数学物理学报, 2020, 40(2): 501-514 doi:

论文

非时齐复合Poisson风险模型的破产特征量分析

邓迎春1, 李满1, 黄娅2, 周杰明,1

On the Analysis of Ruin-Related Quantities in the Nonhomogeneous Compound Poisson Risk Model

Deng Yingchun1, Li Man1, Huang Ya2, Zhou Jieming,1

通讯作者: 周杰明, E-mail: jmzhou@hunnu.edu.cn

收稿日期: 2018-12-24  

基金资助: 湖南省哲学社会科学基金.  17YBA290

Received: 2018-12-24  

Fund supported: the Philosophy and Social Science Fund of Hunan Province.  17YBA290

摘要

该文将经典风险模型推广到非时齐复合Poisson风险模型.首先,运用经典方法和时变方法,计算了该模型下的破产特征量,且得到了更新方程的解析表达式.其次,定义了时变后相应模型的一个广义的Gerber-Shiu函数,验证了时变方法对非时齐Poisson风险模型的有效性.最后,当单次索赔量服从指数分布时,计算了相应的破产概率和Gerber-Shiu函数.

关键词: 破产概率 ; 时变方法 ; 非时齐Poisson过程 ; Gerber-Shiu函数 ; 更新方程

Abstract

In this paper, the classical risk model is extended to nonhomogeneous compound Poisson risk model. Firstly, both the classical method and the time-varying method are used to calculate the ruin-related quantities for this model, and the analytical expression of the renewal equation is obtained. Secondly, for the time-varying model, the generalized Gerber-Shiu function is defined, which is to verify the effectiveness of the time-varying method for the nonhomogeneous compound Poisson risk model. Finally, when each claim follows an exponentially distribution, the corresponding ruin probability and Gerber-Shiu function are calculated.

Keywords: Ruin probability ; Time-varying method ; Nonhomogeneous Poisson process ; Gerber-Shiu function ; Renewal equation

PDF (467KB) 元数据 多维度评价 相关文章 导出 EndNote| Ris| Bibtex  收藏本文

本文引用格式

邓迎春, 李满, 黄娅, 周杰明. 非时齐复合Poisson风险模型的破产特征量分析. 数学物理学报[J], 2020, 40(2): 501-514 doi:

Deng Yingchun, Li Man, Huang Ya, Zhou Jieming. On the Analysis of Ruin-Related Quantities in the Nonhomogeneous Compound Poisson Risk Model. Acta Mathematica Scientia[J], 2020, 40(2): 501-514 doi:

1 模型的建立

保险公司的风险无处不在,如通货膨胀风险,承保风险,利率的改变,投资风险和无偿付能力的风险.而无偿付能力风险是保险公司的最根本风险,因此不同风险模型的破产特征量研究一直深受保险公司经营者和许多学者的关注.自从一个世纪以前, Lundberg和Cramér提出了经典的风险模型[1]之后,破产概率问题被很多学者广泛地研究,如文献[2-20]. Yuen和Guo研究了复合二项模型中时间相关索赔的破产概率[4]. Gerber等研究了带有投资的跳跃扩散模型中的绝对破产概率[5]. Li等在文献[6-7]中分别研究了MAP风险模型和Markov-modulated风险模型的一些破产概率问题. Liu和Li等研究了更新风险模型的无穷时间上的破产概率[8-9].尹传存研究了关于破产概率的一个局部定理[10].刘艳和胡亦钧在文献[11]中研究了马氏环境下带扰动的Cox相关的风险模型破产概率的上界估计. Gerber和Shiu在文献[19]中提出了Gerber-Shiu函数.自此之后, Gerber-Shiu问题受到了很多学者的关注和研究,参见文献[21-26].

经典模型中单位时间的保费收入为常数,单次索赔量是独立同分布的随机变量.这个模型的主要特点是独立,齐次,常数保费.然而,由于保险业的发展,投资市场的不稳定,以及投保者诚信度等因素的影响,保险公司的保费收入具有波动性.根据市场的扩大和收缩,保险公司合理地调整保费收取方式,更符合现代保险公司的实际情况.因此,保费收入依然按照线性增长的假设就具有很大的局限性.对于这个问题,在文献[27]中, Zhang等考虑了一个具有变保费率的干扰复合Poisson模型,运用Laplace变换的方法研究了Gerber-Shiu函数.基于复合泊松过程, Li等研究了带有随机利率的风险模型及其在精算中的应用[28].韦晓等基于变保费率扰动的风险模型研究了有限时间的破产概率和大偏差问题[29].同时,由于自然灾害(如地震,海啸,干旱)等不确定性的因素,虽然持续时间不长,但是很大程度上影响着保险公司的索赔,因此风险模型的齐次性假设在应用于实际方面有很大的局限性.因此,非齐次风险模型被一些学者研究,例如:Picard在文献[30]中研究了保险公司的非齐次风险模型. Kartashov考虑了一个具有连续时间的经典更新方程的非时齐干扰和具有变保费率的风险模型[31]. Kartashov研究了具有半轴连续时间的非时齐扰动问题[32]. Blaževičius等分别在文献[33-34]中研究了非齐次索赔情形下有限时间和无限时间的破产概率问题.对于非齐次和变保费率问题,更多的研究结果可参考文献[35-37].

根据这些文献,具有非线性保费收入的非时齐风险模型与经典风险模型相比,更能反映真实的保险市场.在此基础上,我们考虑保费收入$C(t)$是一个非线性函数,索赔到达过程为非时齐Poisson过程的一类连续时间风险模型,盈余过程表示为

$ \label{1.2}U(t)=u+C(t)-S(t), $

其中$u$为初始盈余, $C(t)$表示保险公司的保费收入. $\{S(t)=\sum\limits_{i=1}^{N(t)}Y_{i}, t\geq0\}$表示保险公司的总索赔过程,其中单次索赔量$\{Y_{i}, ~i=1, 2, 3, \cdots\}$是一个独立同分布的非负随机变量序列,记为$Y\sim F_{Y}(y)$,则对于单次索赔量$Y$,均值为$E(Y)=\mu_{1}<\infty$,且矩母函数为$M_{Y}(r)=E(e^{rY})=\int_{0}^{\infty}e^{ry}{\rm d}F_{Y}(y)$. $\{N(t), t\geq0\}$是索赔到达速率为$\lambda(t)>0(t\geq0)$的非时齐Poisson过程,假设与$\{Y_{i}, i=1, 2, 3, \cdots\}$相互独立.令$m(t)=\int_{0}^{t}\lambda(s){\rm d}s$,则$m(t)$表示非时齐Poisson过程$\{N(t), t\geq0\}$的均值函数(累计强度函数).

