RQE EQECAT s new platform. Risk Quantification & Engineering

Transcription

1 RQE EQECAT s new platform Risk Quantification & Engineering

2 EQECAT Models Global Coverage Covering 90% of Global GDP spanning the insured world; 181 peril models for 95 countries Americas/Caribbean: 39 EQ countries 32 TC countries 3 WS countries GOM Offshore Energy US EQ Workers Comp US Flood US Tornado/Hail US Wildfire US Winterstorm Europe: 22 EQ countries 28 WS countries 5 Flood countries Africa/Middle East: 12 EQ countries 4 TC countries 4 WS countries Asia/Pacific: 14 EQ countries 11 TC countries 1 WS country

3 Catastrophe Risk

4 Foundation of CAT Modeling Multi-Disciplinary Research Statistician Meteorologist Structure Engineer Operation Researcher Mathematician Seismologist Actuary Hydrologist Geotech Engineer Software Engineer

5 EQECAT Catastrophe Models A story of innovation Wholly-Owned by EQE EQECAT Transformation begins RQE V 13 Released EQE International Founded EQECAT founded as JV between EQE and GC Professional Service Provider of Year Review EQE International Acquired by ABS Group 1. Professional Service Provider of Year - Review Warm AMO HU Freq. 1 st to use soils based attenuation 1 st to adopt UCERF 1.0 Named "Best Risk Modeling Firm" at 2011 Reactions Global Awards EQE Develops EQ Model for Engineering Support EQE Develops 1 st EQ FF Model Develops 1 st EQ Sprinkler Leakage Model Eurowind based on Historic SLP Records Develops 3D EQ Fault Model/ Spectral Parameters Develops Time Dependency EQ Model Conducts for 1 st Securitization Develops NE Asymmetrical Storm Methodology Helps in Formation of CEA Euroquake Regional model T&H and Winterstorm Models Eurowind Measured Wind Speeds Dynamic Financial 150,000 Year Model Euroflood Aggregate Model Releases WORLDCATenterprise platform Eurowind Hybrid Model Germany Austria Flood in WCe Named "Best Risk Modeling Firm" at 2012 Reactions Global Awards New basin-wide Asia Typhoon (incorporating wind, storm surge, and rainfall flooding) wins Innovation of the Year at Asia Insurance Awards

6 RQE The Principles for RQE Which part of Cat Modeling is important to you? The Treatment of uncertainty Transparency Methodology The Modeling the complexity of the (re)insurance conditions Confidence in the results The Integration into the workflow Scalability for performance

7 RQE - The RQE

8 RQE The Cat modeling deals with uncertainty We improve our understanding in order to Quantify the random element Reduce the epistemic element Following are some examples Types of Uncertainty Epistemic lack of understanding, reduced by better knowledge Aleatoric inherent randomness, can quantify but not reduce

9 UNDERSTANDING AND MODELING OF UNCERTAINTY Understanding uncertainty in Hazard Vulnerability Modeling

10 RQE The : Earthquake Methodology Soil-based attenuation Time dependency World-class scientists Dr. Ken Campbell, Vice President of Model Development: With over 40 years of seismic hazard evaluation experience is one of the world s leading experts on ground motion attenuation, formerly with the USGS, and a current member of the USGS National Seismic Hazard Mapping Advisory Panel Paul Thenhaus, Senior Geologist: With 36 years of seismic hazard evaluation experience, spent the first 17 years of his career with the USGS on the National Seismic Hazard Mapping Team and has served as the lead investigator on numerous seismic hazard evaluations worldwide. Relationships Close collaborative relationships with many of the world s national seismic hazard agencies, including: the United States Geological Survey (USGS), Geological Survey of Canada (GSC), Geoscience Australia (GA), and Geological and Nuclear s New Zealand (GNS), and many others. CEA Exclusive: EQECAT, maintains a close relationship with the California Earthquake Authority (CEA) and remains the only commercially available earthquake model used by the CEA for pricing their policies

11 RQE Earthquake: Soil-Based Attenuation Leading Edge Technology Attenuation is a source of significant uncertainty New methodology developed to avoid unnecessary bias Hollywood Rock (orange & brown) Burbank Soil (yellow & green) Soil-Based Attenuation: Captures exposure better more exposure is located on soil Reflects the measured data more ground motion recordings are located on soil Smaller amplification factors Less uncertainty in losses Downtown LA

12 The : Topical Cyclone Model RQE Pioneers in creating basin wide model- North Atlantic Model and Asia Typhoon Model North Atlantic Hurricane Model Methodology: High resolution, 16 Directional time stepping wind field model Near Term view of risk based on AMP (Atlantic Multi-Decadal Oscillation) Expert review Robert Tuleya, formerly with the GFDL Hurricane group and presently developing and upgrading the next generation HWRF, reviewed EQECAT s hurricane windfield model Dr. Gamil Serag Eldin and Dr. Kashif Ali performed independent reviews of the computer science aspects of the model in 2013 Model Certification EQECAT s North Atlantic Hurricane Model has been successfully certificated by the FCHLPM since the inception of the process in Asia Typhoon Model Widest Geography coverages- 9 territories (7 countries) First to introduce a basin wide model, with detailed winds, storms surge and rainfall flooding model Expert review Prof. Johnny Chan, City Univ. of Hong Kong; Director, Guy Carpenter Asia-Pacific Climate Impact Centre and others Date Stamping for all events within weather peril models.

