 Research
 Open Access
 Published:
On a discrete risk model with delayed claims and a randomized dividend strategy
Advances in Difference Equations volume 2015, Article number: 284 (2015)
Abstract
In this paper, we consider a discrete risk model with delayed claims and randomized dividend strategy. The expected discounted dividends before ruin are studied. Difference equations for the expected discounted dividends are derived and solved.
Introduction
In the compound binomial model, the surplus process for an insurance company is described as follows:
where u is a nonnegative integer denoting the initial surplus, \(\{X_{t}\}_{t=1}^{\infty}\) is a sequence of i.i.d. random variables denoting the individual claim sizes. Let X denote the generic version of \(X_{t}\)’s and define their common probability function by
The Bernoulli sequence \(\{ \xi_{t} \}_{t=1}^{\infty}\) is used to denote claim occurrence such that \(\xi_{t}=1\) if a claim occurs in the time period \((t1, t]\), and \(\xi_{t}=0\) if no claim occurs in the time period \((t1, t]\). It is assumed that
where \(0< q<1\).
The compound binomial risk model has been studied by many authors, for example, Gerber [1], Shiu [2], Willmot [3] and Dickson [4]. Recently, some extensions have been made on this model. Yang et al. [5] study the ruin probabilities in a discrete Markov risk model. Yang and Zhang [6] consider a discrete renewal risk model with twosided jumps. Gerber et al. [7] modify the compound binomial risk model by dividend payments. Chen et al. [8] study the survival probabilities in a discrete semiMarkov risk model.
In reality, insurance claims may be delayed due to various reasons. The compound binomial risk model can be extended by involving two types of insurance claims, namely the main claims and the byclaims. We use \(\{X_{t}\}\) and \(\{\xi_{t}\}\) to denote the main claim sizes and the indicators for their occurrences, respectively. We assume that each main claim induces a byclaim. The byclaim and its associated main claim may occur simultaneously with probability p (\(0< p<1\)), or the occurrence of the byclaim may be delayed to the next time period with probability \(1p\). Let \(\{Y_{t}\}_{t=1}^{\infty}\) be an i.i.d. sequence to denote byclaim sizes and let Y be a generic variable of the byclaim. Define the probability mass function of Y by
Let \(S_{t}^{X}\) and \(S_{t}^{Y}\) be, respectively, the total main claims and byclaims up to time t, where the superscripts X and Y are used to indicate main claim and byclaim, respectively. Then the delayed risk model \(U^{\infty}=\{U_{t}^{\infty}\}_{t=0}^{\infty}\) can be described as follows: \(U^{\infty}_{0}=u\) and for \(t=1,2,\ldots\) ,
For the study on risk models with delayed claims, we refer the interested readers to Yuen and Guo [9], Yuen et al. [10] and Xiao and Guo [11].
Recently, risk models with randomized dividend strategy have received a lot of attention in the literature. Albrecher et al. [12] study the expected discounted dividends in the compound Poisson model with randomized dividend decision times. Avanzi et al. [13] consider a periodic dividend strategy in the dual model. Zhang [14] considers a perturbed compound Poisson risk model with a randomized dividend strategy. Zhang and Cheung [15] investigate the randomized dividend strategy in a Markov additive risk model. For the discrete risk model, Tan and Yang [16] propose a randomized dividend strategy by modifying the compound binomial model. In their model, whenever the surplus process is larger or equal to a barrier b (a positive integer), the company will possibly pay dividends at the end of the next period. He and Yang [17] consider a compound binomial model, where dividends are randomly paid to shareholders and policyholders. In this paper, we employ a randomized dividend strategy to modify the delayed risk model \(U^{\infty}\), and denote the modified model by \(U^{b}=\{U_{t}^{b}\}_{t=1}^{\infty}\). As in Tan and Yang [16], for \(t=0,1,\ldots\) , we assume that whenever \(U_{t}^{b}\geq b\), a dividend of size \(\eta_{t+1}\) is possibly paid at the beginning of the \((t+1)\)th period \((t, t+1]\), where \(\{\eta_{t}\}_{t=1}^{\infty}\) is a Bernoulli sequence such that
Now the total dividends paid up to time t can be expressed as
Starting from the initial surplus \(U_{0}^{b}=u\), we have for \(t=1,2,\ldots\) ,
Associated with the model \(U^{b}\), we define the ruin time by
where \(\tau^{b}=\infty\) if \(U_{t}^{b}\geq0\) for all t. The total discounted dividends paid off before ruin are given by
where \(0< v<1\) is a discount factor. Given the initial surplus u, we define
as the expected present value of discounted dividends paid off prior to ruin.