由于保险公司收取的保费要满足预期的赔付,即$C(t)>E[S(t)]$,定义$\theta$为保险公司的相对安全负荷,假设保险公司根据期望保费原理收取保费,则

$ \label{1.3} C(t)=\mu_{1}(1+\theta)m(t), $

其中$\theta>0$, $m(t)$严格单调递增,且假设$\lim\limits_{t\rightarrow\infty}m(t)=+\infty$.

2 非时齐复合Poisson风险模型的破产概率

这一部分考虑非时齐复合Poisson风险模型的破产概率.定义破产时间为$T=\inf\{t\geq0, $$U(t)<0\}$ (如果$T=\infty$,则破产不发生).破产概率为

首先,考虑非时齐复合Poisson风险模型调节系数的存在性.由

$ \label{2.1}E[e^{rS(t)-rC(t)}]=1 $

$E[e^{rN(t)}]=M_{N(t)}(r)=e^{m(t)(e^{r}-1)}$,可得

$ \label{2.2} E[e^{rS(t)}]=e^{m(t)(M_{Y}(r)-1)}. $

把(2.2)式代入(2.1)式,可得

再由(1.2)式,可得方程

$ \label{2.3}M_{Y}(r)=1+\mu_{1}(1+\theta)r. $

若关于$r$的方程(2.3)有正数解,我们称它的最小正数解$r=R$为调节系数.接下来的一个定理说明调节系数$R$的存在性.

定理2.1  在非时齐复合Poisson风险模型(1.1)中,若$M_{Y}(r)$$(0, \gamma)$上有定义且连续,并满足

则关于$r$的方程(2.3)在$(0, \gamma)$上存在唯一的正根.

  令$g(r)=M_{Y}(r)-1-\mu_{1}(1+\theta)r$,只要能证明$g(r)$满足下面的条件(1)-(4),则定理得证. (1) $g(0)=0$; (2) $g'(0)<0$; (3) $\lim\limits_{r\rightarrow\gamma_{-}}g(r)=+\infty$; (4) $g''(r)>0~(0 <r <+\infty)$.

(1)和(2)可证.

下证(3). (ⅰ)当$\gamma<+\infty$时,由$\lim\limits_{r\rightarrow\gamma_{-}}M_{Y}(r)=+\infty$,可得

(ⅱ)当$\gamma=+\infty$时,由$M_{Y}(r)=1+\mu_{1}r+\frac{\mu_{2}}{2}r^{2}+\cdots>\frac{\mu_{2}}{2}r^{2}$,从而当$r\rightarrow+\infty$时,有

由(ⅰ), (ⅱ)可证(3).

考虑到$g''(r)=M''_{Y}(r)=E(Y^{2}e^{rY})>0$, (4)可证.从而定理2.1可证.

接下来,运用时变方法得到非时齐复合Poisson风险模型的破产概率.

定理2.2   在模型(1.1)中,对于$u\geq0$,有

$\begin{eqnarray}\label{2.4} \psi(u)=\frac{e^{-Ru}}{E[e^{-RU(T)}|T<\infty]}. \end{eqnarray}$

  采用时变方法将模型(1.1)转化为经典风险模型的形式.令$\tilde{U}(t)=U[m^{-1}(t)]$, $\tilde{N}(t)=N[m^{-1}(t)]$,由(1.2)式可知

$ \label{2.5} \tilde{C}(t)=C[m^{-1}(t)]=\mu_{1}(1+\theta)t. $

$\tilde{S}(t)=\sum\limits_{i=1}^{\tilde{N}(t)}Y_{i}$,由文献[40]中的定理3.5可知$\{\tilde{N}(t), t\geq0\}$是参数为1的齐次Poisson过程.所以,模型(1.1)变成了经典风险模型的形式,其盈余过程为

$ \label{2.6}\tilde{U}(t)=u+\mu_{1}(1+\theta)t-\tilde{S}(t). $

我们令

$ \label{2.7}\tilde{T}=\inf\{t\geq0, \tilde{U}(t)<0\}. $

由时间变换易知$\tilde{T}=m(T)$,因此, $T<\infty$当且仅当$\tilde{T}<\infty$.

由经典风险模型的破产概率,可得

$ \label{2.8} \tilde{\psi}(u)=\frac{e^{-\tilde{R}u}}{E[e^{-\tilde{R}\tilde{U}(\tilde{T})}|\tilde{T}<\infty]}, $

其中$\tilde{R}$是方程

$ \label{2.9}1+\mu_{1}(1+\theta)r=M_{Y}(r) $

的最小正数解.比较调节系数方程(2.3)和(2.9), $\tilde{R}=R$,即经过时间变换之后,调节系数不变.同时,有

因此, $E[e^{-\tilde{R}\tilde{U}(\tilde{T})}|\tilde{T}<\infty]=E[e^{-RU(T)}|T<\infty], $从而得到破产概率表达式(2.4).

注2.1  非时齐复合Poisson过程也可以运用经典模型的方法得到破产概率.对于$t>0, $$ R>0$,运用全期望公式可得

$ \begin{eqnarray}\label{2.10} E[e^{-RU(t)}]=E[e^{-RU(t)}|T\leq t]P(T\leq t)+E[e^{-RU(t)}|T>t]P(T>t). \end{eqnarray} $

由模型(1.1),可知(2.10)式左侧为

$ \begin{eqnarray}\label{2.11}e^{-Ru-RC(t)+m(t)[M_{Y}(R)-1]}=e^{-Ru}. \end{eqnarray} $

我们设$T$的分布函数为$G(t)$, $\mathcal{F}_{⊔}$为由$U(r)$生成的$\sigma$代数, $\mathcal{F}_{⊔}=\sigma\{\mathcal{U}(▽), ▽\leq ⊔\}$.在(2.10)式右侧第一项中,

再由(2.11)式,从而(2.10)式为

$ \begin{eqnarray}\label{2.12}e^{-Ru}=E[e^{-RU(T)}|T\leq t]P(T\leq t)+E[e^{-RU(t)}|T>t]P(T>t). \end{eqnarray} $

$t\rightarrow\infty$时, (2.12)式右侧第一项收敛于$E[e^{-RU(T)}|T<\infty]\psi(u)$.

