应用概率统计第二十二卷ChineseJournalofAppliedProbability第二期2006年5月andStatisticsV01.22No.2May2006
ExplicitAsymptoticsfortheRuinProbability
withRiskyInvestmentIncluded冰
CHENYIQING
(DepartmentofStatisticsandActuarialScience,TheUniversityofHongKong,HongKong)
DONGXIAOJUXIEXIANGSHENG
(SchoolofEconomicsandManagement,GuangdongUniversityofTechnology,Guangzhou,510090)
Abstract
Inthispaper,weinvestigatetheruinprobabilityofadiscrete—timeriskmodel,inwhichthesurplusofaninsurancebusinessiscurrentlyinvestedintoariskyasset.Usingapurelyprobabilis—tictreatment.weestablishexplicitasymptoticrelationsfortheinfinite-timeruinprobabilities,henceweextendarecentresultofTangandTsitsiashvili(2003)totheinfinite-timecase.
Keywords:Asymptotics,regularvariation,ruinprobability,stochasticequation.AMSSubjectClassiflcation:62E20,60G50,60G70.
§1.introduction
FollowingNyrhinen(1999),TangandTsitsiashvili(2003),andChenandXie(2005),weconsideradiscrete—timeriskmodel.inwhichthesurplusoftheinsurancecompanyiscurrentlyinvestedintoariskyassetwhichmayleadtoanegativereturnineachyear.DenotebyAn∈(一∞,∞)thenetincome(thetotalincomingpremiumminusthetotalclaimamount)withinyearnandbyrn∈(一1,∞)thereturnrateatyearn,礼=1,2,…。Lettheinitialsurplusbez≥0.Hence,ifweassumethatthenetincomeAniscalculatedatthebeginningofyearn,thenthesurplusaccumulatedtilltheendofyear礼,characterizedbySn,satisfiestherecurrenceequation
岛=z≥0,S。=(1+rn)(.sn一1+A。),礼=1,2,…;(1.1)
alternatively,ifweassumethatthenetincomeA,liscalculatedattheendofyearn,thenthesurplusaccumulatedtilltheendofyear佗,characterizedby孔,satisfiestherecurrenceequation
To=。≥0,正。=(1+rn)霉。一1+A。,礼=1,2,….(1.2)
Throughoutthepaper,weassumethatthenetincomesAn,礼=1,2,…,constituteasequenceofindependent,identicallydistributed(i.i.d.)randomvariables(r.v.’s),thatthereturnrates‰,礼=1,2,…,alsoconstituteasequenceofi.i.d.r.v.’S,andthatthetwosequences{An,n=1,2,…)andr?l,n=1,2,…)areindependent.
*ThisworkwassupportedbytheGuangdongNaturalScienceFoundation(ProjectNo.980415)Received2004.12.10.Revised2005.7.1.
万 方数据
152应用概率统计第二十二卷
黝・ire
墨l=一A…×l=1+rn’n=1,2,….(1.3)
Ther.、,.Xnisthenetpayoutduringyear扎andther.v.Kisthediscountfactorfromyear竹toyearTt一1,n=1,2,….InWhatfollows,wewritebyXandYthegenericr.v.’softhesequences{xn,n=1,2,…)and{】么,n=1,2,・・・)andwritebyFandGthedistributionfunctions(d.f.’S)ofther.V.’sXandY,respectively.Clearly,ther.v.yisstrictlypositive.WeassumethatthetailprobabilityF(x)=1一F(x)=P(x>。)satisfiesF(x)>0foranyrealnumberz.Inth鬯terminologyofNorberg(1999)andTangandTsitsiashvili(2003),wecallxtheinsuranceriskand】,’thefinancialrisk.
Correspondingtothesurplusprocesses(1.1)and(1.2),wedefinetheruinprobabilitieswithinfinitetime礼=1.2....andinfinitetimeas
妒s(z,n)=P(。≤ra。i9n乳<ol&=。),砂T(z,n)=P(。麴。死<oI%=z),(1.4)respectively
惦(z)=P(唧mi<n。。Sk<0Is0=。),妇(z)=P(逸mi<n。。Tk<01To=。)(1,5)
Inthispaperweareinterestedintheasymptoticbehavioroftheseprobabilities。
Hereafter.alllimitrelationshipsareforo斗00unlessstatedotherwise;fortwopositivefunctionsa(x)and6(z),wewritea(x)Sb(x)iflimsupn(z)/b(。)≤1,writea(x)乏b(x)iflirai《a(x)/b(x)≥1,andwritea(x)一6(z)ifboth.
