Research article

Robust optimal investment-reinsurance strategies with the preferred reinsurance level of reinsurer

  • This paper investigates robust equilibrium investment-reinsurance strategy for a mean variance insurer. With a larger market share, a reinsurer has a greater say in negotiating reinsurance contracts and makes the decision to propose the preferred level of reinsurance and charges extra fees as a penalty for losses that deviate from the preferred level of reinsurance. Once the insurer receives a decision from the reinsurer, the insurer weighs its risk-bearing capacity against the cost of reinsurance in order to find the optimal investment-reinsurance strategy under the mean-variance criterion. The insurer who is ambiguity averse to jump risk and diffusion risk obtains a robust optimal investment-reinsurance strategy by dynamic programming principle. Moreover, the reinsurance strategy is no longer excess-of-loss reinsurance or proportional reinsurance. In particular, the insurer may purchase proportional reinsurance for different ranges of loss and the proportion depends on the extra charge rate, which is more consistent with market practice than the standard excess-of-loss reinsurance and proportional reinsurance. The optimal reinsurance policy depends on the degree of ambiguous aversion to jump risk and not on the degree of ambiguous aversion to diffusion risk. Finally, several numerical experiments are presented to illustrate the impacts of key parameters on the value function and the optimal reinsurance contracts.

    Citation: Wanlu Zhang, Hui Meng. Robust optimal investment-reinsurance strategies with the preferred reinsurance level of reinsurer[J]. AIMS Mathematics, 2022, 7(6): 10024-10051. doi: 10.3934/math.2022559

    Related Papers:

    [1] Hasanen A. Hammad, Maryam G. Alshehri . Generalized Ξ-metric-like space and new fixed point results with an application. AIMS Mathematics, 2023, 8(2): 2453-2472. doi: 10.3934/math.2023127
    [2] Hongyan Guan, Jinze Gou, Yan Hao . On some weak contractive mappings of integral type and fixed point results in b-metric spaces. AIMS Mathematics, 2024, 9(2): 4729-4748. doi: 10.3934/math.2024228
    [3] Dong Ji, Yao Yu, Chaobo Li . Fixed point and endpoint theorems of multivalued mappings in convex b-metric spaces with an application. AIMS Mathematics, 2024, 9(3): 7589-7609. doi: 10.3934/math.2024368
    [4] Muhammad Rashid, Muhammad Sarwar, Muhammad Fawad, Saber Mansour, Hassen Aydi . On generalized b-contractions and related applications. AIMS Mathematics, 2023, 8(9): 20892-20913. doi: 10.3934/math.20231064
    [5] Afrah Ahmad Noman Abdou . Chatterjea type theorems for complex valued extended b-metric spaces with applications. AIMS Mathematics, 2023, 8(8): 19142-19160. doi: 10.3934/math.2023977
    [6] Senthil Kumar Prakasam, Arul Joseph Gnanaprakasam, Ozgur Ege, Gunaseelan Mani, Salma Haque, Nabil Mlaiki . Fixed point for an OgF-c in O-complete b-metric-like spaces. AIMS Mathematics, 2023, 8(1): 1022-1039. doi: 10.3934/math.2023050
    [7] Badriah Alamri, Jamshaid Ahmad . Fixed point results in b-metric spaces with applications to integral equations. AIMS Mathematics, 2023, 8(4): 9443-9460. doi: 10.3934/math.2023476
    [8] Rashid Ali, Faisar Mehmood, Aqib Saghir, Hassen Aydi, Saber Mansour, Wajdi Kallel . Solution of integral equations for multivalued maps in fuzzy b-metric spaces using Geraghty type contractions. AIMS Mathematics, 2023, 8(7): 16633-16654. doi: 10.3934/math.2023851
    [9] Naeem Saleem, Salman Furqan, Mujahid Abbas, Fahd Jarad . Extended rectangular fuzzy b-metric space with application. AIMS Mathematics, 2022, 7(9): 16208-16230. doi: 10.3934/math.2022885
    [10] Ying Chang, Hongyan Guan . Generalized (αs,ξ,,τ)-Geraghty contractive mappings and common fixed point results in partial b-metric spaces. AIMS Mathematics, 2024, 9(7): 19299-19331. doi: 10.3934/math.2024940
  • This paper investigates robust equilibrium investment-reinsurance strategy for a mean variance insurer. With a larger market share, a reinsurer has a greater say in negotiating reinsurance contracts and makes the decision to propose the preferred level of reinsurance and charges extra fees as a penalty for losses that deviate from the preferred level of reinsurance. Once the insurer receives a decision from the reinsurer, the insurer weighs its risk-bearing capacity against the cost of reinsurance in order to find the optimal investment-reinsurance strategy under the mean-variance criterion. The insurer who is ambiguity averse to jump risk and diffusion risk obtains a robust optimal investment-reinsurance strategy by dynamic programming principle. Moreover, the reinsurance strategy is no longer excess-of-loss reinsurance or proportional reinsurance. In particular, the insurer may purchase proportional reinsurance for different ranges of loss and the proportion depends on the extra charge rate, which is more consistent with market practice than the standard excess-of-loss reinsurance and proportional reinsurance. The optimal reinsurance policy depends on the degree of ambiguous aversion to jump risk and not on the degree of ambiguous aversion to diffusion risk. Finally, several numerical experiments are presented to illustrate the impacts of key parameters on the value function and the optimal reinsurance contracts.



