專題二32佳達(dá)再保險(xiǎn)經(jīng)紀(jì)有限公司Guy-Carpen匯總課件

1、專題二32佳達(dá)再保險(xiǎn)經(jīng)紀(jì)有限公司Guy_Carpen匯總專題二32佳達(dá)再保險(xiǎn)經(jīng)紀(jì)有限公司Guy_Carpen匯總量化與管理中國巨災(zāi)風(fēng)險(xiǎn)2008年9月26日David Lightfoot 佳達(dá)再保險(xiǎn)經(jīng)紀(jì) 西雅圖分部量化與管理中國巨災(zāi)風(fēng)險(xiǎn)2008年9月26日David LigDiscussion TopicsGlobal Evolution of Cat Models2008 Cat EventsRising Expectations in Cat Risk Measurement and ManagementRegulatory viewRating agency viewManaging Ca

2、tastrophe riskEvent monitoringi-aXsDiscussion TopicsGlobal Evolut討論議題巨災(zāi)模型的全球演進(jìn)2008年巨災(zāi)事件對巨災(zāi)風(fēng)險(xiǎn)評估和管理的新期望從監(jiān)管的角度從評級機(jī)構(gòu)的角度巨災(zāi)風(fēng)險(xiǎn)管理事件監(jiān)測i-aXs討論議題巨災(zāi)模型的全球演進(jìn)Evolution of Catastrophe Models19802008Significant Natural Catastrophes (and insured losses)1989 Hugo US$4.5b1991 Mirellie US$5.2b1990 Vivian US3b1992 Andrew

3、 US$18b1994 Northridge US$12b1995 Kobe US$3b2001 WTC US20b2004 Charley, Frances, Ivan Jeanne US28b2005 Katrina US$50bCat Model EvolutionCat models largely ignored(cat loading based on experience rating)The rise of the CAT ModelIncreased worldwide coverageCatastrophe casualty modelling emergesModels

4、feel the strain losses exceed modelled resultsEvolution of Catastrophe Model巨災(zāi)模型的演進(jìn)19802008重大自然巨災(zāi)事件 (以及保險(xiǎn)損失)1989 颶風(fēng)Hugo 45億美元1991 颶風(fēng)Mirellie 52億美元1990 颶風(fēng)Vivian 30億美元 1992 颶風(fēng)Andrew 180億美元1994 Northridge地震 120億美元1995 Kobe地震 30億美元2001 911事件 200億美元2004 颶風(fēng)Charley, Frances, Ivan Jeanne 280億美元2005 颶風(fēng)Katrin

5、a 500億美元巨災(zāi)模型演進(jìn)巨災(zāi)模型很大程度上被忽視 (巨災(zāi)附加費(fèi)率建立在經(jīng)驗(yàn)值上）巨災(zāi)模型的出現(xiàn)和發(fā)展世界范圍的覆蓋面不斷擴(kuò)大巨災(zāi)傷亡模型出現(xiàn)模型面臨壓力 - 實(shí)際損失超過模型結(jié)果巨災(zāi)模型的演進(jìn)19802008重大自然巨災(zāi)事件 1989 Value of Cat Models1st Layer2nd Layer3rd LayerNon-workingLayerWorkingLayerExperience RatingExposure Rating / Cat ModelingLossValue of Cat Models1st Layer2n巨災(zāi)模型的價(jià)值第一層第二層第三層非工作層（有理賠的

6、概率較低）工作層（有理賠的概率較高）根據(jù)經(jīng)驗(yàn)值定價(jià)根據(jù)風(fēng)險(xiǎn) / 模型定價(jià)理賠額巨災(zāi)模型的價(jià)值第一層第二層第三層非工作層工作層（有理賠的概率The Simple Method to Exposure RatingEarthquake ExampleThe largest magnitude earthquake you might expect near Beijing is 8.0The probability of this event could be about 1 in 500 years (based on historical data)1 such event in the la

7、st 500 years (1679)A PML estimate for this could range from 5% to 15% of Beijing total sums insured Probability of a 6.0 to 7.9 magnitude earthquake near Beijing is about 2.00%About 10 events in the last 500 years. the damage in these events could range from 1% to 10%A significant level of uncertain