Nyrhinen(1999)investigatedtheasymptoticbehavioroftheruinprobability妒s@).Intermsofthemodeldescribedabove,wecanobtainacombinationofTheorems3.3and3.4ofNyrhinen(1999)asfollows:if(1)w=sup{tlHY。≤1)∈(0,。。),(2)EY。andElxI£arefiniteforsomet>W,(3)F(o)>0,and(4)someoftheconvolutionpowersofthedistributionoflogYhasanon—trivialabsolutelycontinuouscomponent,thentherelation
Cs(z、一Cx一”(1.6)
holdsfor。somepositive,butimplicit,constantC.
Theobjectiveofthepresentpaperistoestablishexplicitasymptoticrelationsfortheruinprobabilities妒s(z)andCT(z)undersomeotherassumptionsonthetailsofther.v.’sXandY.Intheproof,wewillusearecentresultofTangandTsitsiashvili(2003),whoinvestigatedthefinite・timeruinprobabilitiesCs(z,n)and妒T(。,佗)undertheassumptionsthatFisheavytailedandthatthetailG’isdominatedbythetailF.WewillalsoapplysomeresultsobtainedbyVervaat(1979)intheliteratureofstochasticdifferenceequations.
Therestofthispaperconsistsofthreesections:Section2presentsthemainresults’afterrecallinganimportantclassofheavy—taileddistributions,Section3collectssomelemmas,andSection4provesthetheorems.
万 方数据
第二期陈宜清董效菊谢湘盔。在风险投资下破产概率的一个渐近显式153
§2.MainResults
Wewillassumethatthed.f.FoftheinsuranceriskXhasaregularlyvaryingtail,denotedbyF∈死.Bydefinition,ad.f.Fconcentratedon(一。o,oo)belongstotheclass冗ifthereissomea≥0suchthat—
Xl-im-+'00裂F(x=∥一、。(2_1)、…7
foranyY>0.Forsimplicity,wedenotebyF∈冗一atheregularitypropertyin(2.1).Ifithiscasewehave
瓢)=X--at(咖xp∽学dⅣ)'z>d,i,(2.2)
forsomen>0,wherec(x)_÷c∈(0,OO)and£(z)-÷0;seeBinghamet,a1.(1987,page21).Thecorrespondingr.V.Xsatisfies
E(X+)p<∞for0≤P<n,E(X+)p=∞forP>0:,(2.3)
wherez+=max{x,o).
UndertheassumptionsthatFbelongstoaheavy・taileddistributionclass.whichisslightlylargerthantheclass冗,andthatthetailGisdominatedbythetailF(tobeprecise,EYP<0(3forsomePlargerthantheupperMatuszewskaindexofthed.f.F),TangandTsitsiashvili(2003,Theorem5.1andRemark5.1)obtainedpreciseasymptoticestimatesforthefinite.timeruin
fs(z,n)andfr(x,n)foreachfixed礼=1,2,….Thereallyapplicableresultsof
paperwereobtainedinthecasethat
F∈冗一nandEYp<ooforsomeP>OL>0.(2.4)
thiscase,itholdsforeachfixed似叩)一等黔77,=1,2,…that
(2.5)
TangandTsitsiashvili(2003,Theorem5.2(3)andRemark5.1).
Iffurtherweassumethat
Eya<1.(2.6)
theterm(EYo)“inrelation(2.5)vanishesasnincreases.Thoughrelation(2.5)wasonlybyTangandTsitsiashVili(2003)toholdforfixedln,Weintuitivelybelievethattheinfinite.ruinprobabilityfs(x)shouldsatisfy
fs(x)一1一EYaF(z).(2.7)
Thefollowingresultprovesthatthisistrue.
Theorem2.1Underassumptions(2.4)and(2.6),theasymptoticrelation(2.7)holds.
Theorem2.1successfullyestablishesanasymptoticrelationfortheruinprobabilityfs(x)infullyexplicitform.
万 方数据probabilitiestheirInseethenprovedtimea
154应用概率统计第二+二卷
AsfortheruinprobabilityCr(x),wehaveasimilarexplicitasymptoticrelationbelow.Theorem2.2Underassumptions(2.4)and(2.6),itholdsthat
Cr(z)Eya
1一EYOCF(x)(2.8)
§3.SomeLemmas
WefirstpointoutasimplerelationshipbetweentheruinprobabilitiesindexedbySandthoseindexedbyT.