现在只要能证明当$t\rightarrow\infty$时(2.12)式右侧第二项收敛于$0$,即可证定理$2.2$.为此,记$\alpha=\mu_{1}\theta, \beta^{2}=\mu _{2}$,从而有

因为$m(t)$严格单调递增,并且由$\lim\limits_{t\rightarrow\infty}m(t)=+\infty$,所以只要当$t$充分大, $u+\alpha m(t)-\beta (m(t))^{\frac{2}{3}}>0$.从而(2.12)式右侧第二项可以变形,经过整理可以得到

由切比雪夫不等式,有

从而有

$ \label{2.13}E[e^{-RU(t)}|T>t]P(T>t)\leq m(t)^{-\frac{1}{3}}+e^{-R(u+\alpha m(t)-\beta (m(t))^{\frac{2}{3}})}. $

$t\rightarrow\infty$时,从而(2.13)式收敛于$0$.定理$2.2$得证.

一般来说,直接计算出(2.4)式的分母非常困难.但是,当$T<\infty$时,有$U(T)<0$,总有$E[e^{-RU(T)}|T<\infty]>1$,所以推论2.1中的不等式限定了破产概率的上界.

推论2.1    $\psi(u)\leq e^{-Ru}.$

推论2.2   如果单次索赔量$Y$有上界$M(M>0)$,当$0\leq Y\leq M$时, $\psi(u)\geq e^{-R(u+M)}$;当$M\ll u$时, $\psi(u)\approx e^{-Ru}.$

例2.1   当单次赔付量$Y$服从参数为$\beta$的指数分布时,由于$\mu_{1}=\frac{1}{\beta}$,则

进一步得到

$R$代入破产概率公式为

3   非时齐复合Poisson风险模型的Gerber-Shiu函数

3.1    Gerber-Shiu函数的计算

随机变量$U(T-)$表示破产前的瞬时盈余, $|U(T)|$表示破产时的赤字,破产时间$T=\inf\{t\geq0, U(t)<0\}$(如果$T=\infty$,则破产不发生). $f(u_{1}, u_{2}, t|u)$$U(T-)$, $|U(T)|$$T$的联合概率密度函数.破产概率为

其中$U(0)=u\geq0$表示初始盈余,根据(1.2)式, $C(t)>E[S(t)]$, $\psi(u) <1$, $f(u_{1}, u_{2}, t|u)$是瑕疵概率密度函数,如果$u_{1}>u+C(t)$,则$f(u_{1}, u_{2}, t|u)=0$.

对于$\delta\geq0$,定义

其中$\delta$是折现因子或者称为破产时间$T$的Laplace项.定义Gerber-Shiu函数为

$\begin{eqnarray}\label{3.0}\phi(u):&=&E[e^{-\delta T}\omega(U(T-), |U(T)|)I_{\{T<\infty\}}|U(0)=u]\nonumber\\&=&\int_{0}^{\infty}\int_{0}^{\infty}\int_{0}^{\infty}\omega(u_{1}, u_{2})e^{-\delta t}f(u_{1}, u_{2}, t|u){\rm d}t{\rm d}u_{1}{\rm d}u_{2}, \end{eqnarray}$

其中$I_{\{T<\infty\}}$是示性函数. $\omega: [0, \infty)\times[0, \infty)\rightarrow[0, \infty)$是关于$u_{1}, u_{2}$的一个可测二元函数.

记第一次索赔发生时间为$\tau=\inf\{t>0, N(t)>0\}$.此时, $\tau$的密度函数是$f_{\tau}(t)=\lambda(t)e^{-m(t)}, t>0$.如果$h>0$足够小,在时间间隔$[0, h]$上考虑如下三种情况.

(1)  直到时间$h$,没有索赔;

(2)  直到时间$h$,第一次索赔发生,但是保险公司没有破产;

(3)  直到时间$h$,第一次索赔发生,且索赔量$y>u+C(t)$,保险公司破产.

根据上面的三种情况,由全概率公式, (3.1)式可写为

$\begin{eqnarray}\label{3.1} \phi(u)&=&e^{-(\delta h+m(h))}\phi(u+C(h))+\int_{0}^{h}\bigg[\int_{0}^{u+C(t)}\phi(u+C(t)-y)f_{Y}(y){\rm d}y\bigg]\lambda(t) e^{-(\delta t+m(t))}{\rm d}t \nonumber\\ &&+\int_{0}^{h}\bigg[\int_{u+C(t)}^{\infty}\omega(u+C(t), y-u-C(t))f_{Y}(y){\rm d}y\bigg]\lambda(t) e^{-(\delta t+m(t))}{\rm d}t. \end{eqnarray} $

根据$m'(t)=\lambda(t)$,关于$h$求导,可得

$\lambda(h)_{h\rightarrow0}=\lambda(0)$.根据(1.2)式, $C(h)=\mu_{1}(1+\theta)m(h)=\mu_{1}(1+\theta)\int_{0}^{h}\lambda(s){\rm d}s$,因此$\frac{{\rm d}C(h)}{{\rm d}h}=\mu_{1}(1+\theta)\lambda(h)$.$h\rightarrow0$时, $C(h)\rightarrow0$,可得

$ \begin{eqnarray}\label{3.2} 0&=&-(\delta+\lambda(0))\phi(u)+\mu_{1}(1+\theta)\lambda(0)\phi'(u)+\lambda(0)\int_{0}^{u}\phi(u-y)f_{Y}(y){\rm d}y \nonumber\\ &&+\lambda(0)\int_{u}^{\infty}\omega(u, y-u)f_{Y}(y){\rm d}y. \end{eqnarray} $

引理3.1  令$l_{1}(\xi)=\delta+\lambda(0)-\mu_{1}(1+\theta)\lambda(0)\xi$,则$l_{1}(\xi)=\lambda(0)\hat{f}_{Y}(\xi)$有唯一的非负根,其中$\hat{f}_{Y}(\xi)=\int_{0}^{\infty}e^{-\xi y}f_{Y}(y){\rm d}y$$f_{Y}(\xi)$的Laplace变换.

