Catastrophe Risk Models for Asia from the User Perspective

1
Catastrophe Risk Models for Asia from the User
Perspective
Financing the Risks of Natural Disasters

• George Walker
• Head of Strategic Developments
• Aon Re Australia

2
Hypothetical Case Study
OJUDAKAN
Population 10 Million
Dwellings 2 Million
GDP/Person 15 US
GDP growth 4 / year
Significant Earthquake Typhoon Risk
Faults
Typhoon Tracks
3
Catastrophe Insurance Situation
Insurance Vulnerability
Large Industrial (Multi-National) 100
Low Smaller Industrial/Commercial 40
Moderate Public Infrastructure 0
Low Housing 5 High

• Ojudakan Government under pressure from
  international funding agencies to
• Reduce vulnerability of housing
• Introduce a national disaster insurance scheme

4
Design of Disaster Insurance Schemes
5
Key Output From Loss Risk Analysis
6
Premium Analysis
Market Value Premium Function (Average
Annual Loss, Standard Deviation) Local Factors
7
Sustainability Modelling
Model statistically performance over time
8
Sustainability Analysis Output
Initial Fund Size Zero
Annual Growth Rate PML Premium)
4 Investment return rate 6 Loan
rate 7 Admin Costs 5 Initial Premium
US10/dwelling
No Reinsurance Prob of Ruin 7.2
Full Reinsurance Prob of Ruin 3.6
9
Earthquake Loss Model
10
GIS Typhoon Loss Model
11
Modelling Problem - Hazard Risk
Lack of Reliable Scientific Data
Data Probable Information
Faults Poor Earthquake Records (Mgt5)
Moderate Typhoon Records Moderate Soil
Mapping Poor Flood Mapping
Poor Topographical Mapping Poor
12
Modelling Problem - Portfolio Data

• Information often lacking of national inventory
  of buildings.
• Where information exists likely to be deficient
  in respect of
• Value
• Precise Location often aggregated at
• coarse level
• Building characteristics relevant to
• vulnerability - eg age, construction type,
  
• roof type, number of stories, occupancy type

13
Modelling Problem - Vulnerability

• Information generally lacking on vulnerability of
  local forms of construction
• Further complicated by need to to
• Allow for effect of mitigation measures such
• as building code changes in modelling future
  
• losses
• Be able to model losses when using non-
• standard policy conditions eg total loss
  
• claims only.

14
Consequences
Heavy Reliance on Expert Opinion And Extrapolati
on of 1st World Models

• Result
• Models may not be relevant eg Typhoon loss
  
• models based on wind damage when flooding
• main hazard
• Different models may give widely differing
• answers

15
Example
16
Underlying Issue
17
Thank You