    Fixed point (FP) theory has been applied in many domains of science, including approximation theory, biology, chemistry, dynamic systems, economics, engineering, fractals, game theory, logic programming, optimization problems, and physics. Undoubtedly, the FP theory is an active area in mathematics and beautifully combines analysis, topology, and geometry. In the past few decades, it has been clear that the FP theory is a very effective and significant instrument for investigating nonlinear processes. Its application relies on the existence of solutions to mathematical problems that are based on the contraction principle. After Banach [1] presented his principle, the FP theory approaches became more successful and appealing to scientists, see [2,3,4,5,6,7,8,9,10].

    Now, let's start with the definition of FP and the Banach contraction principle.

    Let (W,D) be a metric space. A point ζW is called a FP of a mapping S:WW, if

    S(ζ)=ζ.

    Theorem 1. (Banach contraction principle) Let S be a self-mapping on a complete metric space (W,D). If there exists a constant θ [0,1) such that

    D(S(ζ),S(ξ))θD(ζ,ξ) (1.1)

    for all ζ,ξW, then S has a unique FP.

    The term "nonexpansive mapping" refers to a mapping S satisfying the condition (1.1) for θ=1. Nonexpansive mappings emerged as a direct extension of the concept of contraction mappings initially introduced by Banach [1]. In 1965, Browder [11] and Göhde [12] established an FP theorem for nonexpansive mappings within uniformly convex Banach spaces.

    In 1975, Zhang [13] introduced the broader category of mean nonexpansive mappings, encompassing nonexpansive mappings, and demonstrated the existence and uniqueness of FPs for this class of mappings within Banach spaces exhibiting normal structural properties.

    Definition 1. [13] Let (W,D) be a metric space. A mapping S:WW is called mean nonexpansive if there exist α,β0 with α+β1 such that

    D(S(ζ),S(ξ))αD(ζ,ξ)+βD(ζ,S(ξ)),  ζ,ξW. (1.2)

    In a recent development, Mebawondu et al. [14] introduced the notion of T-mean nonexpansive mapping within the framework of metric spaces. Subsequently, they provided rigorous proof establishing the existence of a unique FP for this particular mapping.

    Definition 2. [14] Let T,S be self-mappings on a metric space (W,D). S is said to be a T-mean nonexpansive mapping if there exist α,β0 with α+β1 such that

    D(T(S(ζ)),T(S(ξ)))αD(T(ζ),T(ξ))+βD(T(ζ),T(S(ξ)),     ζ,ξW. (1.3)

    Remark 1. If we take T=I, where I is the identity function, in (1.3), then we obtain that S is a mean nonexpansive mapping defined by (1.2).

    Also, Mebawondu et al. [14] presented the following example with T-mean nonexpansive mapping but not mean nonexpansive mapping in metric spaces.

    Example 1. [14, Examples 2.1 and 2.2] (i) Let W=[1,) and D(ζ,ξ)=|ζξ| for all ζ,ξW. Let T,S:WW be defined by T(ζ)=1+lnζ and

    S(ζ)={10ζ,   if ζ2,  1,      if ζ=2.

    (ii) Let W=[0,2] and D be as in (i). Let T,S:WW be defined by

    T(ζ)={1ζ,  if ζ[0,1],2ζ,   if ζ(1,2],

    and

    S(ζ)={1,   if ζ[0,1),2,   if ζ[1,2].

    Then, in the above examples, S is T-mean nonexpansive, but not mean nonexpansive.

    The oldest known iteration method is the Picard iteration, which is used in the Banach contraction principle. In 2009, Morales and Rojas [15] offered a novel iteration scheme called the T-Picard iteration, which is defined as follows.

    Definition 3. [15] Let T,S be self-mappings on a metric space (W,D) and ζ0W. The sequence {T(ζn)}W defined by

    T(ζn+1)=T(S(ζn))=T(Sn(ζ0)),  nN0=N{0} (1.4)

    is called the T-Picard iteration associated with S.

    Remark 2. If we take T=I in (1.4), then we obtain the Picard iteration.

    Another domain where FP theory can be extended pertains to spatial considerations. Recently, significant advancements have been made in this direction, particularly within the realm of metric spaces. The concept of b-metric space was introduced by Bakhtin [16] and Czerwik [17] in different periods. In a different line of development, Matthews [18] introduced the notion of partial metric space and adapted the Banach contraction principle for applications in program verification.

    In 2012, Amini-Harandi [19] introduced the concept of metric-like space, wherein a point's self-distance need not be equal to zero. Within these novel spaces, Amini-Harandi [19] presented various FP theorems that extended and enhanced existing results in both partial metric and b-metric spaces. Subsequently, in the following year, Alghamdi et al. [20] introduced the notion of b-metric-like space, thereby generalizing the concepts of b-metric and metric-like spaces, and established various associated FP results. Since then, numerous outcomes related to FPs of mappings under specific contractive conditions within these spaces have been obtained, as exemplified by works such as [21,22,23,24].

    Motivated by these seminal results, we investigated the existence and uniqueness of FPs for T-mean nonexpansive mappings and provided a stability theorem for T-Picard iteration within the framework of b-metric-like spaces. Furthermore, we explored the existence and uniqueness of solutions to the Fredholm-Hammerstein integral equations on time scales as an application of our primary theorem. Finally, we presented two illustrative numerical examples from different time scales. It is noteworthy that our results were significant as they generalized the corresponding findings within b-metric and metric-like spaces, both of which are special instances of b-metric-like spaces.

    In this subsection, several well-known definitions and two lemmas in b-metric-like spaces are listed.

    Definition 4. [20, Definition 2.4] Let W be a nonempty set and k1 be a given real number. A function Dbl:W×WR+ is b-metric-like (in brief, bl) if the following conditions are satisfied:

    (Dbl1) Dbl(ζ,ξ)=0ζ=ξ (indistancy implies equality);

    (Dbl2) Dbl(ζ,ξ)=Dbl(ξ,ζ) (symmetry);

    (Dbl3) Dbl(ζ,ξ)k[Dbl(ζ,η)+Dbl(η,ξ)] (weakened triangularity)

    for all  ζ,ξ,ηW. The pair (W,Dbl) is called a bl space.

