Download - 2015.10.05 ANRA Session Paolo Bazzuro
MODELLING AND QUANTIFICATIONOF THE ECONOMICIMPACT OF EXTREME CLIMATE-RELATED RISKSProf. Paolo BazzurroEmail: [email protected]
Institute for Advanced Study IUSS PaviaWorkshop
BRUSSELS, 20-21 October www.ferma.eu
FORUM 2015Venice, Italy 4-7 October
What is risk? Risk to Whom or to What? How is Risk
Measured? Exposure
Population Buildings (residential,
commercial, public) Infrastructure (bridges, dams,
ports, etc.) Crops and Livestock
Impact Fatalities, injuries Monetary losses Resiliency
BRUSSELS, 20-21 October www.ferma.eu
FORUM 2015Venice, Italy 4-7 October 3
Why Are Cat Risk Models Built?
To support decisions about risk management
Risk Retention Risk Mitigation and Control Risk Financing/Transfer Risk Avoidance
BRUSSELS, 20-21 October www.ferma.eu
FORUM 2015Venice, Italy 4-7 October
IndividualPolicy Holder
Retro MarketPrimary Insurance Market
Buys
PrimaryPolicy
ReinsuranceContract
CorporatePolicy Holder
Sells SellsBuys
Insurer Reinsurer
Cedes
Reinsurance Market
CatastropheBond
SPV(Special Purpose
Vehicle)
CedesCapital Markets(Hedge Funds, Money
Managers, etc.)Buys Sells
Invests Broker
Cat Risk Models Serve All the Segments of the Insurance Industry
4
BRUSSELS, 20-21 October www.ferma.eu
FORUM 2015Venice, Italy 4-7 October
Who Else Uses CatastropheRisk Models?
• Catastrophe/relief funds and risk pools (e.g., California Earthquake Authority, Caribbean Catastrophe Risk Insurance Facility, Pacific Catastrophe Risk Insurance Pilot)
• Rating Agencies (S&P, Moody’s, Fitch)• Real estate management and investment• Mortgage lending Institutions• Major banks (WB, ADB, IADB, etc.)• In US, Fannie Mae, Freddie Mac and the secondary mortgage market• Major Corporations (Dow Chemical, Devon Energy, Arkema, Sony,
General Motors, etc.)• Governments and their departments (U.S. Department of Homeland
Security, the U.S. Navy, the Risk Management Agency, a division of the U.S.D.A., Mexican Government, etc.)
5
BRUSSELS, 20-21 October www.ferma.eu
FORUM 2015Venice, Italy 4-7 October
Why Do Corporations Need Catastrophe Risk Models?
• Accurately quantify, help mitigate, and manage catastrophe risk
• Are the conventional Catastrophe Risk Models enough?• Catastrophe Risk Engineering Models more appropriate
Primary needs– Risk assessment and mitigation
• Explicitly includes business interruption– Post-catastrophe response
• Damage assessment and repair concepts
6
BRUSSELS, 20-21 October www.ferma.eu
FORUM 2015Venice, Italy 4-7 October
A Brief History of Catastrophe Modeling in the Insurance
Industry
US Hurricane, Tornado &
Earthquake Models
• Introduced
Late1980s 2000
First Catastrophe Bond Issued Transferring
Risk to Capital Markets
Introduction of Performance-
Based Earthquake Engineering
(PBEE)
Mid- 1990s
Cat Models capture the
attention of the industry
Hurricane Andrew (1992)
• $16B
Northridge EQ (1994)
• $12B
1992,1994
• World Trade Center Attack
• First Terrorism, Workers’
Comp Models
Introduced One Year
Later2001,2002
Decade of the 00s: Introduction of Wildfire, Winter Storm, Crop Loss and Offshore Assets
(Gulf of Mexico), Pandemic Flu Models
2004,2005 2009
11 Insolvencies after Hurricane
Andrew and several after Northridge
3 Insolvencies after Hurricane
Katrina
Record Breaking Hurricane
Seasons of 2004 & 2005
Hurricane Katrina (2005)
• $41B
2006
Introduction of First Hurricane
Models Conditioned on
Warm Sea-Surface
Temperatures
M O D E L D E V E L O P M E N T T I M E L I N E
Introduction of Tsunami Model for the Pacific
20122011
Tohoku Earthquake
7
BRUSSELS, 20-21 October www.ferma.eu
FORUM 2015Venice, Italy 4-7 October 8
What is the Coverage Worldwide of Cat Risk Models for the
Insurance Industry? Most of them are for the developed parts of the world or
for fast-growing countries where the insurance industry is mature
A large portion of the world (i.e., most of the developing countries) is currently left out
Things are slowly changing Sovereign Risk Financing Initiatives (Caribbean, Pacific Islands
Countries, South America, Mexico, etc.) The Global earthquake Model (GEM) (http://
www.globalquakemodel.org/) CAPRA (http://www.ecapra.org/)
8
BRUSSELS, 20-21 October www.ferma.eu
FORUM 2015Venice, Italy 4-7 October
What is Risk Assessment?
