WIXLIB

ICLR Friday ForumCANADIAN CATASTROPHE RISK & MODELING –2012 AND BEYONDSeptember 14th, 2012Anya Sri-Skanda-Rajah, FCAS, FCIAKen Mok, CRM, FCIP

Catastrophe Modeling – 2012 and BeyondAgenda Current landscape: Recent events and the state of the market Model availability and usage for Canadian perils Model enhancements expected in the future Managing model change and uncertainty Model result blending and customizationGUY CARPENTER1

Section 1Current Landscape: Recent Events and theState of the Market

Current Landscape2011: Active Cat Year for CanadaDateMarch 5-7April 14 - May 31April 27-28May 14-17July 18-19August 21August 28-30November 27PlaceQuebec, OntarioManitobaOntario, QuebecAlbertaAlberta, Manitoba, SaskatchewanOntarioNew Brunswick, Ontario, QuebecCalgary, AlbertaInsured LossEvent(CAD)Winter Weather, Heavy Rainfall50,000,000Assiniboine River Floods160,000,000Storms with gusts up to 100km/hr, hail210,000,000Slave Lake wildfire700,000,000Thunderstorms, heavy winds, rain, hail, tornadoes 185,000,000F3 tornado, hail, winds, flooding135,000,000Hurricane Irene post tropical flooding, wind130,000,000Windstorm up to 149 km/hr200,000,0001,770,000,000Source: Swiss Re 2011 Sigma Report 1998 and 2009 were higher– 1998 2.3 bln CAD in 2011 dollars- Primarily Jan ’98 Quebec Icestorm– 2009 2.1 bln CAD in 2011 dollars- Primarily 3 Alberta weather eventsGUY CARPENTER3

Current LandscapeGlobal Cat Events2011 and 2012 Compared at June 30th 76 Billion 11 BillionGUY CARPENTER4

Current LandscapeEvents Shape Catastrophe MarketJanuary 2011Catastrophe Cover RenewalsHardGUY CARPENTERSoftAuto/CasualtyCatastropheProperty RiskUpward PricingLimited CapacityFew MarketsDownward PricingStrong CapacityMany MarketsDownward PricingStrong CapacityMany Markets5

Current LandscapeEvents Shape Catastrophe MarketJanuary 2012Catastrophe Cover RenewalsHardSoftCatastropheProperty RiskAuto CasualtyUpward PricingLimited CapacityMany MarketsEven PricingGood CapacityMany MarketsEven PricingGood CapacityGrowing Markets The global market was well positioned to absorb big hits in 2011 Cat pricing hardened in Canada more than some had anticipated Exposures were up, limits increased, and the cat spend grewGUY CARPENTER6

Current LandscapeEvents Shape Catastrophe MarketAnticipated Mid 2012 to January 2013Catastrophe Cover RenewalsHardGUY CARPENTERSoftCatastropheProperty RiskAuto CasualtyFlat to Soft PricingLimited CapacityMany MarketsEven PricingGood CapacityMany MarketsEven PricingGood CapacityGrowing Markets7

Section 2Model Availability for Canadian Perils

What Questions Are Catastrophe ModelsDesigned to Answer? Where are future events likely to occur? How intense are they likely to be? For each potential event, what is the estimated range of damage andinsured loss?Catastrophe models are designed to estimate the probability of loss severity. Theyare not intended to forecast future events.GUY CARPENTER9

2008Catastrophe Model EvolutionChanging Risk Landscape 2005FinancialCrisisCat htRe-adjust Frequency,Severity assumptionsRating Agency Increases CapitalRequirementsWTC 9/11Loss Correlation / Data Rapidly ImprovingCapital Markets Enter into ReinsuranceHurricaneAndrewModels Go Worldwide / InfluenceRate AdequacyRating Agencies EmpoweredPoor Data / Limited ModelsWith each market turning event, the industry realized it had more exposure thanpreviously believedGUY CARPENTER10

Catastrophe Modeling and Model Vendors Founded at Stanford University in 1988 World's leading provider of products and services for thequantification and management of catastrophe risks Grew in the 1990s, expanding services and perilscovered Founded in 1987 Pioneered the probabilistic catastrophe modelingtechnology Founded in 1980s One of first catastrophe models inindustryOther modelsGUY CARPENTER Most large reinsurers and other risk managementcompanies have developed their own in-house models11