    In a bl space (W,Dbl), if ζ,ξW and Dbl(ζ,ξ)=0, then ζ=ξ; but the converse may not be true since Dbl(ζ,ζ) may be positive for some ζW.

    Remark 3. As it is excepted, each bl space forms a metric-like space by letting k=1. On the other hand, every b-metric space is a bl space with the same parameter k. Hence, the class of bl spaces is bigger than the class of metric-like spaces (see [19]) or b-metric spaces (see [17,25,26,27]). However, the inverse implications are not true in general.

    Example 2. [20, Examples 2.5 and 2.6] Let W=[0,). Define a function Dbl:W2[0,) by Dbl(ζ,ξ)=(ζ+ξ)2 or Dbl(ζ,ξ)=(max{ζ,ξ})2. Then (W,Dbl) is a bl space with the parameter k=2. Clearly, (W,Dbl) is not a b-metric space or metric-like space.

    Definition 5. [20, Definition 2.9] Let (W,Dbl) be a bl space, {ζn} be a sequence in W, and ζW. We say that

    (i) {ζn} is said to be a convergent sequence if limnDbl(ζn,ζ)=Dbl(ζ,ζ) and a Cauchy sequence if limn,mDbl(ζn,ζm) exists and is finite;

    (ii) (W,Dbl) is called complete if, for every Cauchy sequence {ζn} in W, there exists ζW such that limn,mDbl(ζn,ζm)=Dbl(ζ,ζ)=limnDbl(ζn,ζ).

    Remark 4. In a bl space, the limit for a convergent sequence is not unique in general. However, if {ζn} is a Cauchy sequence with limn,mDbl(ζn,ζm)=0 in the complete bl space (W,Dbl), then the limit of such a sequence is unique. Indeed, in such a case, if ζnζ (Dbl(ζn,ζ)Dbl(ζ,ζ)) as n we get that Dbl(ζ,ζ)=0.

    In 2009, Beiranvand et al. [28] defined the concept of sequentially convergent mapping in metric spaces. Following this, many researchers have worked on this concept (see [14,29,30,31,32]). Inspired by these results, we provide the following definition in bl spaces.

    Definition 6. Let (W,Dbl) be a bl space. A mapping T:WW is said to be sequentially convergent if we have, for every sequence {ξn}, if {T(ξn)} is convergent then {ξn} is also convergent.

    Lemma 1. [33, Lemma 1.10] Let T be a self-mapping on a bl space (W,Dbl). If T is continuous at uW then, for every sequence {ζn} in W such that ζnu, we have T(ζn)T(u), that is,

    limnDbl(T(ζn),T(u))=Dbl(T(u),T(u)).

    In 2023, Calderón et al. [34] studied the concept of (T,S)-stability in metric spaces. Now, we define this concept in the setting of bl spaces as follows.

    Definition 7. Let T,S be self-mappings on a bl space (W,Dbl) and ζ be the FP of S. Let {T(ζn)} be a sequence generated by an iteration scheme, that is,

    T(ζn+1)=h(T,S,ζn),   nN0, (2.1)

    where ζ0W is the initial point and h is a function. Assume that {T(ζn)} converges to T(ζ). Let {T(ξn)}W be an arbitrary sequence, and set ρn=Dbl(T(ξn+1),h(T,S,ξn)), nN0. Then, the iteration scheme (2.1) is said to be (T,S)-stable if and only if

    limnρn=0limnT(ξn)=T(ζ).

    If we take T=I in Definition 7, it is reduced to the concept of the stability of an iteration scheme, which was defined by Harder and Hicks [35].

    Lemma 2. [36, p. 14] Let {ξn}, {ρn} be sequences of non-negative numbers and 0θ<1, so that

    ξn+1θξn+ρn,

    for all nN. If limnρn=0, then limnξn=0.

    In 1988, Hilger [37] introduced the theory of time scales, which has recently garnered a lot of attention, in his Ph. D. thesis to unify continuous and discrete analysis. In this subsection, we will go over several fundamental concepts related to time scales.

    Definition 8. [38] A time scale TS is an arbitrary, nonempty, and closed subset of the real numbers R.

    The forward and backward jump operators are defined by ω(t)=inf{sTS:s>t}, and τ(t)=sup{sTS:s<t}, respectively, where inf=supTS and infTS=sup. The graininess function μ:TS[0,+) is defined by μ(t)=ω(t)t.

    The classification of points on the time scale TS is possible with the jump operators.

    Definition 9. [38, Definition 1.1] A point tTS is said to be right-dense if ω(t)=t, right-scattered if ω(t)>t, left-dense if τ(t)=t, left-scattered if τ(t)<t, isolated if τ(t)<t<ω(t), and dense if τ(t)=t=ω(t).

    In the following definition, the set TκS that will be needed in the delta derivative is given.

    Definition 10. [38] The set TκS is defined as follows:

    TκS={TS(τ(supTS),supTS],      ifsupTS<,          TS,                           ifsupTS=.

    Definition 11. [38, Definition 1.10] Assume f:TS R is a function and fix tTκS. The delta derivative (also Hilger derivative) fΔ(t) exists if, for every ϵ>0, there exists a neighbourhood U=(tδ,t+δ)TS for some δ>0 such that

    |[f(ω(t))f(s)]fΔ(t)[ω(t)s]|ϵ|ω(t)s|      forallsU.

    Proposition 1. [38] Let f:TS R be a function and tTκS. If f is continuous at t and t is right-scattered, then f is differentiable at t with

    fΔ(t)=f(ω(t))f(t)μ(t).