  根据$\hat{f}'_{Y}(\xi)=-\int_{0}^{\infty}e^{-\xi y}yf_{Y}(y){\rm d}y<0$$\hat{f}''_{Y}(\xi)=\int_{0}^{\infty}e^{-\xi y}y^{2}f_{Y}(y){\rm d}y>0, $可知函数$\hat{f}_{Y}(\xi)$是递减的凸函数.

接下来,考虑方程$l_{1}(\xi)=\lambda(0)\hat{f}_{Y}(\xi)$.根据$C(t)>E[S(t)]$,可得$l_{1}'(0)=-\mu_{1}(1+\theta)\lambda(0)<-\lambda(0)\mu_{1}=-\lambda(0)\hat{f}'_{Y}(0)$, $l_{1}(0)=\delta+\lambda(0)\geq\lambda(0)=\lambda(0)\hat{f}_{Y}(0)$.因此, $l_{1}(\xi)=\lambda(0)\hat{f}_{Y}(\xi)$有唯一的非负根$\xi_{1}$,记为$\xi_{1}=\rho$.引理$3.1$得证.

注3.1   如果单次索赔量的密度函数$f_{Y}(\xi)$充分正则,则方程$l_{1}(\xi)=\lambda(0)\hat{f}_{Y}(\xi)$另外有一负根$\xi_{2}$,通常记作$-R$.如果$\delta=0$,则$R$是风险模型的调节系数.

通过变量替换,积分因子的方法(具体过程见附录A)和引理$3.1$,可得更新方程

$ \label{3.13}\phi=\phi\ast k+h, $

其中

$ \label{3.11}k(y)=\frac{1}{\mu_{1}(1+\theta)}\int_{0}^{\infty}e^{-\rho z}f_{Y}(y+z){\rm d}z, $

$ \label{3.12}h(y)=\frac{1}{\mu_{1}(1+\theta)}\int_{0}^{\infty}e^{-\rho z}\int_{0}^{\infty}\omega(y+z, x)f_{Y}(y+z+x){\rm d}x{\rm d}z, $

这里$\rho$为方程$l_{1}(\xi)=\lambda(0)\hat{f}_{Y}(\xi)$的非负根.

根据Laplace变换得

如果给定单次索赔量的密度函数$f_{Y}(\cdot)$,则利用Laplace逆变换可以得到$\phi(u)$的显式表达式(见例3.1).

3.2 时变后风险模型的Gerber-Shiu函数及推广

非时齐复合Poisson风险模型,运用时变的方法,根据(1.2)式和$m(t)=\int_{0}^{t}\lambda(s){\rm d}s$,索赔到达过程$\tilde{N}(t)=N[m^{-1}(t)], t\geq0\}$是强度为$\tilde{\lambda}=1$的齐次Poisson过程(文献[40,定理3.5]).记第一次索赔发生时间为$\tilde{\tau}=\inf\{t>0, \tilde{N}(t)>0\}$, $\tilde{\tau}$的密度函数是$\tilde{f}_{\tilde{\tau}}(t)=e^{-t}, t>0$,破产时间$\tilde{T}=\inf\{t\geq0, \tilde{U}(t)<0\}$.$\tilde{C}(t)=C[m^{-1}(t)]=\mu_{1}(1+\theta)t, $可得时变后的盈余过程(2.6)为

定义Gerber-Shiu函数为

$ \begin{eqnarray}\label{3.16} \Phi(u):&=&E[e^{-\tilde{\delta} \tilde{T}}\omega(U(\tilde{T}-), |U(\tilde{T})|)I_{\{\tilde{T}<\infty\}}|U(0)=u], \end{eqnarray} $

其中$I_{\{\tilde{T}<\infty\}}$是示性函数, $\omega(U(\tilde{T}-), |U(\tilde{T})|)=\omega(U(T-), |U(T)|)$是一个可测二元函数.

如果$\tilde{h}>0$足够小,在时间间隔$[0, \tilde{h}]$上, (3.7)式为

$\begin{eqnarray}\label{3.17}\Phi(u)&=&e^{-(\tilde{\delta} \tilde{h}+\tilde{h})}\Phi(u+\mu_{1}(1+\theta)\tilde{h})\nonumber\\&&+\int_{0}^{\tilde{h}}e^{-(\tilde{\delta} t+t)}\bigg[\int_{0}^{u+\mu_{1}(1+\theta)t}\Phi(u+\mu_{1}(1+\theta)t-y)f_{Y}(y){\rm d}y\bigg]{\rm d}t\nonumber\\&&+\int_{0}^{\tilde{h}}e^{-(\tilde{\delta} t+t)}\bigg[\int_{u+\mu_{1}(1+\theta)t}^{\infty}\omega(u+\mu_{1}(1+\theta)t, y-u-\mu_{1}(1+\theta)t)f_{Y}(y){\rm d}y\bigg]{\rm d}t.\end{eqnarray}$

关于$\tilde{h}$求导可得

$\tilde{h}\rightarrow0$,则

$ \label{3.18}0=-(\tilde{\delta}+1)\Phi(u)+\mu_{1}(1+\theta)\Phi'(u)+\int_{0}^{u}\Phi(u-y)f_{Y}(y){\rm d}y+\int_{u}^{\infty}\omega(u, y-u)f_{Y}(y){\rm d}y. $

引理3.2  令$l_{2}(\tilde{\xi})=\tilde{\delta}+1-\mu_{1}(1+\theta)\tilde{\xi}$,则$l_{2}(\tilde{\xi})=\hat{f}_{Y}(\tilde{\xi})$有唯一的非负根,其中$\hat{f}_{Y}(\tilde{\xi})=\int_{0}^{\infty}e^{-\tilde{\xi} y}f_{Y}(y){\rm d}y$$f_{Y}(\tilde{\xi})$的Laplace变换.

  证明与引理3.1相似,此处省略.

