搜索
从保险人的角度谈风险管理何华,博士中国保险学会年会,2011年5月电子邮箱:hhe@munichre.com为什么要重视巨灾?巨灾风险评估巨灾风险分担&定价巨灾风险积累&业务组合优化议程为什么要重视巨灾?巨灾风险评估巨灾风险分担&定价巨灾风险积累&业务组合优化议程©GeoRisksResearch,MunichRe,2009为什么要重视巨灾?巨灾风险高MunichReGlobeofNaturalHazardsSpatialDistributionPerMagnitudeRange(1973–Present)地震的空间分布(1973–2010)为什么要重视巨灾?巨灾风险高为什么要重视巨灾?巨灾风险越来越高地震的时间分布(1973–2010)为什么要重视巨灾?巨灾风险越来越高地震的时间分布(1973–2010)YearYearLoss(US$,m)(BasedonPCSData)美国飓风保险损失为什么要重视巨灾?巨灾风险越来越高(US$bn)Overalllosses(in2010values)Insuredlosses(in2010values)©2011MünchenerRückversicherungs-Gesellschaft,GeoRisksResearch,NatCatSERVICE–AsatJanuary2011501001502002503001980198219841986198819901992199419961998200020022004200620082010Averageofthelast10years:OveralllossUS$110bnInsuredlossUS$35bn全球巨灾经济和保险损失（1980–2010）为什么要重视巨灾?巨灾风险越来越高PML=ProbableMaximumLoss?(BasedonAIRModeledIndustryLossesforUSHurricanes)为什么要重视巨灾?巨灾风险概率分布的长尾性E[loss|a100-year-eventoccurs]=123billions79billionsPML=ProbableMaximumLoss?(BasedonAIRModeledIndustryLossesforUSHurricanes)为什么要重视巨灾?巨灾风险概率分布的长尾性(BasedonAIRModeledIndustryLossesforUSHurricanes)巨灾损失风险的尾巴比恐龙尾巴都要长！PDFLossPML=ProbableMaximumLoss?为什么要重视巨灾?巨灾风险概率分布的长尾性Parma2009台风碰到台风?为什么要重视巨灾?巨灾风险的复杂性Lupit2009为什么要重视巨灾?巨灾风险的复杂性台风碰到高气压?(1nauticalmile=1.15statutemiles)134889128164242050100150200250300012243648607284Error(nmi)ForecastPeriod(hours)OfficialTrackForecastErrorsAtlanticBasin,1989-1998(Source:NHC,NOAA)为什么要重视巨灾?巨灾风险的复杂性ForecastPeriodMaximumstrikeprobability72hours10%-15%48hours20%-25%36hours25%-35%24hours40%-50%12hours75%-85%(Source:NHC,NOAA)ChancetogetStrickenwhenExpertsSaySo?为什么要重视巨灾?巨灾风险的复杂性为什么要重视巨灾?巨灾风险评估巨灾风险分担&定价巨灾风险积累&业务组合优化议程2011/5/1018September2009,Padangearthquake,M7.6,~1300dead巨灾风险评估地震振动破坏(AIR-WorldwideCorp.,Boston)RelativelyMinorDamage(Residential)巨灾风险评估地震振动破坏TransAlaskaPipeline巨灾风险评估地震振动破坏TransAlaskaPipeline巨灾风险评估地震振动破坏(AIR-WorldwideCorp.,Boston)PredominantDamage(Commercial)巨灾风险评估地震振动破坏(AIR-WorldwideCorp.,Boston)PredominantDamage(Government)巨灾风险评估地震振动破坏Whiplash巨灾风险评估地震振动破坏Highriselowrise?巨灾风险评估地震振动破坏Highriselowrise?巨灾风险评估地震振动破坏高层楼房比低层楼房抗震性能好?巨灾风险评估地震振动破坏巨灾风险评估地震液化破坏10.05.201129TitelderPräsentationundNamedesRednersChristchurch,NewZealand,2011巨灾风险评估地震液化破坏(Source:RenatoU.SolidumJr,PHIVOLCS,2007)DagupanCity,Phillipines巨灾风险评估地震液化破坏CityofNiigata,1964巨灾风险评估地震液化破坏(中国地震局袁晓铭等,2009)Wenchuanearthquake巨灾风险评估地震液化破坏(CourtesyofAIR-Worldwide)巨灾风险评估地震液化破坏巨灾风险评估地震海啸破坏巨灾风险评估风载碎片冲击破坏2011/5/1036巨灾风险评估风载碎片冲击破坏巨灾风险评估风载碎片冲击破坏2011/5/1038巨灾风险评估风载碎片冲击破坏ChaseTower,Houston,IKE,2008ASCE7-05(ReferredbyIBC03):“GlazinginCategoryII,III,orIVbuildingslocatedover60ft(18.3m)abovethegroundandover30ft(9.2m)aboveaggregatesurfaceroofslocatedwithin1,500ft(458m)ofthebuildingshallbepermittedtobeunprotected.”