    Remark 5. [38] (i) If  TS= R, then fΔ(t)=f(t) is the derivative used in the standard calculus.

    (ii) If  TS= Z, then fΔ(t)=f(t+1)f(t)=Δf(t) is the forward difference operator used in difference equations.

    Definition 12. [38, Definition 1.1] A function f:TS R is called rd-continuous if it is continuous at all right-dense points in TS and its left-sided limits exist (finite) at all left-dense points in TS.

    If f is rd-continuous, then there exists a function F such that FΔ(t)=f(t) (see [38, Theorem 1.74]). In this case, the (Cauchy) delta integral of f is defined by

    edf(t)Δt=F(e)F(d)   forall d,eTS.

    Remark 6. [38] (i) If TS= R, then

    edf(t)Δt=edf(t)dt

    is the Riemann integral used in the standard calculus.

    (ii) If TS= Z, then

    edf(t)Δt=e1t=df(t)    foralld,eZwithd<e.

    Lemma 3. [39] (Cauchy-Schwarz inequality) Let f,g:TS R be two rd-continuous mappings. For all d,eTS with de, we have

    ed|f(t)g(t)|Δt{ed|f(t)|2Δt}.{ed|g(t)|2Δt}.

    Proposition 2. [40] Let d and e be arbitrary points in TS.

    (i) Every constant function f(t)=c (tTS) is Δ-integrable from d to e and edcΔt=c(ed).

    (ii) If f and g are Δ-integrable on [d,e) and f(t)g(t) for all t[d,e), then

    edf(t)Δtedg(t)Δt.

    (iii) If f is Δ-integrable on [d,e), then |f| is Δ-integrable on [d,e) and

    |edf(t)Δt|ed|f(t)|Δt.

    (iv) If f and g are Δ-integrable on [d,e), then their product f.g is Δ-integrable on [d,e).

    We recommend the reader to [38,41,42,43] for more information on time scales and their applications.

    Let's start this section with an example of T-mean nonexpansive mapping, which doesn't mean nonexpansive in a bl space with the same constants.

    Example 3. Let W=[0,1] and Dbl(ζ,ξ)=(max{ζ,ξ})2 for all ζ,ξW. Then (W,Dbl) is a bl space with the parameter k=2. Let T,S:WW be defined by

    T(ζ)={3ζ8,    ifζ[0,1),0,      ifζ=1,

    and

    S(ζ)={1,     ifζ[0,1),0,   ifζ=1.

    Let α=0 and β=1. Then, we get

    Dbl(T(S(ζ)),T(S(ξ)))=0αDbl(T(ζ),T(ξ))+βDbl(T(ζ),T(S(ξ)))

    for all ζ,ξW. As a result, S is T-mean nonexpansive. To demonstrate that S does not mean nonexpansive, we assume that ζ=0 and ξ=1. In this case, we have

    Dbl(S(ζ),S(ξ))=1>0=0.1+1.0=αDbl(ζ,ξ)+βDbl(ζ,S(ξ)).

    Hence, S does not mean nonexpansive.

    Now, let's give the FP result of the paper.

    Theorem 2. Let (W,Dbl) be a complete bl space and T:WW be a continuous, one-to-one, and sequentially convergent mapping. If S:WW is a T-mean nonexpansive mapping such that k(α+2βk)<1, then S has a unique FP. Morever, the sequence {T(ζn)} defined by (1.4) converges strongly to T(ζ), where ζ is the FP of S.

    Proof. Using (1.3), (1.4), and (Dbl3), we get

    Dbl(T(ζn+1),T(ζn))=Dbl(T(S(ζn)),T(S(ζn1)))αDbl(T(ζn),T(ζn1))+βDbl(T(ζn),T(S(ζn1)))=αDbl(T(ζn),T(ζn1))+βDbl(T(ζn),T(ζn))αDbl(T(ζn),T(ζn1))+βk[Dbl(T(ζn),T(ζn1))+Dbl(T(ζn1),T(ζn))]=(α+2βk)Dbl(T(ζn),T(ζn1)). (3.1)

    Also, we obtain

    Dbl(T(ζn),T(ζn1))=Dbl(T(S(ζn1)),T(S(ζn2)))αDbl(T(ζn1),T(ζn2))+βDbl(T(ζn1),T(S(ζn2)))=αDbl(T(ζn1),T(ζn2))+βDbl(T(ζn1),T(ζn1))αDbl(T(ζn1),T(ζn2))+βk[Dbl(T(ζn1),T(ζn2))+Dbl(T(ζn2),T(ζn1))]=(α+2βk)Dbl(T(ζn1),T(ζn2)). (3.2)

    Substituting (3.2) into (3.1), we have

    Dbl(T(ζn+1),T(ζn))(α+2βk)2Dbl(T(ζn1),T(ζn2)),

    and inductively, we get

    Dbl(T(ζn+1),T(ζn))(α+2βk)n1Dbl(T(ζ2),T(ζ1)). (3.3)

    Furthermore, for n>m, we have

    Dbl(T(ζm),T(ζn))k[Dbl(T(ζm),T(ζm+1))+Dbl(T(ζm+1),T(ζn))]kDbl(T(ζm),T(ζm+1))+k2[Dbl(T(ζm+1),T(ζm+2))+Dbl(T(ζm+2),T(ζn))]kDbl(T(ζm),T(ζm+1))+k2Dbl(T(ζm+1),T(ζm+2))+...+knm1Dbl(T(ζn2),T(ζn1))+knmDbl(T(ζn1),T(ζn)). (3.4)

    Now, (3.3) and (3.4) imply that

    Dbl(T(ζm),T(ζn))[k(α+2βk)m1+k2(α+2βk)m+k3(α+2βk)m+1+...+knm(α+2βk)n2]Dbl(T(ζ2),T(ζ1))=k(α+2βk)m1[1+k(α+2βk)+(k(α+2βk))2+...+(k(α+2βk))nm1]Dbl(T(ζ2),T(ζ1))k(α+2βk)m11k(α+2βk)Dbl(T(ζ2),T(ζ1)).