8、ty is associated with these methodsThe Simple Method to Exposure 風(fēng)險(xiǎn)定價(jià)的簡單方法以地震為例在北京附近發(fā)生的可以預(yù)見的最大震級地震是8.0級可能的最大損失將達(dá)到大約北京地區(qū)總保額的15%根據(jù)歷史數(shù)據(jù)，該事件發(fā)生的可能性大約是500年一遇這樣的事件在過去500年中只發(fā)生過1次（1679年）在北京周圍發(fā)生6.0 至7.9級地震的可能性大約是2.00% （50年1遇）在過去500年中發(fā)生過大約10次. 這些事件可能造成的損失介于北京地區(qū)總保額的1%至10%如果具備北京地區(qū)的風(fēng)險(xiǎn)累計(jì)數(shù)字，就可以得到這些事件的大約損失估計(jì)這些方法存在很大

9、的不確定性風(fēng)險(xiǎn)定價(jià)的簡單方法以地震為例在北京附近發(fā)生的可以預(yù)見的最大China Catastrophe ModelsAvailable Models“Detailed” and “Aggregate” models licensed from key global vendors:AIR Worldwide Corporation (AIR)CLASIC/2 (Detailed) & CATRADER versions 10.0 (Aggregate)Perils: Earthquake & TyphoonRisk: PropertyEQECAT (EQE) WORLDCATenterpris

10、e version 3.10 (Aggregate)Perils: Earthquake & TyphoonRisk: PropertyRisk Management Solutions (RMS) RiskLink version 8.0 (Detailed)Perils: EarthquakeRisk: Property & Casualty* Images supplied by Guy Carpenters China Catastrophe ModelsAvail中國巨災(zāi)模型可用的模型經(jīng)世界主要模型公司授權(quán)的“具體” 和 “累計(jì)” 模型 AIR Worldwide Corporati

11、on (AIR)CLASIC/2 (具體) & CATRADER versions 10.0 (累計(jì))模擬災(zāi)害: 地震 & 臺風(fēng)風(fēng)險(xiǎn)類別: 財(cái)產(chǎn)EQECAT (EQE) WORLDCATenterprise version 3.10 (累計(jì))模擬災(zāi)害: 地震 & 臺風(fēng)風(fēng)險(xiǎn)類別: 財(cái)產(chǎn)Risk Management Solutions (RMS) RiskLink version 8.0 (具體)模擬災(zāi)害: 地震風(fēng)險(xiǎn)類別: 財(cái)產(chǎn)，傷亡* Images supplied by Guy Carpenters 中國巨災(zāi)模型可用的模型經(jīng)世界主要模型公司授權(quán)的“具體” 和Rising Expectatio

12、ns in Catastrophic Risk ManagementRegulatory ViewCIRCCircular No. 402 (April 2007), in response to increased catastrophic risk, companies are encouraged to (among other things):Appropriately purchase catastrophic reinsurance to leverage international reinsurance sourcesLeverage tools developed by th

13、ird parties to gradually improve the state of catastrophic risk modeling in ChinaGlobally move to define appropriate capital levels through risk-based capital measuresEuropean Union through Solvency 2 regulation is setting target capital to be 99.5% VaR (1 in 200 year worst outcome)Scenarios would g

14、enerally involve a catRising Expectations in Catastr對巨災(zāi)風(fēng)險(xiǎn)管理的日益提升的期望從監(jiān)管的角度中國保監(jiān)會(huì)2007年4月第402號通知用以應(yīng)對日益頻繁的巨災(zāi)風(fēng)險(xiǎn)，鼓勵(lì)保險(xiǎn)公司 (但不限于):妥善安排巨災(zāi)再保險(xiǎn)，充分利用再保險(xiǎn)資源利用外界開發(fā)的工具， 來逐步改進(jìn)中國的巨災(zāi)風(fēng)險(xiǎn)模型 從全球來講，對資本充足性的判定已逐步轉(zhuǎn)向以利用風(fēng)險(xiǎn)資本的測量為依據(jù)歐盟通過的Solvency 2 監(jiān)管規(guī)定，將資本標(biāo)準(zhǔn)設(shè)定在99.5%的在險(xiǎn)價(jià)值水平（換言之就是200年一遇的水平）不同的狀況通常都考量了巨災(zāi)的因素對巨災(zāi)風(fēng)險(xiǎn)管理的日益提升的期望從監(jiān)管的角度中國保監(jiān)會(huì)Ris

