematics and Economics 40 (3),
509–532.
Wang, G.J., Wu, R., 2008. The expected discounted penalty function for the perturbed
compound Poisson risk process with constant interest. Insurance: Mathematics
and Economics 42 (1), 59–64.
Wang, C.W., Yin, C.C., Li, E.Q., 2010. On the classical risk model with credit and debit in-
terests under absolute ruin. Statistics & Probability Letters 80 (15), 427–436.
Yang, H., Zhang, Z.M., Lan, C.M., 2008. On the time value of absolute ruin for a multi-
layer compound Poisson model under interest force. Statistics & Probability Letters
78 (13), 1835–1845.
Yuan, H.L., Hu, Y.J., 2008. Absolute ruin in the compound Poisson risk model with
constant dividend barrier. Statistics & Probability Letters 78 (14), 2086–2094.
Zhang, S.Q., 2012. Impulse stochastic control for the optimization of the dividend pay-
ments of the compound Poisson risk model perturbed by diffusion. Stochastic
Analysis and Applications 30 (4), 642–661.
Zhang, Z.M., Yang, H., Li, S.M., 2010. The perturbed compound Poisson risk model with
two-sided jumps. Journal of Computational and Applied Mathematics 233 (8),
1773–1784.
Zhang, Z.M., Yang, H.L., Yang, H., 2011. On the absolute ruin in a map risk model with
debit interest. Advances in Applied Probability 43 (1), 77–96.
Zhu, J.X., Yang, H.L., 2008. Estimates for the absolute ruin probability in the compound
Poisson risk model with credit and debit interest. Journal of Applied Probability 45
(3), 818–830.