Current Canadian Licensed Modeling Capabilities– RMS- Earthquake- Fire-Following Earthquake- Severe Convective Storm (Tornado, Hail, Lightning and Straight-LineWinds)- Winterstorm (Freeze, Snow, Wind and Ice)- North Atlantic Hurricane (new to RMS v11.0)– EQECAT- Earthquake- Fire-Following Earthquake– AIR- Earthquake- Fire-Following Earthquake (not for automobiles)- Severe ThunderstormGUY CARPENTER12

Current Model re Following200920052011Severe Storm20082005NAWinter Storm2008NANAHurricane2011NANAModel / PerilGUY CARPENTEREQECAT201113

Catastrophe Model Use in Canada by PerilEarthquake Available Models: RMS, AIR, EQECAT Usage:– RMS is widely used as compared to AIR / EQECAT inthe industry Latest Updates:– RMS: 2009- Western Canada PMLs increased about 30%- Clients generally bought more cat limit and there wereno capacity issues with the reinsurers– EQECAT: 2011– AIR: 2005GUY CARPENTER14

Catastrophe Model Use in Canada by PerilSevere Storm Available models: RMS and AIR Usage:– RMS is widely used as compared to AIR in the industry Latest Updates:– RMS updated their SCS model in 2008 (RiskLink v8.0)- Losses decreased approximately 30% pendinggeography and LOB– AIR's last Severe Storm model update was in 2005.GUY CARPENTER15

Catastrophe Model Use in Canada by PerilWinter Storm and Hurricane Available models: RMS Usage:– RMS only Latest Updates:– Winterstorm initial release in 2008– Hurricane initial release in 2011GUY CARPENTER16

Catastrophe Model Industry LossAverage Annual Loss ( m)RMS Industry Loss Contribution by Province / PerilGUY CARPENTER17

Catastrophe Model Industry LossRMS Industry Loss Contribution by Province / PerilGUY CARPENTER400 Year Return Period OEP ( b)18

Earthquake Industry Loss Comparison!!!!!"!% & &'!!!!"""!GUY CARPENTER""" #"!"#! !""!"!!!!!"19

Severe Storm Industry Loss Comparison&!"!!"!#"""!!""!!!!!! "!""!!!#!% & &'( ) ' &*!"""#GUY CARPENTER!"!"!!"!"!20

Non-Modelled PerilsTsunami Tsunami Risk– Canada working on a NationalHazard Map for Tsunami– Expected to be an area ofinterest for commercial modeldevelopment in the future Canada Tsunami Risk– Vancouver Island– B.C. Coast– Baffin Island– Mackenzie Delta– St. Lawrence Estuary– Atlantic CoastGUY CARPENTERGraphic from Natural Resources Canada21

Non-Modelled PerilsWildfire Slave Lake – 700- 750M CAD Insured Loss– 2nd largest cat loss in Canada history– 50% of wildfires caused by humansPhotograph by: Caezer Ng/The Lakeside Leader, edmontonjournal.comGUY CARPENTER22

Non-Modelled PerilsWildfireSlave Lake Wildfires – May 2011GUY CARPENTER2011 Exposure – i-aXs RealCat23

Non-Modelled PerilsFlood Flood – May/June 2011– Insured Loss- 160M CAD– Total Economic Loss- 815M CADThe Assiniboine River threatens to breach the 18th Street Bridge in Brandon, Man. onMay 11. The province has announced that a controlled breach of a dike is set to occur onMay 12. (David Lipnowski/Canadian Press)GUY CARPENTER24

Section 3Model Enhancements Expected in the Future

Upcoming Model Changes RMS– No Canadian model updates currently planned– Next Generation Platform scheduled for 2014– Respond to 2015 GSC update EQECAT– No Canadian model updates currently planned– RQE scheduled for late 2012– Research efforts underway on liquefaction, underwater landslide,tsunami, windstorm, hail, and flood AIR– Earthquake, Severe Storm, and Winter Storm likely in 2014– New Hurricane model likely in 2014GUY CARPENTER26

Section 4Managing Model Change and Uncertainty

Major Sources of Uncertainty Cat Models Data––––Address InformationMulti locationInsurance to value and other coverage issuesImperfect sight into risk characteristics HazardCourtesy of AIR Worldwide / EERI– Lack of history of large earthquakes limits our ability to simulate them- We rely on indirect sources of information, like GPS measurements orpaleo-seismology (e.g. historical liquefaction)– We know relatively little about seismicity potential in areas of low seismicity(like in Eastern Canada)– Even in areas of high seismicity like southwestern Canada, there is a limitedamount of data and there is room for the unexpected (Japan M9.0 in 2011,for example) Engineering– Limited claims data for catastrophic events– Lack of understanding of structural behavior under severe loadsGUY CARPENTER28