15、ing Expectations in Catastrophic Risk ManagementRating Agency ViewCatastrophic risk a key element in capital solvency calculationA.M. Best uses greater of 1/100 year wind or 1/250 year earthquake net cat loss amount as a benchmarkS&P uses 1 in 250 year aggregate net loss amount as a benchmarkA.M. Be

16、st view of a strong cat risk management programEmphasis on data quality Monitoring exposure throughUsing one or more catastrophe modeling toolsReviewing aggregate loss exposure accumulationConsidering potential loss exposure scenarios in addition to modeled outputManaging to realistic loss scenarios

17、, not the lowest case loss estimatesImplementing controls, including:Establishing specific aggregate limits using a reasonable and defensible basisPurchasing an appropriate reinsurance program Integration of catastrophe management into the underwriting processRising Expectations in Catastr對巨災(zāi)風(fēng)險(xiǎn)管理的日益

18、提升的期望值從監(jiān)管的角度巨災(zāi)風(fēng)險(xiǎn)是測算資本及償付能力的關(guān)鍵要素A.M. Best 利用100年一遇風(fēng)災(zāi)或者250年一遇地震可能產(chǎn)生的凈自留損失，兩者取高者為準(zhǔn)，來作為衡量的標(biāo)準(zhǔn)S&P 利用250年1遇累計(jì)凈自留損失金額來作為衡量的標(biāo)準(zhǔn)A.M. Best 判斷一個(gè)好的巨災(zāi)風(fēng)險(xiǎn)管理項(xiàng)目時(shí)強(qiáng)調(diào)數(shù)據(jù)質(zhì)量監(jiān)測風(fēng)險(xiǎn)的方法包括：使用一個(gè)或一個(gè)以上的巨災(zāi)模型工具監(jiān)測風(fēng)險(xiǎn)累積情況在模型結(jié)果的基礎(chǔ)上，考慮其他可能的損失情境考察真實(shí)可能的損失狀況, 而不是最低的損失評估監(jiān)控, 包括:利用合理和站得住腳的方法建立詳細(xì)的累積限額購買適當(dāng)?shù)脑俦ｋU(xiǎn)保障將巨災(zāi)管理整合于核保流程當(dāng)中對巨災(zāi)風(fēng)險(xiǎn)管理的日益提升的期望值從監(jiān)管的角度

19、巨災(zāi)風(fēng)險(xiǎn)是測算Data QualityRequirements for Earthquake Location dataAIRs analysis is run at district level (approximately 2,900 districts)The RMS model can be run at any level (down to postal code) (approximately 33,000)Data QualityRequirements for 數(shù)據(jù)質(zhì)量以地震模型數(shù)據(jù)要求為例地址信息：AIR的分析可以達(dá)到區(qū)一級的水平 (全國大約 2,900 個(gè)區(qū))RMS則可以達(dá)

20、到6位郵政編碼水平 (全國大約33,000個(gè)郵政編碼 )數(shù)據(jù)質(zhì)量以地震模型數(shù)據(jù)要求為例地址信息：Data QualityRequirements for Earthquake, ContinuedConstructionAIR has 46 construction typesRMS has 40 unique construction typesOccupancyAIR has 5 unique occupancy types given a known construction type RMS has 5 unique occupancy types given a known cons

21、truction typeYear built(reflects changing building codes)Building height / number of storiesData QualityRequirements for 數(shù)據(jù)質(zhì)量以地震模型數(shù)據(jù)要求為例（續(xù)）建筑類型AIR 有46種建筑類型RMS 有40種建筑類型行業(yè)類別AIR 針對每一個(gè)建筑類型都有5個(gè)單獨(dú)的行業(yè)類別RMS針對每一個(gè)建筑類型都有5個(gè)單獨(dú)的行業(yè)類別建筑年份反映不同時(shí)期的建筑標(biāo)準(zhǔn)建筑高度/樓層數(shù)量數(shù)據(jù)質(zhì)量以地震模型數(shù)據(jù)要求為例（續(xù)）建筑類型Data QualityLoss Estimates by Constr