Difference equations
In this section, we derive difference equations for the expected discounted dividends paid before ruin. First, we introduce an auxiliary process \(\bar{U}^{b}_{t}\) defined as \(\bar{U}_{0}^{b}=u\) and for \(t=1,2,\ldots\) ,
where Ȳ independent of other random variables is distributed like Y, and
Accordingly, we define the ruin time by
with \(\bar{\tau}^{b}=\infty\) if \(\bar{U}_{t}^{b}\geq0\) for all t. For the risk model \(\bar{U}^{b}\), the discounted dividends paid before ruin are given by
Define the expected present value of discounted dividends paid before ruin by
For the surplus process \(U^{b}\), consider the following situations:

(1)
no claim occurs in \((0, 1]\) and no dividend is paid in \((0, 1]\);

(2)
no claim occurs in \((0, 1]\) and a dividend of 1 is paid in \((0, 1]\) (if \(u < b\), this case does not exist);

(3)
a main claim and its byclaim occur simultaneously in \((0, 1]\), and no dividend is paid in \((0, 1]\);

(4)
a main claim and its byclaim occur simultaneously in \((0, 1]\), and a dividend of 1 is paid in \((0, 1]\) (if \(u < b\), this case does not exist);

(5)
a main claim occurs in \((0, 1]\) and its byclaim is delayed to the next period, and no dividend is paid in \((0, 1]\);

(6)
a main claim occurs in \((0, 1]\) and its byclaim is delayed to the next period, and a dividend of 1 is paid in \((0, 1]\) (if \(u < b\), this case does not exist).
Note that in situations (1)(4), the surplus process \(U^{b}\) will regenerate itself after the first period; whereas in (5)(6), \(U^{b}\) will switch to \(\bar{U}^{b}\). For \(0\leq u< b\), no dividends will be paid in the first time period, then we have
where we use the convention \(\sum_{x=i}^{j}\cdot=0\) for \(i>j\). Whereas for \(u\geq b\), a dividend will be paid at the beginning of the first time period with probability θ, then we have
Similarly, for model \(\bar{U}^{b}\), we have for \(0\leq u< b\),
and for \(u\geq b\),
The case \(0\leq u< b\)
In this section, we consider the case \(0\leq u< b\). In order to simplify (2.1) and (2.3), we define the following auxiliary functions:
It is easily seen that the difference equations (2.1) and (2.3) can be rewritten as follows:
Now we relax the restriction \(0\leq u< b\) to \(u\geq0\) in (3.1), and let \((\chi_{1}(u), \chi_{2}(u))\) be the corresponding solution, i.e.
In order to get \((\chi_{1}(u), \chi_{2}(u))\), we use the generating function method. In the rest of this paper, we put a hat on top of a function to denote its generating function. For example,
For the convolution
since \(f*g(0)=f*g(1)=0\), its generating function is given by
It is not hard to see that
For example,
For \(0<z<1\), multiplying the first equation in (3.2) by \(z^{u}\) and then summing over u from 0 to ∞, we obtain
which leads to
Similarly, from the second equation in (3.2) we can obtain
Immediately, solving (3.3) and (3.4) gives
where we have used the fact
Hence, we conclude that the solution to the difference system (3.2) is uniquely determined by the initial value \(\chi_{1}(0)\), from which we know that the solution to (3.1) can be expressed as follows:
where α is an unknown constant, \(h_{,1}(u)\) and \(h_{,2}(u)\) are determined by the generating functions
with
Note that \(\hat {w}_{,1}(z)\), \(\hat {w}_{,2}(z) \) are both analytic inside the unit circle. In fact, since \(q_{,21}(0)=q_{,22}(0)=0\), we have
Hence, upon inverting the above generating functions we obtain
To continue, we introduce the discrete DicksonHipp operator defined as
for some function \(f(x)\) defined on \(\{0, ,1,\ldots\}\). As a matter of fact, \(\mathcal{T}_{z}f(y)\) is the generating function of \(f(y+\cdot)\). One of the nice properties of \(\mathcal{T}_{z}\) is the commutative property, i.e.