Lemma3.1FortheriskmodelintroducedinSection1,theruinprobabilitiesdefinedby(1.4)and(1.5)satisfytherelations
妒T(z,n)=Z”妒s(z/耖,竹)G(d!,),札=1,2,…,(3.1)
and
妒r(z)=/o”妒s(z/可)G(d!,)(3.2)
ProofIteratingtherecurrenceequation(1.1)yieldsthat
忆nn
.%=z,S。=z兀(1+勺)+∑At丌(1+勺),竹=1,2,….(3.3)
j=li----1j=i
WewritethediscountedvalueofthesurplusSlin(3.3)as
~nni-1ni--1
So2o.S。=鼠兀巧=z+∑Ai兀巧=。一∑蜀兀巧,
j=li=1j=li=1j=x
0
where1-I=1byconvention.Itisclearthatforeachn=1,2,…
j=l
・
似叩)=P(。臻圭蕊墨i--1pK
。),t一、
惦(z)=P1鼢蚤托旦1p。).(3・4)
Similarly,itholdsthat
kt
Cr(z,佗)一、。m蜓ax,。蚤冠旦pz),州z)=P(。姒max。。量k置重巧>z)(3.5)Hencebythei.i.d.assumptionsmadeinSection1,relations(3.1)and(3.2)holds.孝
ThelemmabelowisacombinationofseveralresultsofVervaat(1979).Lemma3.2Let{(xmK),n=1,2,…)beasequenceofi.i.d.randompairswithgenericrandompair(X,y).Considerthestochasticdifferenceequation
V。=V,K,l+‰,竹=1,2,…(3.6)
If一∞≤Elog】7<0andE(109IxI)+<。o,thenther.v.’sKconvergesindistributiontosomereal—valuedr.v.比,whichisinvariantindistributionforallinitialr.V.’sv0.
万 方数据
第二期陈宜清董效菊谢湘生:在风险投资下破产概率的一个渐近显式155
ProofByTheorem1.6(b,c)ofVervaat(1979),thestochasticequation
影:d譬移+受.矿isindependentof(贾,矿),(3.7)
11asasolutioll,where≯denotesequalityindistribution.Then,byTheorem1.50)ofvervaat(1979),‘hbis
s~olutionuniqueindistributi。nand该definedby(3.6)conyergesindistributi。n
anyinitialr.v.弱.Since,byLemma1.1。fVervaat(1979),anylimit
1‘.v.比(K)should‘。s。me》(%),saY,forbeasolutionofequation(3.7),weproVethatforall谣ther.v.,s‰(讫)areidenticalindistribution.私
ThefollowinglemmaiswellknownandisfromPropositionofFeller(1971,p.278)orLemma1.3.1ofEmbrechtseta1.(1997).
Lemma3.3LetF1andBbetwod.f.’‘concentratedsoInen≥0,thentheirconvolutionF1宰F2∈冗一ax)+≯2(z).Theand蕊(z)~≯1on[0roo).IfF1∈冗~口andF2∈冗~。followinglemmaisfromBreiman(1965).
Lemma3.4LetXandYbetwoindependentr.v.’sdistributedbyFandG,respectively,Yisnonnegative・Ifd・f.F∈冗一aforsome0<a<ooandEYV<ooforsomep>△,then
。氅可丽2lim1P万(X万Y—>__x):Ey。.Ey。・
§4.ProofsoftheMainResuIt8
4.1ProofofTheorem2.1
From(3.4)and(2.5),itis妒s(。)≥妒s(茁,n)一鬻≯(z)clearthatforeachn:1,2,…,followsthatforeachn=1,2,
lirainf
o—■o。j擎堕>—1-(E—Y'1)nF(x1—1一Ey口‘
theright—handside
Cs(z)之1一EYQF(x)(4.1)
ItremainstoderiVeacorrespondingasymptoticupperboundfortheruinprobability始(茁).thisend,from(3.4)wederivethat
Cs(z)≤P(量肆苜匕、i=Ij=l>z)(4.2)
each礼=1,2,…,。。,set
n
%=E
l=1五+il--11巧.j=l
万 方数据forwhereItLetting扎_。。onToFor
156应用概率统计第二十二卷Bythei.i.d.assumptions,itiseasytoseethatforeach竹=1,2,…,
.n巩=4∑霹兀巧=碥,n
i=1j=i+l(4.3)
wherei兀。巧isequalto1byconvention.Clearly,withVo=0,theJ-----n十Isequence{%,n:1川2..)。’satisfiestherecurrenceequation
K。=M。M,1+砧forn=1,2,….(4.4)
Nowwechecktheconvergenceindistributionofthesequence{K,n=1,2,…).Byassurep—tion(2.6)weeasilyunderstandthat一。o≤ElogY<0shouldholdsincethefunctionf(t1:Eytisconvexint∈【0,Q]and,0(o)=ElogY.HencebyLemma3.2,V乞Convergesindistributiontoar.v・‰,say,whichisinvariantforallv0.Inviewof(4.3),thisactuallyprovesthat%isfinitealmostsurelyandisequaltoV乞indistribution.Specifically,wechoose比tobeindependentofthesequences{%,n=1,2,…)and{K,扎=1,2,…1.