通过变量替换,以及积分因子的方法,由引理$3.2$,方程(3.9)可得更新方程

$\begin{eqnarray}\label{3.19}\Phi=\Phi\ast \tilde{k}+\tilde{h}, \end{eqnarray}$

其中$\tilde{k}, \tilde{h}$

$\begin{eqnarray}\label{3.20}\tilde{k}(y)=\frac{1}{\mu_{1}(1+\theta)}\int_{0}^{\infty}e^{-\tilde{\rho} z}f_{Y}(y+z){\rm d}z, \end{eqnarray}$

$\begin{eqnarray}\label{3.21}\tilde{h}(y)=\frac{1}{\mu_{1}(1+\theta)}\int_{0}^{\infty}e^{-\tilde{\rho} z}\int_{0}^{\infty}\omega(y+z, x)f_{Y}(y+z+x){\rm d}x{\rm d}z, \end{eqnarray}$

其中$\tilde{\rho}$为方程$l_{2}(\tilde{\xi})=\hat{f}_{Y}(\tilde{\xi})$的非负根.

根据Laplace变换得

如果给定单次索赔量的密度函数$f_{Y}(\cdot)$,则利用Laplace逆变换可以得到$\Phi(u)$的显式表达式(见例3.1).

对于时变后的盈余过程,定义一个广义的Gerber-Shiu函数为

$\begin{eqnarray}\label{3.24}\Phi^{\gamma}(u):&=&E[e^{-\int_{0}^{\tilde{T}}\gamma(s){\rm d}s}\omega(U(\tilde{T}-), |U(\tilde{T})|)I_{\{\tilde{T}<\infty\}}|U(0)=u], \end{eqnarray}$

其中$I_{\{\tilde{T}<\infty\}}$是示性函数, $\omega(U(\tilde{T}-), |U(\tilde{T})|)=\omega(U(T-), |U(T)|)$是一个可测二元函数, $\gamma(t)>0$对于任意的$t\geq0$.

如果$\tilde{h}>0$足够小,在时间间隔$[0, \tilde{h}]$上, (3.13)式为

$\begin{eqnarray}\label{3.25}\Phi^{\gamma}(u)&=&e^{-(\int_{0}^{\tilde{h}}\gamma(s){\rm d}s+\tilde{h})}\Phi^{\gamma}(u+\mu_{1}(1+\theta)\tilde{h})\nonumber\\&&+\int_{0}^{\tilde{h}}e^{-(\int_{0}^{t}\gamma(s){\rm d}s+t)}\bigg[\int_{0}^{u+\mu_{1}(1+\theta)t}\Phi^{\gamma}(u+\mu_{1}(1+\theta)t-y)f_{Y}(y){\rm d}y\bigg]{\rm d}t\nonumber\\&&+\int_{0}^{\tilde{h}}e^{-(\int_{0}^{t}\gamma(s){\rm d}s+t)}\bigg[\int_{u+\mu_{1}(1+\theta)t}^{\infty}\omega(u+\mu_{1}(1+\theta)t, y-u-\mu_{1}(1+\theta)t)f_{Y}(y){\rm d}y\bigg]{\rm d}t.\\\end{eqnarray}$

关于$\tilde{h}$求导可得

$\begin{eqnarray}0&=&-(\gamma(\tilde{h})+1)e^{-(\int_{0}^{\tilde{h}}\gamma(s){\rm d}s+\tilde{h})}\Phi^{\gamma}(u+\mu_{1}(1+\theta)\tilde{h})\nonumber\\&&+\mu_{1}(1+\theta)e^{-(\int_{0}^{\tilde{h}}\gamma(s){\rm d}s+\tilde{h})}(\Phi^{\gamma}(u+\mu_{1}(1+\theta)\tilde{h}))'\nonumber\\&&+e^{-(\int_{0}^{\tilde{h}}\gamma(s){\rm d}s+\tilde{h})}\int_{0}^{u+\mu_{1}(1+\theta)\tilde{h}}\Phi^{\gamma}(u+\mu_{1}(1+\theta)\tilde{h}-y)f_{Y}(y){\rm d}y\nonumber \\&&+e^{-(\int_{0}^{\tilde{h}}\gamma(s){\rm d}s+\tilde{h})}\int_{u+\mu_{1}(1+\theta)\tilde{h}}^{\infty}\omega(u+\mu_{1}(1+\theta)\tilde{h}, y-u-\mu_{1}(1+\theta)\tilde{h})f_{Y}(y){\rm d}y.\end{eqnarray}$

$\tilde{h}\rightarrow0$,记$\gamma(\tilde{h})|_{\tilde{h}=0}=\gamma(0)$,可得

$\begin{eqnarray}\label{3.26}0&=&-(\gamma(0)+1)\Phi^{\gamma}(u)+\mu_{1}(1+\theta)(\Phi^{\gamma}(u))'+\int_{0}^{u}\Phi^{\gamma}(u-y)f_{Y}(y){\rm d}y\nonumber\\&&+\int_{u}^{\infty}\omega(u, y-u)f_{Y}(y){\rm d}y.\end{eqnarray}$

引理3.3   令$l_{3}(\bar{\xi})=\gamma(0)+1-\mu_{1}(1+\theta)\bar{\xi}$,则$l_{3}(\bar{\xi})=\hat{f}_{Y}(\bar{\xi})$有唯一的非负根,其中$\hat{f}_{Y}(\bar{\xi})=\int_{0}^{\infty}e^{-\bar{\xi} y}f_{Y}(y){\rm d}y$$f_{Y}(\bar{\xi})$的Laplace变换.

  证明过程与引理3.1和3.2相似,此处省略.

通过变量替换,由积分因子的方法,运用引理$3.3$,方程(3.16)的更新方程为

其中

这里$\bar{\rho}$为方程$l_{3}(\bar{\xi})=\hat{f}_{Y}(\bar{\xi})$的非负根.

根据Laplace变换得

注3.2   非时齐复合Poisson过程,对于广义的Gerber-Shiu函数$\Phi^{\gamma}(u)$,有

$1.$如果$\int_{0}^{\tilde{T}}\gamma(s){\rm d}s=\delta m^{-1}(\tilde{T})$, $\gamma(0)=\delta (m^{-1}(\tilde{h}))'|_{\tilde{h}=0}=\frac{\delta}{\lambda(0)}, $则方程(3.16)为

两边同时乘以$\lambda(0)$,可得

与方程(3.3)比较,此时$\Phi^{\gamma}(u)=\phi(u)$.