    It follows that {T(ζn)} is a Cauchy sequence with limn,mDbl(T(ζn),T(ζm))=0 and since (W,Dbl) is a complete bl space, there exists ξ0W such that

    limnDbl(T(ζn),ξ0)=Dbl(ξ0,ξ0)=0. (3.5)

    Since the mapping T is sequentially convergent and the sequence {T(ζn)} is convergent, then the sequence {ζn} is convergent. So, there exists ζW such that

    limnDbl(ζn,ζ)=Dbl(ζ,ζ). (3.6)

    Because T is continuous, by Lemma 1 and (3.6), we have

    limnDbl(T(ζn),T(ζ))=Dbl(T(ζ),T(ζ)). (3.7)

    From (3.5) and (3.7), we obtain ξ0=T(ζ). So

    Dbl(T(S(ζ)),T(ζ))k[Dbl(T(S(ζ)),T(S(ζn)))+Dbl(T(S(ζn)),T(ζ))]k[αDbl(T(ζ),T(ζn))+βDbl(T(ζ),T(S(ζn)))]+kDbl(T(S(ζn),T(ζ))=kαDbl(T(ζ),T(ζn))+k(β+1)Dbl(T(ζ),T(ζn+1))0   asn.

    Therefore, Dbl(T(Sζ),T(ζ))=0, which implies that T(S(ζ))=T(ζ). Since T is one-to-one, we have that

    S(ζ)=ζ.

    Hence, ζ is an FP of S. To show that ζ is a unique FP of S, we suppose on the contrary that there exists another FP, say ξ, such that ζξ. That is, S(ζ)=ζ and S(ξ)=ξ. Thus, we get

    Dbl(T(ζ),T(ξ))=Dbl(T(S(ζ)),T(S(ξ)))αDbl(T(ζ),T(ξ))+βDbl(T(ζ),T(S(ξ)))=(α+β)Dbl(T(ζ),T(ξ)),

    which implies that

    (1(α+β))Dbl(T(ζ),T(ξ))0.

    Since k(α+2βk)<1 and k1, we have α+β<1. That is, we obtain

    Dbl(T(ζ),T(ξ))=0,

    which implies that T(ζ)=T(ξ). Since T is one-to-one, we get ζ=ξ.

    Remark 7. Theorem 2 can be regarded as an extension of Theorem 2.3 in [14] from a metric space to a bl space.

    If we choose T=I in Theorem 2, we obtain the below result, which is new in the literature.

    Corollary 8. Let (W,Dbl) be a complete bl space and S:WW be a mean nonexpansive mapping with k(α+2βk)<1. Then S has a unique FP.

    We obtain the following stability result using Definition 7.

    Theorem 3. Under the hypotheses of Theorem 2, the sequence {Tζn} defined by (1.4) is (T,S)-stable.

    Proof. Suppose that {T(ξn)} is an arbitrary sequence in W and ρn=Dbl(T(ξn+1),T(S(ξn))). Let limnρn=0. Then, by (1.3) and (Dbl3), we obtain

    Dbl(T(ξn+1),T(ζ))k[Dbl(T(ξn+1),T(S(ξn)))+Dbl(T(S(ξn)),T(ζ))]=k[Dbl(T(ξn+1),T(S(ξn)))+Dbl(T(S(ξn)),T(S(ζ)))]kρn+k(α+β)Dbl(T(ξn),T(ζ)).

    By the assumption limnρn=0, it follows from Lemma 2 that limnT(ξn)=T(ζ).

    Conversely, limnDbl(T(ξn+1),T(ζ))=0. Then, using Lemma 2, we have

    ρn=Dbl(T(ξn+1),T(S(ξn)))k[Dbl(T(ξn+1),T(ζ))+Dbl(T(ζ),T(S(ξn)))]=k[Dbl(T(ξn+1),T(ζ))+Dbl(T(S(ξn)),T(S(ζ)))]kDbl(T(ξn+1),T(ζ))+k(α+β)Dbl(T(ξn),T(ζ))0   asn.

    Therefore, {T(ζn)} defined by (1.4) is (T,S)-stable.

    The study of the existence and uniqueness of differential and integral equation solutions is essential in exploring various types of nonlinear analysis and engineering mathematics. One of the most important tools developed in this field is the FP method. Using the FP result in the previous part, we investigated the existence and uniqueness of solutions for the Fredholm-Hammerstein integral equations on time scales in this section.

    First, let's give the definition of the Fredholm-Hammerstein integral equation.

    Definition 13. [44] A Fredholm-Hammerstein integral equation of the second kind is defined as

    ζ(t)=f(t)+λedK(t,s)u(s,ζ(s))ds,         dt,se,  d,eR, (4.1)

    where the kernel function K(t,s) and the function f(t) are given, the unknown function ζ(t) must be determined, λR is a non-zero constant, and the known function u is continuous and nonlinear respect to the variable ζ. It is called homogeneous if f(t)=0 for all t[d,e] in the Eq (4.1).

    Now, we provide the homogeneous Fredholm-Hammerstein integral equation of the second kind on the time scale as follows.