KnowingF∈冗一n,weannouncethat
P(u矗>z)=P(Vk>z)≤P(Wo>z),。≥0,(4.5)
forsomenonnegativeinitialr.V.Yowithad.f.fromtheclass冗一n.Forthispurpose。wechooseanonnegativer.v.Zindependentofthesequences(%,n=1,2,…)and{%,n=1,2,…1suchthatp(z>z)一cF(x)forsomepositiveconstantc,whichwillbespecifiedlater.BvLemma3.4weknowthatthed.f.ofther.V.MZbelongstotheclass冗一aandP(Hz>z)。cEYoff(z):bythisandLemma3.3wefurtherknowthatP(KZ+xt>z)一(cEYa+1)劳(z).Hence,ifwechooseC>0sufficientlylargesuchthatcEYo+1<e,thenthereissomeconstantXo>0suchthatforallz>X0,P(YIZ+xt>z)≤P(Z>z).Bythisinequalitywecanprovethatforallz≥0,
P(YlZ+xt>xlz>XO)≤P(Z>zIz>;TO).(4.6)
Infact,for0≤z≤X0,inequality(4.6)triviallyholdssinceP(Z>xlz>zo)=1;forz>。o,inequality(4.6)canbeverifiedinthefollowingway:
P(YaZ+zt>xlZ>zo)=堕%≤器葛-一P(z>。fZ>Xo)P(Z>oo)“7“l.‘
Weidentifytheinitialr.v.vosuchthatitisindependentofthesequences{‰,n=1,2,…)and{K,n=1,2,…)andisP(Vo>z)=P(z>茁Jz>z。)“高F(z)・equalindistributiontother.v.Zconditionalon(z>xo).Hence,
(4.7)
Itfollowsfrom(4.6)thatforallz≥0,
P(Ⅵ>z)=P(YlVo+xt>z)=P(YlZ+耐>xlz>Xo)<P(Vo>。).万 方数据
第二期陈宜清董效菊谢湘生t在风险投资下破产概率的一个渐近显式157Successively,
P(V2>。)=P(Y2V1+x≯>z)≤P(Y2Vo-I-x}>。)=P(Y1Vo+x产>z)
=P(Ⅵ>z)≤P(%>。).
Applyingthemathematicalinductionmethodweknowthatforall佗=1,2,。・‘andallX≥0
P(V。>z)≤…≤P(%>。)≤P(Ⅵ>。)≤P(Vo>。)
SinceKconvergesindistributiontoYoo,takingn_ooyieldsthatforallz≥0,
P(u磊>。)=P(Yoo>。)≤P(V0>z),
aSannouncedin(4.5).
WecontinuetheproofofTheorem幺1.Forany
onE∈(0,1)andany礼=1,2,…,wetwopartsassplittheprobabilitytheright-handsideofinequality(4.2)into妒s(。)≤Pf壹耐i兀-1巧、i=1j:l>(1-咖)+P(象砧野∑耐兀b>锄)>£。)J=1+17
=^(z,E,n)七h(x,£,n)
Clearly,from(2.5)and(3.4)withxt№^小等张1叫z)beingreplacedbyxt,weobtainthatSince‰isindependentofthesequences{‰,礼=1,2,一・)and{K,n=1,2,…),
77、,£,n)=P((∑Ⅸ产兀巧)丌巧>£z)=P(比n巧>£。)p£。)j=nf、、i=n+1
≤P,:r4l巧>£。)一i晕黯F(£z),
relation(4.7)and+1117whereinthelaststepweapplied
佗=1,2,…,vS(z)s等F((1一s)。)十i;{器F(£。).