Catastrophe ModelsAn Imprecise ScienceGUY CARPENTERSource: RMS, GC

Catastrophe ModelsAn Imprecise ScienceGUY CARPENTERSource: RMS, GC

Catastrophe ModelsAn Estimation of Model UncertaintyGUY CARPENTERSource: AIR, EQECAT, RMS, GC

Section 5Catastrophe Model Result Blending andCustomizationGUY CARPENTER

Current Cat Modeling ApproachCLIENT EXPOSURESDATA QASINGLE MODEL ANALYTICSCAT REPORTVENDOR IMPLIED VIEW OF RISKGUY CARPENTER3333

The Future: Guy Carpenter’s Customized Cat Modeling ApproachCLIENT EXPOSURESDATA QA AND AUGMENTATIONMODELED PERILSNON-MODELED PERILS (NMP)Superior Understanding of ModelsMULTI-MODELANALYTICSObtain availableviews of riskMODEL SUITABILITYANALYSIS (MSA)Develop appropriate view ofrisk with objective testingPARAMETERIZATION &SCENARIO ANALYSESComplement view of risk fornon-modeled perilsSuperior DocumentationCAT REPORTMSA REPORTNMP REPORTCLIENT’S OWN COMPREHENSIVE VIEW OF RISKGUY CARPENTER3434

Guy Carpenter’s Model Suitability Analysis (MSA)SMClient’s View of RiskEVALUATIONINTEGRATIONCOMMUNICATIONCOMPONENT 1COMPONENT 4COMPONENT 7SENSITIVITYTESTINGMSA GRIDDOCUMENTATIONCOMPONENT 5COMPONENT 8MODELENHANCEMENTKNOWLEDGESHARINGCOMPONENT 2LOSSVALIDATIONCOMPONENT 6COMPONENT 3SCIENTIFICAPPRAISALGUY CARPENTERRISKCUSTOMIZATION3535

Guy Carpenter’s Model Suitability Analysis (MSA)SMClient’s View of RiskINTEGRATIONCOMMUNICATIONCOMPONENT 4COMPONENT 7MSA GRIDDOCUMENTATIONGC Client DefinedGUY CARPENTERC1: STC2: LVC2: LVC3: SAC3: SARelative RCBldg CodeKlaus LossValidationHi-Freq EPValidationAgreementDmg FuncsAgreementEvent FtprtsMODEL 1GOOD10%ERRORSO-SOMATCHMATCHESRESEARCHMATCHESUK METMODEL 2GOOD200%ERRORPOORMATCHNO MATCHMATCHESUK METMODEL 3NO LATESTAGE BAND50%ERRORSO-SOMATCHMATCHESCLAIMSNO MATCHUK METMSA C3: SA CHAPTERSII3636

Future Of Catastrophe Risk AnalysisModel Customization– Markets more willingto accept companyspecific views– Rating agenciesallowing – if notencouraging - a“Corporate View”– Next GenerationPlatforms encouragesensitivity testing,transparency andmulti model approachGUY CARPENTER

ConclusionGUY CARPENTER

ConclusionSome Basic Things to Remember Regarding Cat Risk Widespread and deep usage of models is relatively young Models are models: there are many uncertainties in them Data and scientific hypotheses all matter a lot Suitability analyses of models will be increasingly important Don’t assume a catastrophe model is useful – prove it useful yourself orrely on another to helpGUY CARPENTER3939

Recent Briefings and White Papers Responding to Catastrophe Model ChangeGC Briefing, October 2011 Managing Catastrophe Model Uncertainty: Issues and ChallengesGC Analytics White Paper, December 2011 Beyond PML: A 360 Degree View of RiskGC Analytics White Paper, February 2012 Spring Conditions Suggest Tempered Atlantic Hurricane SeasonGC Analytics White Paper, May 2012 Cold Spots Heating Up: The Impact of Insured Catastrophe Losses in NewGrowth MarketsGC Report, September 2012GUY CARPENTER40

mok@guycarp.comGUY CARPENTER41

quantification and management of catastrophe risks Grew in the 1990s, expanding services and perils covered Founded in 1987 Pioneered the probabilistic catastrophe modeling technology Founded in 1980s One of first catastrophe models in industry GUY CARPENTER 11 Other models Most large reinsurers and other risk management