$2.$如果$\int_{0}^{\tilde{T}}\gamma(s){\rm d}s=\tilde{\delta}\tilde{T}$,方程(3.16)为

与方程(3.9)比较,此时$\Phi^{\gamma}(u)=\Phi(u)$.

注3.3   如果$C(t)=ct$,则模型(1.1)是齐次复合Poisson过程

其中$c$是保费率. $\{Y_{i}, ~i=1, 2, 3, \cdots\}$是独立同分布非负随机变量序列,分布函数为$F_{Y}(y)$, $Y_{i}$是第$i$次索赔量.索赔到达过程$\{N(t), t\geq0\}$是强度为$\lambda>0$的齐次Poisson过程, $\frac{{\rm d}C(t)}{{\rm d}t}=c$,则方程(3.3)为

这是文献[19]的结果.

例3.1   若单次索赔量$Y$服从参数为$\beta$的指数分布.考虑两种特殊情况.

$1.$如果$\delta\neq0$,可得

根据Laplace变换和Laplace逆变换,可得

$2.$如果$\delta=0$,则$\rho=\tilde{\rho}=\xi_{1}=0$,和$-R=-\tilde{R}=\xi_{2}=-\frac{\theta\beta}{1+\theta}$.

(1) $\omega(u_{1}, u_{2})=\omega_{2}(u_{2})=e^{-r_{2}u_{2}}, r_{2}\geq0$,

根据Laplace变换和Laplace逆变换,可得

(2) $\omega(u_{1}, u_{2})=1$,

$\Phi(u)=\phi(u)$是破产概率公式,根据Laplace变换和Laplace逆变换,可得

4   总结和展望

本文运用时变方法和经典方法,讨论了非时齐复合Poisson风险模型的破产概率,模型考虑保险公司的保费收入呈非线性增长,索赔到达过程为非齐次复合Poisson过程,这使得保险公司根据市场的波动及时合理调整保费收入方式和索赔方式. Poisson过程也被应用于网络模型和排队论中,为此我们有以下展望.

(1)大部分网络模型考虑节点数量等时间间隔增长[38],这有局限性.在随机网络中,网络中节点数量可能呈非线性增长,节点的到达过程与时间和权重有关,所以考虑关于非齐次复合Poisson网络模型,对网络中信息冗余,规避拥堵有现实意义.

(2)排队论中[39],顾客到达时间是随机的,不一定按固定时间到达,接待顾客数量与时间有关,即服从非齐次Poisson过程.考虑到对每个顾客的服务时间,非齐次复合Poisson过程更符合现实,可以有效的提高服务效率和服务水平,提高顾客满意度.

附录A 非时齐方程的经典Gerber-Shiu函数求解

为了得到Gerber-Shiu函数的表达式,由方程(3.3):

通过变量替换$x=y-u$,则

$\begin{eqnarray}\label{3.3}0&=&-(\delta+\lambda(0))\phi(u)+\mu_{1}(1+\theta)\lambda(0)\phi'(u)+\lambda(0)\int_{0}^{u}\phi(u-y)f_{Y}(y){\rm d}y \nonumber\\&&+\lambda(0)\int_{0}^{\infty}\omega(u,x)f_{Y}(u+x){\rm d}x.\end{eqnarray}$

由积分因子的方法,令$\phi_{\rho}(u)=e^{-\rho u}\phi(u), \rho\geq0$, $\phi'_{\rho}(u)=-\rho e^{-\rho u}\phi(u)+e^{-\rho u}\phi'(u)$. (A.1)式两端乘以$e^{-\rho u}$,则

$\begin{eqnarray}\label{3.4}\mu_{1}(1+\theta)\lambda(0)\phi_{\rho}'(u)&=&(\delta+\lambda(0)-\mu_{1}(1+\theta)\lambda(0)\rho)\phi_{\rho}(u)\nonumber\\&&-\lambda(0)\int_{0}^{u}\phi_{\rho}(u-y)e^{-\rho y}f_{Y}(y){\rm d}y \nonumber\\&&-\lambda(0)e^{-\rho u}\int_{0}^{\infty}\omega(u,x)f_{Y}(u+x){\rm d}x.\end{eqnarray}$

为了解方程(A.2),运用引理3.1, $l_{1}(\xi)=\lambda(0)\hat{f}_{Y}(\xi)$有唯一的非负根$\xi_{1}$,记$\rho=\xi_{1}$.则方程(A.2)就变为

$\begin{eqnarray}\label{3.5}\mu_{1}(1+\theta)\lambda(0)\phi_{\rho}'(u)&=&\lambda(0)\hat{f}_{Y}(\rho)\phi_{\rho}(u)-\lambda(0)\int_{0}^{u}\phi_{\rho}(u-y)e^{-\rho y}f_{Y}(y){\rm d}y \nonumber\\&&-\lambda(0)e^{-\rho u}\int_{0}^{\infty}\omega(u,x)f_{Y}(u+x){\rm d}x\nonumber\\&=&\lambda(0)\bigg[\hat{f}_{Y}(\rho)\phi_{\rho}(u)-\int_{0}^{u}\phi_{\rho}(y)e^{-\rho (u-y)}f_{Y}(u-y){\rm d}y \nonumber\\&&-e^{-\rho u}\int_{0}^{\infty}\omega(u,x)f_{Y}(u+x){\rm d}x\bigg].\end{eqnarray}$