    Definition 14. Let TS be a time scale with the delta derivative operator Δ, and let d,eTS. A homogeneous Fredholm-Hammerstein integral equation of the second kind on the time scale TS is defined as

    ζ(t)=λedK(t,s)u(s,ζ(s))Δs,      t,s[d,e]TS=[d,e]TS, (4.2)

    where u:[d,e]TS×RR and K:[d,e]TS×[d,e]TSR are given functions, ζ:[d,e]TSR is an unknown function, and λR{0} is a parameter.

    Let W=C([d,e]TS) be the set of all real continuous functions defined on [d,e]TS. We endowed W with the Dbl

    Dbl(ζ,ξ)=supt[d,e]TS(|ζ(t)|+|ξ(t)|)2,    ζ,ξW.

    It was shown in [20, p. 21] that (W,Dbl) is a complete bl space with the parameter k=2.

    In the following theorem, we proved the existence and uniqueness of the integral Eq (4.2) on the complete bl space C([d,e]TS).

    Theorem 4. Consider the Fredholm-Hammerstein integral Eq (4.2) such that the functions K and u are Δ-integrable on [d,e]TS. Assume that

    (|u(t,ζ(t))|+|u(t,ξ(t))|)2(|ζ(t)|+|ξ(t)|)2 (4.3)

    for all ζ,ξC([d,e]TS) and

    supt[d,e]TS(edK2(t,s)Δs)L, (4.4)

    where L(0,12λ2(ed)) is a constant. Then the Eq (4.2) has a unique solution in C([d,e]TS).

    Proof. We define the mappings T and S as follows:

    T(ζ(t))=23ζ(t)

    and

    S(ζ(t))=λedK(t,s)u(s,ζ(s))Δs,  t[d,e]TS.

    Since the functions K and u are Δ-integrable, then, clearly S is a self-mapping on C([d,e]TS), that is, S:C([d,e]TS)C([d,e]TS). By the Cauchy-Schwarz inequality in Lemma 3, Proposition 2, and (4.3), for all ζ,ξC([d,e]TS), we obtain

    (|T(S(ζ(t)))|+|T(S(ξ(t)))|)2=(|2λ3edK(t,s)u(s,ζ(s))Δs |+|2λ3edK(t,s)u(s,ξ(s))Δs |)24λ29(ed|K(t,s)|.|u(s,ζ(s))|Δs+ed|K(t,s)|.|u(s,ξ(s))|Δs)2=4λ29(ed|K(t,s)|.(|u(s,ζ(s))|+|u(s,ξ(s))|)Δs)24λ29(edK2(t,s)Δs)(ed(|u(s,ζ(s))|+|u(s,ξ(s))|)2Δs)4λ29(edK2(t,s)Δs)(ed(|ζ(s)|+|ξ(s)|)2Δs)4λ29supt[d,e]TS(edK2(t,s)Δs)(ed)supt[d,e]TS(|ζ(t)|+|ξ(t)|)2,

    which implies that

    supt[d,e]TS(|T(S(ζ(t)))|+|T(S(ξ(t)))|)24λ29supt[d,e]TS(edK2(t,s)Δs)(ed)supt[d,e]TS(|ζ(t)|+|ξ(t)|)2.

    Taking the advantage of (4.4), we have

    supt[d,e]TS(|T(S(ζ(t)))|+|T(S(ξ(t)))|)24λ29L(ed)Dbl(ζ,ξ).

    Thus, if we say λ2L(ed)=α, we obtain

    Dbl(T(S(ζ)),T(S(ξ)))αDbl(T(ζ),T(ξ))αDbl(T(ζ),T(ξ))+βDbl(T(ζ),T(S(ξ))).

    Since L<12λ2(ed), then we have α<12. Then S is a T-mean nonexpansive mapping on C([d,e]TS) with α<12 and β=0 implying that 2(α+4β)<1. Then it is clear that T is a continuous, one-to-one, and sequentially convergent mapping. Therefore, all conditions of Theorem 2 are satisfied, and so S has a unique FP that is a unique solution of the integral Eq (4.2) for |λ|<12L(ed).

    The following examples illustrate the result of Theorem 4.

    Example 4. Let TS=2N0={1,2,4,...}. For all t2N0, ω(t)=2t. Consider the following integral equation

    ζ(t)=λ41ζ(s)s2Δs,   t[1,4]TS.

    Here, K(t,s)=s2 and u(t,ζ(t))=ζ(t) for all ζC([1,4]TS). Thus, by Proposition 1, we obtain

    41K2(t,s)Δs=41(s2)2Δs=41s4Δs=41(87s3)ΔΔs=87(s3)|s=4s=1=98.

    Therefore, the condition (4.4) holds with L=98. Clearly, we get

    (|u(t,ζ(t))|+|u(t,ξ(t))|)2=(|ζ(t)|+|ξ(t)|)2.

    Thus, the condition (4.3) is provided. By Theorem 4, the given integral equation has a unique solution for L<16λ2, that is, |λ|<233. Consequently, we conclude that this integral equation has a unique solution, which is the trivial solution ζ(t)0.

    In fact, it can be shown that the equation has only the solution ζ(t)0 for |λ|<233 by using direct computation. Let

    c=41ζ(s)s2Δs.

    Then we have ζ(t)=cλ, and so,

    c=41cλs2Δs=cλ411s2Δs=32cλ.

    This equation can be written as

    c(132λ)=0,

    that is, for λ23, there is only one solution, which is c=0. Hence, we obtain that this integral equation has a unique solution when λ23.

    Example 5. Let TS=Z. Consider the following integral equation

    ζ(t)=λ40(s+ω(s))ζ(s)1+(ζ(s))2Δs,   t[0,4]TS. (4.5)

    Here, K(t,s)=s+ω(s), u(t,ζ(t))=ζ(t)1+(ζ(t))2 for all ζC([0,4]TS) and ω(t)=t+1 for all tZ. By Remark 6 (ii), we obtain

    40K2(t,s)Δs=40(2s+1)2Δs=40(4s2+4s+1)Δs=3s=0(4s2+4s+1)=84.