=等c・一£,一口+i裂翳E—a
n_∞andLemma3,4.Thus,foranyE∈(0,1)andanyItfollowsfromF∈冗一athat1;…一Cs(x),1一(E】,o)“1裟p葡百sSincenTi,觋p等萨+耥・罂等thenletting£_0leadtotheand£arearbitraryandEYo<1,firstletting
desiredresultthat
Cs(z)S1一EY口F(z)。(4.8)
万方数据
158应用概率统计第二十二卷
Combining(4.1)and(4.8)weobtain(2.7).Thisendstheproof.社
4.2ProofofTheorem2.2
Introduceasurvivald.f.RsbyRs(x)=(1一惦(z))1(¥>o),whichisastandardd.f.concen—tratedonf0,oo)withamassRs({o))=1一Cs(O)at0.Theorem2.1hasprovedthatRs∈冗一。.Hence,applyingLemma3.4toCT(z)~EY。妒s(z)“赤F(z)・relation(3.2)yieldsthat
亡、,Q
Thisendstheproof.社
References
【1】Bingham,N.H.,Goldie,C.M.,Teugels,J.L.,RegularVariation,CambridgeUniversityPress,Cam—
bridge,1987.
【2]Breiman,L.,Onsomelimittheoremssimilartothearc—sinlaw,Theor.ProbabilityAppl.,10(1965),
323—331.
[3】Chen,Y.,Xie,X.,Thefinitetimeruinprobabilitywiththesameheavy-tmledinsuranceandfinancial
risks,ActaMath.Appl.Sin.Engt.Ser.,21(1)(2005),153—156.
[4】Embrechts,P.,Klfippelberg,C.,Mikosch,T.,ModellingExtremalEventslorInsuranceandFinance,
Springer・Verlag,Berlin,1997.
f5】Feller,W,,AnIntroductiontoProbabilityTheoryandItsApplications,V01.II,SecondeditionJohn
Wiley&Sons,Inc.,NewYork-London-Sydney,1971.
【6】Norberg,R.,Ruinproblemswithassetsandliabilitiesofdiffusiontype,StochasticProcess.Appl.,
81(2)(1999),255—269.
c7JNyrhinen,H.,Ontheruinprobabilitiesinageneraleconomicenvironment,StochasticProcess.Appl.,
83(2)(1999),319—330.
[8】Tang,Q.,Tsitsiashvili,G.,Preciseestimatesfortheruinprobabilityinfinitehorizoninadiscrete.
timemodelwithheavy—tailedinsuranceandfinancialrisks,StochasticProcess.Appl.,108(2)(2003),299—325.
[9】Vervaat,W.,Onastochasticdifferenceequationandarepresentationofnonnegativeinfinitelydivis.
iblerandomvariables,Adv.inAppl.Probab.,11(4)(1979),750-783.
在风险投资下破产概率的一个渐近显式
陈宜清董效菊谢湘生
(香港大学统计与精算学系,香港)(广东工业大学经济管理学院。广州,510090)
在本文中,我们研究了一个离散时间风险模型的破产概率.在此风险模型中,保险公司的剩余资本被用于进行风险投资.我们运用纯概率的手法建立了无限时间破产概率的渐近显式,从而将Tang和Tsitsiashvili(2003)近期的一个结果推广到了无限时间的场合.
关键词:渐近式,正则变化,破产概率,随机方程.学科分类号:0211.3.
万 方数据
在风险投资下破产概率的一个渐近显式
作者:
作者单位:
刊名:
英文刊名:
年,卷(期):陈宜清, 董效菊, 谢湘生, CHEN YIQING, DONG XIAOJU, XIE XIANGSHENG陈宜清,CHEN YIQING(香港大学统计与精算学系,香港), 董效菊,谢湘生,DONG XIAOJU,XIEXIANGSHENG(广东工业大学经济管理学院,广州,510090)应用概率统计CHINESE JOURNAL OF APPLIED PROBABILITY AND STATISTICS2006,22(2)
参考文献(9条)
W On a stochastic difference equation and a representation of nonnegative infinitelydivisible random variables[外文期刊] 1979(04)
Q;Tsitsiashvili G Precise estimates for the ruin probability in finite horizon in adiscretetime model with heavy-tailed insurance and financial risks[外文期刊] 2003(02)
H On the ruin probabilities in a general economic environment[外文期刊] 1999(02)
R Ruin problems with assets and liabilities of diffusion type[外文期刊] 1999(02)
W;John Wiley;Sons An Introduction to Probability Theory and Its Applications 1971
P;Kliippelberg C;Mikosch T Modelling Extremal Events for Insurance and Finance 1997
Y;Xie X The finite time ruin probability with the same heavy-tailed insurance and financialrisks[期刊论文]-Acta Math Appli Sinica English Series 2005(01)
L On some limit theorems similar to the arc-sin law,Theor 1965
N H;Goldie C M;Teugels J L Regular Variation 1987
本文链接:
1