巨灾风险评估风载碎片冲击破坏ChaseTower,Houston,IKE,2008InterfirstBankAlicia,1983(photobyProf.Joe.Minor)ASCE7-05(ReferredbyIBC03):“GlazinginCategoryII,III,orIVbuildingslocatedover60ft(18.3m)abovethegroundandover30ft(9.2m)aboveaggregatesurfaceroofslocatedwithin1,500ft(458m)ofthebuildingshallbepermittedtobeunprotected.”巨灾风险评估风载碎片冲击破坏巨灾风险评估强风破坏为什么要重视巨灾?巨灾风险评估巨灾风险分担&定价巨灾风险积累&业务组合优化议程ConceptionofmodelsSetofscenariosStatisticsExpectedlossorlossoccurrenceprobabilityRiskcurveIndividualexposureValuedistribution巨灾风险分担与定价为什么需要模型Time19002000Loss巨灾风险分担与定价为什么需要模型Time19002000历史巨灾事件Loss巨灾风险分担与定价为什么需要模型Time19002000模拟巨灾事件Loss巨灾风险分担与定价为什么需要模型历史巨灾事件PDF投影到概率空间Loss巨灾风险分担与定价为什么需要模型PMLCurve巨灾风险分担与定价为什么需要模型PMLAALPMLCurve巨灾风险分担与定价为什么需要模型ContributionofIntensitiestotechnicalrate(AAL)00.010.020.030.040.050.060.070.080.090.1678910MMIrelativeContributiontoAAL00.10.20.30.40.50.60.70.80.91acumulatedContributintoAALAALAAL_cumulative巨灾风险分担与定价为什么需要模型比例合约AALEPLossDeductibleLimit巨灾风险分担与定价如何用模型定价非比例合约EPLossLayer1Layer2Layer3Layer4巨灾风险分担与定价如何用模型定价盲人摸象巨灾风险分担与定价模型面对的挑战巨灾风险分担与定价模型面对的挑战盲人摸象案例分析:飓风IKE,CommercialLOBGround-Up($millions)Gross($millions)真实损失49.18.7模型损失31.91.3ModeledLossesincludedemandsurge巨灾风险分担与定价模型面对的挑战RiskCountExposureActualGround-UpModeledGround-Up13502,529,414,94620,000,00013,882,436SelectedPolicy1小身子,大钱包问题1:为什么低估Ground-uploss?巨灾风险分担与定价模型面对的挑战RisksExposureActualGround-UpModeledGround-Up465,699,62318,000,000348,595关连风险！SelectedPolicy2巨灾风险分担与定价模型面对的挑战问题1:为什么低估Ground-uploss?05001,0001,5002,0002,5003,000GrossLoss(USD)ClaimsModeledPolicyCountsGround-upLoss(USD)RisksCount巨灾风险分担与定价模型面对的挑战问题2:为什么严重低估Grossloss?分布错误!75%ProbabilityDamageRatioat100mph100%25%80%ProbabilityDamageRatioat100mph100%15%25%5%平均损失率=25%ActualModeled巨灾风险分担与定价模型面对的挑战平均损失率=25%问题2:为什么严重低估Grossloss?100,000independentandidenticalhouses;50dollarsdeductibleforeachhouse.Lossdistr.Ground-uplossGrosslossActual75Kintact,25Ktotaled25K*100=2.5M25K*50=1.25MModel15Kintact,80Kwith25%loss,5Ktotaled80K*25%*100+5K*100%*100=2.5M5K*50=250K巨灾风险分担与定价模型面对的挑战问题2:为什么严重低估Grossloss?100,000independentandidenticalhouses;50dollarsdeductibleforeachhouse.Lossdistr.Ground-uplossGrosslossActual75Kintact,25Ktotaled25K*100=2.5M25K*50=1.25MModel15Kintact,80Kwith25%loss,5Ktotaled80K*25%*100+5K*100%*100=2.5M5K*50=250K巨灾风险分担与定价模型面对的挑战问题2:为什么严重低估Grossloss?分布（不确定性）致关重要！2011/5/1062巨灾风险分担与定价模型面对的挑战为什么要重视巨灾?巨灾