    Therefore, the condition (4.4) holds with L=84. Also, we get

    (|u(t,ζ(t))|+|u(t,ξ(t))|)2=(|ζ(t)1+(ζ(t))2|+|ξ(t)1+(ξ(t))2|)2(|ζ(t)|+|ξ(t)|)2,

    that is, the condition (4.3) is satisfied. By Theorem 4, the integral Eq (4.5) has a unique solution for L<18λ2. Therefore, we have |λ|<1442. We deduce that the given integral equation has a unique solution, which is the trivial solution ζ(t)0.

    In fact, by using direct computation, it can be shown that the equation has a unique solution ζ(t)0. Let

    c=40(s+ω(s))ζ(s)1+(ζ(s))2Δs. (4.6)

    Thus, we can write

    ζ(t)=cλ. (4.7)

    By substituting the value of ζ(t) given by (4.7) into (4.6), we obtain

    c=40(s+s+1)cλ1+(cλ)2Δs=cλ1+(cλ)240(2s+1)Δs=16cλ1+(cλ)2,

    and from here, we get

    c3λ2+c(116λ)=0.

    For λ<116, there is only one solution, which is c=0. Hence, the integral Eq (4.5) has a unique solution when λ<116.

    In the above examples, Theorem 4 provides a small interval for λ. However, on this interval, we can guarantee the existence and uniqueness of the solution without computing it.

    We have demonstrated the existence and uniqueness of the FPs of T-mean nonexpansive mappings and provided the stability result for the T-Picard iteration in bl spaces. Additionally, we presented an application of the Fredholm-Hammerstein integral equations and two numerical examples on time scales. Using similar approaches in our results, the integral type of T-mean nonexpansive mappings described in [14] can be studied in bl spaces. Also, the types of T-mean nonexpansive mappings can be studied in different spaces, such as hyperbolic metric spaces defined by Kohlenbach [45].

    The authors declare they have not used Artificial Intelligence (AI) tools in the creation of this article.

    The authors, A. Aloqaily and N. Mlaiki, would like to thank Prince Sultan University for paying the APC and for the support through the TAS research lab.

    The authors declare that they have no conflict of interest.



