Managing a Global Catastrophe Portfolio CARe May 19, 2005 Managing a Global Catastrophe Portfolio May 19, 2005 Agenda Motivation Model overview: Input data Dependencies Measure of profitability Sensitivity Analysis Architecture Applications: Reporting Portfolio optimization: Scenario analysis Efficient frontier Capital Charge 2 Managing a Global Catastrophe Portfolio May 19, 2005 Motivation to build a portfolio model 1. Dynamic monitoring of Portfolio ROE and Capital deployed 2. Rapid and reliable risk profiles for reporting to internal/external parties 3. Efficient planning, scenario evaluation and portfolio optimisation 4. Evaluation of the capital implications of nonstandard products 3 Managing a Global Catastrophe Portfolio May 19, 2005 Model Overview: input data Data Source on exposure to natural catastrophe Risk Rates and exposure entered by underwriters in operating systems: Give frequency and severity for particular peril/region hitting a layer Very complete inventory of natural perils (300 separate combinations of region and peril modeled) Outputs from Cat models stored in PRECED: Simulated loss for particular natural peril event and for particular cedant Actual model uses a mix of both types of data. 4 Managing a Global Catastrophe Portfolio May 19, 2005 5 Model overview: input data Combination of risk-rate data and cat model output (loss files) allows a very complete description of catastrophe exposure. Percentage of expected loss Modeled perils US quake Japan quake US wind Europe wind Japan wind Total 6% 5% 17% 20% 7% 56% Non / poorly modeled perils Unusual in the industry Other quake Flood Hail Others Total 18% 9% 3% 14% 44% Managing a Global Catastrophe Portfolio May 19, 2005 Model overview: input data Cat model outputs: List of losses to a particular cedant for all events in catalogue of Cat model Loss file The portfolio model handles both our internal CatFocusTM suite of models and commercial models: AIR, RMS, EQECat Primary advantage of using loss files is the ability to aggregate losses across different portfolios 6 Managing a Global Catastrophe Portfolio May 19, 2005 Model Overview: Dependencies Dependencies Achilles heel of any portfolio model Overall capital and its allocation are very sensitive to dependency structure. Methodology for Risk rates: Same peril-Same region: fully correlated Correlation matrices: Atlantic Hurricane, EU Wind, EU Flood based on simulated events/meteorological study. Otherwise Independent Methodology for cat model outputs: Natural correlation via aggregation at the event level Same event may affect different cedants/regions Portfolios within the same region are only partially correlated 7 Managing a Global Catastrophe Portfolio May 19, 2005 Model overview: measure of profitability Overall capital for Cat portfolio Statistical measure on distribution of financial results Use of Tail Value at Risk: Mean of losses exceeding the corresponding VaR 8 Managing a Global Catastrophe Portfolio May 19, 2005 Model overview: measure of profitability Allocation of capital It serves two purposes: Portfolio optimization by over-/under-weighting segments with profitability higher/lower than overall portfolio Calibration of capital charge for different key markets We use contribution to portfolio TailVar Credit for diversification to each segment according to how it correlates with the main risks in the overall portfolio Marginal allocation ensures that profitability at segment level is a good indicator of where to grow/reduce business. 9 Managing a Global Catastrophe Portfolio May 19, 2005 Model overview: sensitivity analysis Sensitivity analysis is essential in order to build confidence in model and to assess its limitations We reviewed the impact of different correlation models on aggregate loss distributions as well as profitability: Dependency structures (copulas) for methodology based on riskrates Correlation inherent in cat model outputs for several models Ultimately we have several views of our portfolio based on different models. 10 Managing a Global Catastrophe Portfolio May 19, 2005 11 Model overview: sensitivity analysis Annual aggregate losses (arbitrary unit) 30 25 20 15 10 5 0 10 100 1000 Return periods (years) Model 1 Model 2 Gaussian model Managing a Global Catastrophe Portfolio May 19, 2005 12 Model overview: architecture Graphical User Interface Process new Report / Visualization / Drives Core Engine Core Engine Generates loss distributions using mixed methodology System Report EL, Cover, etc, for in-force portfolio at particular date Loss files for treaties not in force PM Database Process Report and generated loss curves Loss file DB Loss files and event information GIS Loss files for in-force portfolio Managing a Global Catastrophe Portfolio May 19, 2005 Model overview: architecture 13 Managing a Global Catastrophe Portfolio May 19, 2005 Applications: most damaging event 14 Managing a Global Catastrophe Portfolio May 19, 2005 Applications: scenario analysis for planning Scenario analysis rather than full-blown automatic optimization: Scenario based on underwriters projections rather than theoretical model of rate changes Criteria to define scenario: Total EPI/exposure is fixed. Scenario should increase overall profitability Portfolio should be achievable in practice 15 Managing a Global Catastrophe Portfolio May 19, 2005 Applications: scenario analysis for planning Underwriters’ projections Base portfolio to create other scenarios Realized by applying changes to portfolio in-force as of July 1: Change our share Change ROL Apply changes selectively to key markets and to treaties with similar risk rates. 16 Managing a Global Catastrophe Portfolio May 19, 2005 Applications: efficient frontier Efficient frontier: line on risk-return graph showing optimal portfolios that: maximize profit for a given level of capital or minimize capital for a given profit Assumptions for optimizations: Price elasticity: an increase/decrease in market share will result in an decrease/increase in rates Portfolio profile is similar to the reference portfolio, only shares of different markets vary 17 Managing a Global Catastrophe Portfolio May 19, 2005 Applications: efficient frontier 18 Managing a Global Catastrophe Portfolio May 19, 2005 Capital cost / Limit Applications: calibration of Capital Charge for pricing Expected Loss / Limit 19 Managing a Global Catastrophe Portfolio May 19, 2005 Conclusion Cat portfolio model: Aggregates the exposure to all natural catastrophe risks that affect our in-force portfolio Calculates our risk profile and capital needs on a frequent basis Is applied in: reporting, optimization, and planning 20