For more properties on this operator, we refer the interested readers to Li [18].
For \(\gamma_{}(z):=zv[1q+q\hat {f}(z)\hat {g}(z) ]\), we have
which imply that there is a number \(\rho_{}\in(0,1)\) such that \(\gamma_{}(\rho_{})=0\). Furthermore, note that
Then we have
which also yields for \(z\leq1\),
Hence, we conclude that \(\phi(x):=vq\mathcal{T}_{\rho_{}}(f*g)(x+1)\) is a defective probability function.
Now for \(k=1,2\),
After inverting the generating functions in the above formula, we obtain
where
The jfold convolution \(\phi^{*j}(x)\) in (3.7) is recursively defined as
with the starting point \(\phi^{*0}(x)=\mathbf{1}_{(x=0)}\).
The case \(u\geq b\)
In this section, we consider the case \(u\geq b\). First, we introduce the following auxiliary functions to simplify (2.2) and (2.4):
Immediately, (2.2) and (2.4) are simplified to be
We use generating function method to solve (4.1). By some straightforward calculations, we obtain
Similarly, we have
For \(0<z<1\), we have
and similarly
Now multiplying both sides of the first equation in (4.1) and summing over u from b to ∞, we obtain
where
Applying exactly the same arguments to the second equation in (4.1) gives
with
After inverting the generating functions \(\hat {\varphi}_{1}(z)\), \(\hat {\varphi}_{2}(z)\), we obtain for \(u=0,1,\ldots\) ,
Note that
where \(\hat {a}(z)=[1q+q\hat {f}(z)\hat {g}(z) ](1\theta+\theta z)\) is a probability generating function with the corresponding probability function given by
Then solving (4.2) and (4.3) results in
For \(\gamma_{+}(z):=zv\hat {a}(z)\), we have
then there exists a number \(\rho_{+} \in(0, 1)\) such that \(\gamma_{+}(\rho_{+})=0\), which also implies that \(\rho_{+}\) is the zero point of the common denominator of (4.4) and (4.5). Note that \(V(u;b)\) cannot grow with an exponential rate, then we conclude that \(\rho_{+}\) is also zero point of the numerators of (4.4) and (4.5), and this leads to
where the second equality holds since \(\hat {q}_{+,11}(\rho_{+})\hat {q}_{+,22}(\rho_{+})\hat {q}_{+,12}(\rho_{+})\hat {q}_{+,21}(\rho_{+})=0\) thanks to \(\gamma_{+}(\rho_{+})=0\). For convenience, set
Then we have
It follows from (4.6) that the numerators in (4.4) and (4.5) reduce to
and
where
After inverting the generating functions in the above formulas, we obtain for \(x=0,1,2\ldots\) ,
Furthermore, using \(\hat {\zeta}_{j0}(\rho_{+})+\alpha \hat {\zeta}_{j1}(\rho_{+})=0\), \(j=1,2\), we have
Similarly, for the common denominator in (4.4) and (4.5), using \(\gamma_{+}(\rho_{+})=0\) we obtain
which also gives for \(z\leq1\),
Hence, \(\beta(x):=v\mathcal{T}_{\rho_{+}}a(x+1)\) is a defective probability function.
Now plugging (4.9) and (4.10) into (4.4) gives
upon inversion, which yields
where
Similarly, from (4.5) we can obtain
where
where the jfold convolution \(\beta^{*j}\) is defined as in \(\phi^{*j}\).
It remains to determine the unknown constant α. To this end, we set \(u=b1\) in the first equation in (3.1) to obtain
Then plugging (3.5) and (4.12) into the above equation gives
Finally, we summarize the main results in the following theorem.
Theorem 1
The expected present values of dividends \(V(u;b)\), \(\bar{V}(u;b)\) can be expressed as follows:
where α is given by (4.14).