    [1] E. Anderson, L. P. Hansen, T. Sargent, Robustness, detection and the price of risk, Technical report, mimeo, 2000.
    [2] K. J. Arrow, Uncertainty and the welfare economics of medical care, Am. Econ. Rev., 53 (1963), 941–973.
    [3] L. Bai, J. Guo, Optimal proportional reinsurance and investment with multiple risky assets and no-shorting constraint, Insur. Math. Econom., 42 (2008), 968–975. https://doi.org/10.1016/j.insmatheco.2007.11.002 doi: 10.1016/j.insmatheco.2007.11.002
    [4] T. Björk, M. Khapko, A. Murgoci, On time-inconsistent stochastic control in continuous time, Finance Stoch., 21 (2017), 331–360. https://doi.org/10.1007/s00780-017-0327-5 doi: 10.1007/s00780-017-0327-5
    [5] T. Bjork, A. Murgoci, A general theory of markovian time inconsistent stochastic control problems, 2017. Available from: https://ssrn.com/abstract=1694759.
    [6] T. Björk, A. Murgoci, X. Y. Zhou, Mean–variance portfolio optimization with state-dependent risk aversion, Math. Financ., 24 (2014), 1–24. https://doi.org/10.1111/j.1467-9965.2011.00515.x doi: 10.1111/j.1467-9965.2011.00515.x
    [7] K. Borch, Reciprocal reinsurance treaties, Astin Bull., 1 (1960), 170–191. https://doi.org/10.1017/S0515036100009557 doi: 10.1017/S0515036100009557
    [8] K. Borch, The optimal reinsurance treaty, Astin Bull., 5 (1969), 293–297. https://doi.org/10.1017/S051503610000814X doi: 10.1017/S051503610000814X
    [9] N. Branger, L. S. Larsen, Robust portfolio choice with uncertainty about jump and diffusion risk, J. Bank. Financ., 37 (2013), 5036–5047. https://doi.org/10.1016/j.jbankfin.2013.08.023 doi: 10.1016/j.jbankfin.2013.08.023
    [10] S. Browne, Optimal investment policies for a firm with a random risk process: exponential utility and minimizing the probability of ruin, Math. Oper. Res., 20 (1995), 937–958. https://doi.org/10.1287/moor.20.4.937 doi: 10.1287/moor.20.4.937
    [11] L. Chen, Y. Shen, Stochastic stackelberg differential reinsurance games under time-inconsistent mean–variance framework, Insur. Math. Econom., 88 (2019), 120–137. https://doi.org/10.1016/j.insmatheco.2019.06.006 doi: 10.1016/j.insmatheco.2019.06.006
    [12] M. Chen, K. C. Yuen, W. Wang, Optimal reinsurance and dividends with transaction costs and taxes under thinning structure, Scand. Actuar. J., 2021 (2021), 198–217. https://doi.org/10.1080/03461238.2020.1824158 doi: 10.1080/03461238.2020.1824158
    [13] Y. Chi, K. S. Tan, S. C. Zhuang, A bowley solution with limited ceded risk for a monopolistic reinsurer, Insur. Math. Econom., 91 (2020), 188–201. https://doi.org/10.1016/j.insmatheco.2020.02.002 doi: 10.1016/j.insmatheco.2020.02.002
    [14] X. Guo, J. Liu, X. Y. Zhou, A constrained non-linear regular-singular stochastic control problem, with applications, Stoch. Proc. Appl., 109 (2004), 167–187. https://doi.org/10.1016/j.spa.2003.09.008 doi: 10.1016/j.spa.2003.09.008
    [15] D. Hu, S. Chen, H. Wang, Robust reinsurance contracts with uncertainty about jump risk, Eur. J. Oper. Res., 266 (2018), 1175–1188. https://doi.org/10.1016/j.ejor.2017.10.061 doi: 10.1016/j.ejor.2017.10.061
    [16] D. Landriault, B. Li, D. Li, V. R. Young, Equilibrium strategies for the mean-variance investment problem over a random horizon, SIAM J. Financ. Math., 9 (2018), 1046–1073. https://doi.org/10.1137/17M1153479 doi: 10.1137/17M1153479
    [17] D. Li, D. Li, V. R. Young, Optimality of excess-loss reinsurance under a mean–variance criterion, Insur. Math. Econom., 75 (2017), 82–89. https://doi.org/10.1016/j.insmatheco.2017.05.001 doi: 10.1016/j.insmatheco.2017.05.001
    [18] D. Li, Y. Zeng, H. Yang, Robust optimal excess-of-loss reinsurance and investment strategy for an insurer in a model with jumps, Scand. Actuar. J., 2018 (2018), 145–171. https://doi.org/10.1080/03461238.2017.1309679 doi: 10.1080/03461238.2017.1309679
    [19] X. Liang, V. R. Young, Minimizing the discounted probability of exponential parisian ruin via reinsurance, SIAM J. Control Optim., 58 (2020), 937–964. https://doi.org/10.1137/19M1282714 doi: 10.1137/19M1282714
    [20] P. J. Maenhout, Robust portfolio rules and asset pricing, Rev. Financ. Stud., 17 (2004), 951–983. https://doi.org/10.1093/rfs/hhh003 doi: 10.1093/rfs/hhh003
    [21] P. J. Maenhout, Robust portfolio rules and detection-error probabilities for a mean-reverting risk premium, J. Econ. Theory, 128 (2006), 136–163. https://doi.org/10.1016/j.jet.2005.12.012 doi: 10.1016/j.jet.2005.12.012
    [22] C. S. Pun, Robust time-inconsistent stochastic control problems, Automatica, 94 (2018), 249–257. https://doi.org/10.1016/j.automatica.2018.04.038 doi: 10.1016/j.automatica.2018.04.038
    [23] H. Schmidli, minimizing the ruin probability by investment and reinsurance, Ann. Appl. Probab., 12 (2002), 890–907. https://doi.org/10.1214/aoap/1031863173 doi: 10.1214/aoap/1031863173
    [24] Z. Sun, X. Zheng, X. Zhang, Robust optimal investment and reinsurance of an insurer under variance premium principle and default risk, J. Math. Anal. Appl., 446 (2017), 1666–1686. https://doi.org/10.1016/j.jmaa.2016.09.053 doi: 10.1016/j.jmaa.2016.09.053
    [25] R. Uppal, T. Wang, Model misspecification and underdiversification, J. Financ., 58 (2003), 2465–2486. https://doi.org/10.1046/j.1540-6261.2003.00612.x doi: 10.1046/j.1540-6261.2003.00612.x
    [26] W. Wang, D. Muravey, Y. Shen, Y. Zeng, Optimal investment and reinsurance strategies under 4/2 stochastic volatility model, 2021. http://dx.doi.org/10.13140/RG.2.2.24915.50721
    [27] Y. Yuan, Z. Liang, X. Han, Robust reinsurance contract with asymmetric information in a stochastic stackelberg differential game, Scand. Actuar. J., in press. https://doi.org/10.1080/03461238.2021.1971756
    [28] Y. Zeng, D. Li, A. Gu, Robust equilibrium reinsurance-investment strategy for a mean–variance insurer in a model with jumps, Insur. Math. Econom., 66 (2016), 138–152. https://doi.org/10.1016/j.insmatheco.2015.10.012 doi: 10.1016/j.insmatheco.2015.10.012
    [29] H. Zhao, Y. Shen, Y. Zeng, W. Zhang, Robust equilibrium excess-of-loss reinsurance and cds investment strategies for a mean–variance insurer with ambiguity aversion, Insur. Math. Econom., 88 (2019), 159–180. https://doi.org/10.1016/j.insmatheco.2019.06.007 doi: 10.1016/j.insmatheco.2019.06.007
    [30] X. Zheng, J. Zhou, Z. Sun, Robust optimal portfolio and proportional reinsurance for an insurer under a cev model, Insur. Math. Econom., 67 (2016), 77–87. https://doi.org/10.1016/j.insmatheco.2015.12.008 doi: 10.1016/j.insmatheco.2015.12.008
  • This article has been cited by:

    1. Nihan Turan, Metin Başarır, Aynur Şahin, On the solutions of a nonlinear system of q-difference equations, 2024, 2024, 1687-2770, 10.1186/s13661-024-01896-6
    2. Aynur Şahin, Zeynep Kalkan, The AA-iterative algorithm in hyperbolic spaces with applications to integral equations on time scales, 2024, 9, 2473-6988, 24480, 10.3934/math.20241192
  • Reader Comments
  • © 2022 the Author(s), licensee AIMS Press. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0)
通讯作者: 陈斌, bchen63@163.com
  • 1. 

    沈阳化工大学材料科学与工程学院 沈阳 110142

  1. 本站搜索
  2. 百度学术搜索
  3. 万方数据库搜索
  4. CNKI搜索

Metrics

Article views(1714) PDF downloads(121) Cited by(2)

Figures and Tables

Figures(3)

Other Articles By Authors

/

DownLoad:  Full-Size Img  PowerPoint
Return
Return

Catalog