22、uction Class and Building Age Comparisons1M RMB Buildings Exposure in Beijing, “Unknown” Occupancy Earthquake Model, Average Annual Loss30% difference because of year builtData QualityLoss Estimates by數(shù)據(jù)質(zhì)量比較不同的建筑結(jié)構(gòu)和年限下的預(yù)估理賠額在北京一百萬的大廈保額, “未知”行業(yè)地震模型,年平均損失因?yàn)槟晗薏煌?，預(yù)估年平均損失相差30%數(shù)據(jù)質(zhì)量比較不同的建筑結(jié)構(gòu)和年限下的預(yù)估理賠額在北京一百

23、萬Understanding UncertaintyDamage range 1% - 90% Understanding UncertaintyDamag理解不確定性Damage range 1% - 90% 理解不確定性Damage range 1% - 90% Ratio of gross AAL to premium or TIV Explore profitability issues by product, province, city, etc., down to the location levelMonitoring Exposure and Implementing Con

24、trolsIntegration of Cat Modeling into Underwriting ProcessCat Loss Ratios ExampleRatio of gross AAL to premium 年平均損失與保費(fèi)或者總保額之間的比率關(guān)系按險(xiǎn)種，省份，市直到具體地點(diǎn)來考察盈利監(jiān)控風(fēng)險(xiǎn)將巨災(zāi)模型整合到核保流程（以巨災(zāi)損失率為例）年平均損失與保費(fèi)或者總保額之間的比率關(guān)系監(jiān)控風(fēng)險(xiǎn)將巨災(zāi)模型27Managing to Realistic ScenariosMetaRisk (Guy Carpenter DFA tool) Timeline Simulation29Managin

25、g to Realistic Scenar28管理現(xiàn)實(shí)情境MetaRisk (Guy Carpenter 動(dòng)態(tài)財(cái)務(wù)分析工具) 時(shí)間軸模擬30管理現(xiàn)實(shí)情境MetaRisk (Guy CarpentImplementing ControlsReinsurance ExampleRisk Management CostRisk Management BenefitBreak EvenImplementing ControlsReinsura風(fēng)險(xiǎn)控制以再保險(xiǎn)為例風(fēng)險(xiǎn)管理成本風(fēng)險(xiǎn)管理收益持平點(diǎn)風(fēng)險(xiǎn)控制以再保險(xiǎn)為例風(fēng)險(xiǎn)管理成本風(fēng)險(xiǎn)管理收益持平點(diǎn)Event MonitoringEvent Monitorin

26、g事件監(jiān)測事件監(jiān)測Video DemonstrationVideo DemonstrationofVideo DemonstrationVideo Demon視頻演示視頻演示Instrat Center of Analytical ExcellenceHistoryFirst reinsurance broker analytical unit founded 20 years agoCurrentlyOver 200 actuaries, cat modelers, financial analysts, scientists, software development20 unified

27、global locationsGrowing resource throughout Asia PacificIntellectual and technical resource for Guy CarpenterInstrat Center of Analytical EInstrat，分析優(yōu)勢中心所在歷史成立于20年前，第一個(gè)再保經(jīng)紀(jì)人獨(dú)立擁有的量化分析團(tuán)隊(duì)現(xiàn)在擁有超過200名精算師，巨災(zāi)模型師，財(cái)務(wù)分析師，科學(xué)家以及軟件開發(fā)人員分布于全球20個(gè)辦公室在亞太區(qū)的團(tuán)隊(duì)逐步壯大是Guy Carpenter重要的智慧和技術(shù)資源Instrat，分析優(yōu)勢中心所在歷史Disclaimer for

28、PresentationThe data and analysis provided by Guy Carpenter herein or in connection herewith are provided for the participants (the Participants) of the 9th Chinese Actuarial Annual Conference “as is,” without warranty of any kind whether express or implied. The analysis is based upon data provided

29、by or obtained from external sources, the accuracy of which has not been independently verified by Guy Carpenter. Neither Guy Carpenter, its affiliates nor their officers, directors, agents, modelers, or subcontractors (collectively, “Providers”) guarantee or warrant the correctness, completeness, currentness, merchantability, or fitness for a particular purpose of such data and analysis. The data and analysis are intended to be used solely for the purpose of the Conference and Participants shall not disclose the work product to any third party, except its reinsurer