Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
T M Risk Assessment Global Case Study: Catastrophe Loss Modelling Dr. Robert Muir-Wood Chief Research Officer Risk Management Solutions ©2014 Risk Management Solutions, Inc. | Confidential THE ORIGINS OF CATASTROPHE MODELLING History does not contain a sufficient population of catastrophes from which to derive a stable mean loss Or address questions like - what loss can we expect once in a hundred years (on average) So we have to create a richer (‘100,000 year’) set - through generating a large population of virtual catastrophes Each event must be credible Each event has a probability And the whole population has to be a complete representation of the ‘universe of possible events’ How do we do this? CREATING THE SYNTHETIC CATALOGUE Research the best historical catalogue Understand catastrophe historiography Thresholds of completeness by period and region Event prehistory (as from geological evidence) Recognise that history is only one realisation of the possible Most synthetic catalogues are some hybrid of: • Statistical Methods • Dynamical Methods (physics based models) Look for independent measures for calibration – such as regional strain rate for earthquakes or river flows as independent of rainfall extremes. Framework for Earthquake Catastrophe Loss Modelling Generate Stoch. Events Earthquake Ground motion Apply Exposure Calculate Damage Quantify Financial Loss In Major Catastrophes - Loss does not occur in isolation Political Context Economic Context Total Event Impact 90% $ Loss Stochastic Hazard Vulnerability Claiming Financial Loss Annual Probability of Exceedance Quantifying Catastrophe Risk: the Exceedance Probability (EP) Curve Exposure = € 1,000 B Event ID Event 1 Event 2 Event 3 Event 4 Event 5 Event 6 Event 7 ... 5.0% 4.0% 3.0% 2.0% 1.0% € 20 € 40 Annual Cumulative AAL(M) Loss M Probability Rate 0.01% € 0.01 0.01% €100 € 50 0.51% € 0.25 0.5% € 40 0.81% € 0.12 0.3% € 25 0.65% 1.46% € 0.16 € 20 2.36% € 0.18 0.9% € 15 3.36% € 0.15 1.0% € 12 4.86% € 0.18 1.5% ... ... Total € 1.05 Loss € 60 € 80 € 100 Visual Display of EP Curves & Return Period Losses • The EP curve provides a visual interpretation of loss potential • Each point of the curve has an associated threshold and probability of exceedance FLOOD DRIVES 80% OF AAL Yangtze and Pearl River Delta economic zones are high risk YANGTZE RIVER DELTA PEARL RIVER DELTA Economic Zones ©2014 Risk Management Solutions, Inc. Confidential MANAGING TSUNAMI RISK IN CHINA ©2014 Risk Management Solutions, Inc. Confidential CATASTROPHE MODELS DRIVE THE BUSINESS OF CATASTROPHE INSURANCE Underwriting & pricing Accumulation Control Portfolio Management Capital Allocation & Regulatory Reporting Reinsurance Pricing & Structuring Alternative Risk Transfer Event Response THE ‘SECRET SAUCE’ OF CAT MODELLING • Needs strongly interdisciplinary approach (Science, Engineering, Statistics, Insurance) – RMS employs c 100 PhDs • High value for global insurance industry based on risk diversification regulatory reporting requirements and investor disclosure – requires ‘industrial strength’ commercial Cat modelling capability and long term model maintenance • The insurance industry is collecting key ‘scientific’ data on exposures and loss. However claims data are also proprietary (and represent competitive IP). Working with/for insurance industry – claims and loss data are utilized to improve vulnerabilities and test short RP modeled losses. • The key to robust catastrophe modelling is multiple rounds of calibration Optimization of Portfolio Composition Return Efficient Frontier Risk Free Return Risk • Each dot represents a portfolio Mitigating the Risk in a Supply-Chain Network AA EDT 100 yrs 500 yrs BI EDT .02 .02 .02 .01 .01 .01 05 05 05 8.3 CBI ratio 31.5 2.89 2.25 CBI ratio with Inventories 1.75 1.91 2.06 1.62 0 0 0 0 0 0 Inventory 60days Supplier C (Semiconductor) Facility (General Assembly) Suppliers B (Body) Facility Facility Facility Supply Chain Supply Chain Supply Chain Supply Chain with Supply Chain with Inventories Supply Chain with Inventories Inventories 0.02 Annual Probability of Exceedance 15 15 15 0.36 0.015 Suppliers A (Engine) 0.01 0.005 10 010 10 0 20 20 20 10 30 30 30 20 40 40 40 30 50 50 50 40 60 60 60 50 EDT (days) 70 70 70 60 80 80 80 70 90 90 90 80 90 Suppliers A Suppliers B Supplier C 13 In ven tory 60d ays Facility Framework for Probabilistic Earthquake Casualty Modelling Generate Stoch. Events Ground motion Footprints Damage & Collapse Building Stock inventory Human 24 hr Exposure Casualties Framework for Storm Surge/Tsunami Evacuation Modelling Stochastic Flood Footprints Human 24 hr Exposure Evacuation Model Casualties RMS TIME STEPPING SURGE MODELLING FOR 50,000 EVENTS IN US HURRICANE MODEL REGIONAL MESH ©2014 Risk Management Solutions, Inc. TRACK SET WIND FIELD LOCAL MESH FLOOD DEPTHS Confidential All US, All Lines, Ground Up Loss Billions RISKY BUSINESS 2014 COASTAL US ‘INSURABLE’ LOSS UNTIL 2100 200 Wind loss only MSL rise / surge range CMIP5 RCP4.5/8.5 model range CMIP3 A1B model range Wind Only - Historical Activity 180 Wind & Surge (Full) - Historical Activity Wind Only - CMIP5 (RCP4.5) 160 Wind & Surge (Full) - CMIP5 (RCP4.5) Wind Only - CMIP5 (RCP4.5) -Active Phase Regionalisation 140 Wind & Surge (Full) - CMIP5 (RCP4.5) - Active Phase Regionalisation 120 100 www.riskybusiness.org 80 60 40 20 2010 2030 2050 2070 Year ©2014 Risk Management Solutions, Inc. Confidential 2090 2110 2130