Conclusion
Dividend problems are hot topics in insurance risk theory. In this paper, we consider a compound binomial model with delayed claims. Suppose that the insurance company will possibly pay dividends when the surplus level is larger than a given barrier b. The expected present values of dividends paid before ruin are studied. We derive systems of difference equations for \(V(u;b)\) and \(\bar{V}(u;b)\), and get the solutions by generating function method. The main results given in Theorem 1 show that the analytic expressions for \(V(u;b)\) and \(\bar {V}(u;b)\) can be obtained.
References
 1.
Gerber, HU: Mathematical fun with the compound binomial process. ASTIN Bull. 24, 177184 (1988)
 2.
Shiu, ESW: The probability of eventual ruin in the compound binomial model. ASTIN Bull. 19, 179190 (1989)
 3.
Willmot, GE: Ruin probabilities in the compound binomial model. Insur. Math. Econ. 12, 133142 (1993)
 4.
Dickson, DCM: Some comments on the compound binomial models. ASTIN Bull. 24, 3345 (1994)
 5.
Yang, H, Zhang, Z, Lan, C: Ruin problems in a discrete Markov risk model. Stat. Probab. Lett. 79, 2128 (2009)
 6.
Yang, H, Zhang, Z: On a discrete risk model with twosided jumps. J. Comput. Appl. Math. 234, 835844 (2010)
 7.
Gerber, HU, Shiu, ESW, Yang, H: An elementary approach to discrete models of dividend strategies. Insur. Math. Econ. 46, 109116 (2010)
 8.
Chen, M, Yuen, KC, Guo, J: Survival probabilities in a discrete semiMarkov risk model. Appl. Math. Comput. 232, 205215 (2014)
 9.
Yuen, KC, Guo, J: Ruin probabilities for timecorrelated claims in the compound binomial model. Insur. Math. Econ. 29, 4757 (2001)
 10.
Yuen, KC, Guo, J, Ng, KW: On ultimate ruin in a delayedclaims risk model. J. Appl. Probab. 42, 163174 (2005)
 11.
Xiao, Y, Guo, J: The compound binomial risk model with timecorrelated claims. Insur. Math. Econ. 41, 124133 (2007)
 12.
Albrecher, H, Cheung, ECK, Thonhauser, S: Randomized observation periods for the compound Poisson risk model: dividends. ASTIN Bull. 41, 645672 (2011)
 13.
Avanzi, B, Cheung, ECK, Wong, B, Woo, JK: On a periodic dividend barrier strategy in the dual model with continuous monitoring of solvency. Insur. Math. Econ. 52, 98113 (2013)
 14.
Zhang, Z: On a risk model with randomized dividenddecision times. J. Ind. Manag. Optim. 10, 10411058 (2014)
 15.
Zhang, Z, Cheung, ECK: The Markov additive risk process under an Erlangized dividend barrier strategy. Methodol. Comput. Appl. Probab. (2014). doi:10.1007/s1100901494147
 16.
Tan, J, Yang, X: The compound binomial model with randomized decisions on paying dividends. Insur. Math. Econ. 39, 118 (2006)
 17.
He, L, Yang, X: The compound binomial model with randomly paying dividends to shareholders and policyholders. Insur. Math. Econ. 46, 443449 (2010)
 18.
Li, S: On a class of discrete time renewal risk models. Scand. Actuar. J. 2005, 241260 (2005)
Acknowledgements
The authors would like to thank two anonymous referees for their helpful comments and suggestions, which improved an earlier version of the paper. This work is supported by the National Natural Science Foundation of China (11101451, 11471058, 11426051), the Natural Science Foundation Project of CQ CSTC of China (cstc2014jcyjA00007), Scientific and Technological Research Program of Chongqing Municipal Education Commission (KJ1400521) and the Fundamental Research Funds for the Central Universities (106112015CDJXY100006).
Author information
Additional information
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
All authors contributed equally to the writing of this paper. All authors read and approved the final manuscript.
Rights and permissions
Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.
About this article
Received
Accepted
Published
DOI
Keywords
 randomized dividend strategy
 expected discounted dividends
 difference equations
 delayed claims