对于$z>0$, (A.3)式的两端对$u$$0$$z$积分.两边除以$\lambda(0)$,则

$\begin{eqnarray}\label{3.6}\mu_{1}(1+\theta)[\phi_{\rho}(z)-\phi_{\rho}(0)]&=&\hat{f}_{Y}(\rho)\int_{0}^{z}\phi_{\rho}(u){\rm d}u-\int_{0}^{z}\int_{0}^{u}\phi_{\rho}(y)e^{-\rho (u-y)}f_{Y}(u-y){\rm d}y{\rm d}u \nonumber\\&&-\int_{0}^{z}e^{-\rho u}\int_{0}^{\infty}\omega(u,x)f_{Y}(u+x){\rm d}x{\rm d}u\nonumber\\&=&\hat{f}_{Y}(\rho)\int_{0}^{z}\phi_{\rho}(u){\rm d}u-\int_{0}^{z}\int_{y}^{z}e^{-\rho (u-y)}f_{Y}(u-y){\rm d}u\phi_{\rho}(y){\rm d}y \nonumber\\&&-\int_{0}^{z}e^{-\rho u}\int_{0}^{\infty}\omega(u,x)f_{Y}(u+x){\rm d}x{\rm d}u\nonumber\\&=&\int_{0}^{z}\phi_{\rho}(y)\int_{z-y}^{\infty}e^{-\rho x}f_{Y}(x){\rm d}x{\rm d}y \nonumber\\&&-\int_{0}^{z}e^{-\rho u}\int_{0}^{\infty}\omega(u,x)f_{Y}(u+x){\rm d}x{\rm d}u.\end{eqnarray}$

对于$z\rightarrow\infty$, (A.4)式两边的第一项都消失,可得

因此

$ \label{3.7}\phi_{\rho}(z)=\frac{1}{\mu_{1}(1+\theta)}\bigg[\int_{0}^{z}\phi_{\rho}(y)\int_{z-y}^{\infty}e^{-\rho x}f_{Y}(x){\rm d}x{\rm d}y+\int_{z}^{\infty}e^{-\rho u}\int_{0}^{\infty}\omega(u,x)f_{Y}(u+x){\rm d}x{\rm d}u\bigg]. $

定义

更新方程(A.5)可重新写为

$ \label{3.10}\phi_{\rho}=\phi_{\rho}\ast k_{\rho}+h_{\rho}, $

$\phi_{\rho}(u)=e^{-\rho u}\phi(u)$,即$\phi(y)=e^{\rho y}\phi_{\rho}(y)$.定义

则方程(A.6)可以写为$ \phi=\phi\ast k+h, $根据Laplace变换得

参考文献

Lundberg F .

On the theory of reinsurance

Transactions Ⅵ International Congress of Actuaries, 1909, 1, 877- 948

URL     [本文引用: 1]

Cramér H. On the Mathematical Theory of Risk. Stockholm: Skandia Jubilee Volume, 1930

URL     [本文引用: 1]

Cramér H. Collective Risk Theory: Asurvey of the Theory from the Point of View of the Theory of Stochastic Processes. Stockholm: Nordiska Bokhandeln, 1955

URL    

Yuen K C , Guo J Y .

Ruin probabilities for time-correlated claims in the compound binomial model

Insurance Mathematics and Economics, 2001, 29 (1): 47- 57

DOI:10.1016/S0167-6687(01)00071-3      [本文引用: 1]

Gerber U , Yang H L .

Absolute ruin probabilities in a jump diffusion model with investment

North American Actuarial Journal, 2007, 11 (3): 159- 169

DOI:10.1080/10920277.2007.10597474      [本文引用: 1]

Li J C , Dickson D C M , Li S M .

Some ruin problems for the MAP risk model

Insurance Mathematics and Economics, 2015, 65, 1- 8

DOI:10.1016/j.insmatheco.2015.08.001      [本文引用: 1]

Li J C , Dickson D C M , Li S M .

Analysis of some ruin-related quantities in a Markov-modulated risk model

Stochastic Models, 2016, 32 (3): 351- 365

DOI:10.1080/15326349.2015.1121399      [本文引用: 1]

Liu R F , Wang D , Peng J .

Infinite-time ruin probability of a renewal risk model with exponential Levy process investment and dependent claims and inter-arrival times

Journal of Industrial and Management Optimization, 2017, 13 (2): 995- 1007

DOI:10.3934/jimo.2016058      [本文引用: 1]

Li J Z .

The infinite-time ruin probability for a bidimensional renewal risk model with constant force of interest and dependent claims

Communications in Statistics-Theory and Methods, 2017, 46 (4): 1959- 1971

DOI:10.1080/03610926.2015.1030428      [本文引用: 1]

尹传存.

关于破产概率的一个局部定理

中国科学(A辑), 2004, 34 (2): 192- 202

URL     [本文引用: 1]

Yin C C .

A local theorem about the ruin probability

Science in China Series A, 2004, 34 (2): 192- 202

URL     [本文引用: 1]

刘艳, 胡亦钧.

马氏环境下带扰动的Cox相关的风险模型破产概率的上界估计

数学年刊, 2004, 25 (5): 579- 586

URL     [本文引用: 1]

Liu Y , Hu Y J .

Upper bound estimation on the ruin probability for a cox correlated risk model disturbed by diffusion in a markovian environment

Chinese Journal of Contemporary Mathematies, 2004, 25 (5): 579- 586

URL     [本文引用: 1]

Konstantinides D , Tang Q H , Tsitsiashvili G .

Estimates for the ruin probability in the classical risk model with constant interest force in the presence of heavy tails

Insurance Mathematics and Economics, 2002, 31 (3): 447- 460

DOI:10.1016/S0167-6687(02)00189-0     

唐启鹤.

重尾索赔下关于破产概率的一个等价式

中国科学(A辑), 2002, 32 (3): 260- 266

URL    

Tang Q H .

An equivalent formula of the ruin probability under a heavy tail claim

Science China Series A, 2002, 32 (3): 260- 266

URL    

郭风龙, 王定成, 张维.

考虑随机投资收益和单边线性索赔的时间依赖更新风险模型的破产概率

中国科学(A辑), 2016, 46A (3): 321- 350

URL    

Guo F L , Wang D C , Zhang W .

Ruin probability of renewal risk model with stochastic investment returns and one-sided linear time-dependent claims

Science China Series A, 2016, 46A (3): 321- 350

URL    

Wang L .

Asymptotic behavior of finite-time ruin probability in a by-claim risk model with constant interest rate

Journal of Mathematics and Statistics, 2014, 10 (3): 339- 357

URL    

Xie J H , Zou W .

Dividend barrier and ruin problems for a risk model with delayed claims

Communications in Statistics-Theory and Methods, 2017, 46 (14): 7063- 7084

DOI:10.1080/03610926.2016.1143010     

Constantinescu C , Samorodnitsky G , Zhu W .