13 The : Eurowind RQE

14 Windspeed Modeling RQE Eurowind is based on actual windspeeds Uncertainty propagates from hazard to loss Using accurate windspeeds reduces a large element of this uncertainty Vulnerability (%) Loss ( ) Maximum Windspeed: over 75m/s (146Kts) Temperature Range: -50 to +70 C Accuracy: 1% of reading between 10 and 55 m/s, 2% above 55m/s Hazard (m/s)

15 Engineering-Based Vulnerabilities Draw upon ABS years of experience and access to global expert engineers Continuous feedback loop with ABS: Thousands of seismic evaluation studies Over 90 Post Earthquake investigations and retrofitting Offshore energy platform classification EQECAT Catastrophe Modeling Insurance and Financial Communities ABS Consulting Risk Management & Engineering Consulting Derived from & validated by Historical damage & claims Engineering analysis: consistency, comprehensiveness Post event investigations (largest experience database) Expert opinion (code, similarity, local conditions, etc.)

16 RQE - The RQE

17 The RQE Cat modeling is about pricing and capital allocation Can we model the complexities of our policy wordings? Are we capturing the tail of the distribution? How confident am I in the results?

18 Simulation with Discrete Samples RQE 100% Sample Damage and Loss Ratio 90% 80% 70% 60% 50% 40% 30% Sample statistics are easy to calculate sum all samples and divide by the number of samples. In this example, the average damage = 12%, and sigma is about 15%. The loss calculation is performed on each individual damage sample. In this example, Deductible =5% Limit = 25% of TIV 20% GU Damage 10% Gross Loss 0% Sample

19 Modeling of Correlation RQE Any aggregation Level (e.g. Portfolio Level) Event Frequency SRA 300,000 Years

20 Simulated Losses RQE Each event becomes multiple outcomes throughout the simulation period Frequency distribution used to set which years Stochastic event set Severity distribution samples set the loss in each year Year Loss Table (YLT) is list of losses in each year

21 Loss Metrics all derived from YLT RQE Expected annual loss Per occurrence exceedance Annual aggregate exceedance Event Loss Table EAL = Mean sum of losses for all years OEP = Sort vector of Max loss per year AEP = Sort vector of Sum of losses in each year ELT = Mean & SD of losses for each event

22 Portfolio Accumulation RQE YLT-1 Accumulation: Simply add the losses for each event in each year Marginal effects: Simply add or subtract the target YLT Combined YLT YLT-2 YLT-n

23 RQE - The RQE

24 The High Performance RQE Cat Modeling now fits efficiently into workflow RQE is: Configurable : to import data directly Scalable : to obtain high speed efficiency Transparent : to validate inputs and outputs

25 The Enabling RQE A Performance Architecture Scalable and Manageable B Insured Exposure Data A model for today s business Analysis Cluster C Output Results For Integration with Models and tools D Graphical and programmatic interfaces The flexibility needed in today s world Workstation Server

26 RQE Performance Architecture RQE 13 2 clients Java GUI XML API Scalable App Server Wintel Proprietary cluster launch SQL Server dbase Standard edition Future 2 clients Java GUI XML API Scalable App Server Wintel HPC / Azure cloud support SQL Server dbase Standard edition Enterprise edition

27 Performance Scalability RQE Analyses are de-composed into a serialized queue of parallel tasks. Large analyses consist of hundreds of tasks Smaller tasks are run in parallel Analyses are processed in FIFO queues Multiple queues enable the prioritization of jobs

28 RQE RQE Databases System EQECAT Common Basic tables to control licensing and features EQECAT System Read-only tables for features User EDB, IDB, RDB A trio of databases for a single exposure grouping (exposure, results, intermediate binary files) User databases can be shared (copied) between RQE users

29 RQE Fitting the Workflow - Import Convenient import from a variety of sources Source type: Single file Relational fileset SQL database Data format: Spreadsheet (.xlsx,.xlsm) Text file (.txt,.csv) SQL Server database

30 Fitting the Workflow - Templates RQE Use of templates allows minimal mouse clicks!

31 RQE Fitting the Workflow - Automation Embed into the corporate workbench Import exposure Retrieve results Manage databases Full functionality API and associated tools XML-based Implement with.net, VBA, etc. Advantages Less user training Fits into the standard workflow

32 EQECAT s Office Locations San Francisco Bay Area (Oakland, CA) Global headquarters; Executive, model and product development New York Area (Hackensack, NJ) Executive, client development, product support Irvine, California Software development London, UK Client development, product support Paris, France Model development, client development Tokyo, Japan Client development, product support India Software development Oakland, CA Hackensack, NJ London, UK

33 Thank You