Ruin probabilities in classical risk models with gamma claims

Scandinavian Actuarial Journal, 2018, 2018 (7): 555- 575

DOI:10.1080/03461238.2017.1402817     

Oshime T , Shimizu Y .

Parametric inference for ruin probability in the classical risk model

Statistics and Probability Letters, 2018, 133, 28- 37

DOI:10.1016/j.spl.2017.09.020     

Gerber U , Shiu S .

On the time value of ruin

North American Actuarial Journal, 1998, 2 (1): 48- 72

DOI:10.1080/10920277.1998.10595671      [本文引用: 2]

Lin X S , Willmot G E .

Analysis of a defective renewal equation arising in ruin theory

Insurance Mathematics and Economics, 1999, 25 (1): 63- 84

DOI:10.1016/S0167-6687(99)00026-8      [本文引用: 1]

Lin X S , Willmot G E , Drekic S .

The classical risk model with a constant dividend barrier:analysis of the Gerber-Shiu discounted penalty function

Insurance Mathematics and Economics, 2004, 33 (3): 551- 566

URL     [本文引用: 1]

Albrecher H , Boxma O J .

A ruin model with dependence between claim sizes and claim intervals

Insurance Mathematics and Economics, 2004, 35 (2): 245- 254

DOI:10.1016/j.insmatheco.2003.09.009     

Albrecher H , Boxma O J .

On the discounted penalty function in a Markov-dependent risk model

Insurance Mathematics and Economics, 2005, 37 (3): 650- 672

DOI:10.1016/j.insmatheco.2005.06.007     

Albrecher H , Cheung E C K , Thonhauser S .

Randomized observation periods for the compound Poisson risk model:the discounted penalty function

Scandinavian Actuarial Journal, 2013, 2013 (6): 424- 452

DOI:10.1080/03461238.2011.624686     

Zhang Z M , Su W .

A new efficient method for estimating the Gerber-Shiu function in the classical risk model

Scandinavian Actuarial Journal, 2018, 2018 (5): 426- 449

DOI:10.1080/03461238.2017.1371068     

Willmot G E, Woo J K. Gerber-Shiu analysis in the classical Poisson risk model//Willmot G E, Woo J K. Surplus Analysis of Sparre Andersen Insurance Risk Processes. New York: Springer, 2017: 45-59

[本文引用: 1]

Zhang Z M , Yang Y , Liu C L .

On a perturbed compound Poisson model with varying premium rates

Journal of Industrial and Management Optimization, 2017, 13 (2): 721- 736

DOI:10.3934/jimo.2016043      [本文引用: 1]

Li S L , Yin C C , Zhao X , Dai H S .

Stochastic interest model based on compound Poisson process and applications in actuarial science

Mathematical Problems in Engineering, 2017, 2, 1- 8

URL     [本文引用: 1]

韦晓, 于金酉, 胡亦钧.

变保费率扰动风险模型的有限时间破产概率和大偏差

数学物理学报, 2007, 27A (4): 616- 623

URL     [本文引用: 1]

Wei X , Yu J Y , Hu Y J .

Large deviations and finite time ruin probability for perturbed risk model with variable premium rate(in Chinese)

Acta Mathematica Scientia, 2007, 27A (4): 616- 623

URL     [本文引用: 1]

Picard P .

A nonhomogeneous risk model for insurance

Computers and Mathematics with Applications, 2006, 51 (2): 325- 334

DOI:10.1016/j.camwa.2005.11.005      [本文引用: 1]

Kartashov M .

Inhomogeneous perturbations of a renewal equation and the Cramér-Lundberg theorem for a risk process with variable premium rates

Theory of Probability and Mathematical Statistics, 2009, 78, 61- 73

DOI:10.1090/S0094-9000-09-00762-5      [本文引用: 1]

Kartashov M .

Boundedness, limits, and stability of solutions of a perturbation of a nonhomogeneous renewal equation on a semiaxis

Theory of Probability and Mathematical Statistics, 2010, 81, 71- 83

DOI:10.1090/S0094-9000-2010-00811-8      [本文引用: 1]

Blaževičius K , Bieliauskieně E , Šiaulys J .

Finite-time ruin probability in the inhomogeneous claim case

Lithuanian Mathematical Journal, 2010, 50 (3): 260- 270

DOI:10.1007/s10986-010-9084-2      [本文引用: 1]

Bieliauskieně E , Šiaulys J .

Infinite time ruin probability in inhomogeneous claims case

Lietuvos Matematikos Rinkinys, 2010, 51, 352- 356

URL     [本文引用: 1]

Andrulyt I M , Bernackait E , Kievinait D , Šiaulys J .

A Lundberg-type inequality for an inhomogeneous renewal risk model

Modern Stochastics:Theory and Applications, 2015, 2 (2): 173- 184

URL     [本文引用: 1]

Abaurrea J , Asín J , Cebrián A C .

Modeling and projecting the occurrence of bivariate extreme heat events using a non-homogeneous common Poisson shock process

Stochastic Environmental Research and Risk Assessment, 2015, 29 (1): 309- 322

URL    

Bernackaitě E , Šiaulys J .

The finite-time ruin probability for an inhomogeneous renewal risk model

Journal of Industrial and Management Optimization, 2017, 13 (1): 207- 222

DOI:10.3934/jimo.2016012      [本文引用: 1]

张波, 张景肖. 应用随机过程. 北京: 清华大学出版社, 2004: 42- 45

[本文引用: 1]

Zhang B , Zhang J X . Applying Stochastic Processes. Beijing: Tsinghua University Press, 2004: 42- 45

[本文引用: 1]

杨琴, 陈云.

基于泊松过程的供应链复杂网络模型

系统工程, 2012, 30 (9): 57- 62

URL     [本文引用: 1]

Yang Q , Chen Y .

Supply chain complex network model based on poisson process

Systems Engineering, 2012, 30 (9): 57- 62

URL     [本文引用: 1]

Albrecher H , Boxma O O , Essifi R R , Kuijstermans R .

A queuing model with a randomized depletion of inventory

Probability in the Engineering and Informational Sciences, 2017, 31 (1): 43- 59

DOI:10.1017/S0269964816000322      [本文引用: 2]

/