Hazard Exposure Vulnerability Impact
Cost to Government, Business and
Society
Fatalities, injuries, displaced persons
Damage to buildings,
infrastructure
Risk = probability of damage and loss to population and assets (Loss Valued in Human and Financial terms)
9
BRUSSELS, 20-21 October www.ferma.eu
FORUM 2015Venice, Italy 4-7 October
The Objectives Dictate the Risk Analysis Approach
•
Incr
easi
ng E
ffort
& C
ompl
exity
Hazard/Risk Index or Profile
Pros: fast, low data requirements, output easy to understand
Cons: Low resolution, subjective
Historical ScenarioPros: based on event-specific data, good for frequent hazards
Cons: misses extreme events, potential impacts of climate change
Probabilistic
Pros: accounts for both frequent/low-impact and rare/extreme events
Cons: high data/expertise requirements, need to ensure outputs can be understood
10
BRUSSELS, 20-21 October www.ferma.eu
FORUM 2015Venice, Italy 4-7 October
#11
The risk modeling approach is conceptually simple
1. Identify Study
Region
3. Define and Overlay
Exposure
5. Estimate Losses
4. Model Damage
2. Characterize
Hazard
11
BRUSSELS, 20-21 October www.ferma.eu
FORUM 2015Venice, Italy 4-7 October
#12
1. Identify Study
Region
3. Define and Overlay
Exposure
• 5. Estimate Losses
4. Model Damage
2. Characterize
Hazard
The risk modeling approach is conceptually simple
#12
BRUSSELS, 20-21 October www.ferma.eu
FORUM 2015Venice, Italy 4-7 October
#13
Identification of the Study RegionFlood hazard modeling at a regional
scale Inaccuracies often tolerated:• Poor accuracy digital
elevation model• Presence of levees,
dykes, and adoption of flood control infrastructures,
• disregard of tide effects
• If levees are modeled, they are added after the inundation area is computed
• Hydraulic models replaced by purely statistical models
13
BRUSSELS, 20-21 October www.ferma.eu
FORUM 2015Venice, Italy 4-7 October
#14
Flood Hazard Modeling at an Urban Scale: none of these
simplifications would be acceptable
14
BRUSSELS, 20-21 October www.ferma.eu
FORUM 2015Venice, Italy 4-7 October
#15
1. Identify Study
Region
3. Define and Overlay
Exposure
• 5. Estimate Losses
4. Model Damage
2. Characterize
Hazard
The risk modeling approach is conceptually simple
15
BRUSSELS, 20-21 October www.ferma.eu
FORUM 2015Venice, Italy 4-7 October
Hazard Characterization – Tropical Cyclone Peril
Hazard
• HAZARD
Intensity Calculation
Hurricane Occurrence
Where?How Strong?
How frequent?
Generating the stochastic catalog of potential future
events
16
BRUSSELS, 20-21 October www.ferma.eu
FORUM 2015Venice, Italy 4-7 October
#17
1. Identify Study
Region
3. Define and Overlay
Exposure
• 5. Estimate Losses
4. Model Damage
2. Characterize
Hazard
The risk modeling approach is conceptually simple
17
BRUSSELS, 20-21 October www.ferma.eu
FORUM 2015Venice, Italy 4-7 October
• $
Physical Location, Size and Shape
Administrative Area
Land Ownership
Land Cover and Land Use
Construction Period
Structural Characteristics
Demographic or Social Characteristics
Economic Characteristics
EXPOSUREDATABASE
Building and maintaining an integrated, centralised and spatially-enabled database is the most effective way to manage exposure information
Assembling an Exposure Database
18
BRUSSELS, 20-21 October www.ferma.eu
FORUM 2015Venice, Italy 4-7 October
The Scale of the Risk Assessment and the Peril Dictate the Exposure Data
Collection Regional vs. urban vs. site-specific studies
19
Level of detail increases
• Perils with very localized effects. The impact of flood, tsunami, landslide, rockslide, hail risk studies require a more detailed exposure database than risk studies for other perils (e.g., earthquake ground shaking and wind)
BRUSSELS, 20-21 October www.ferma.eu
FORUM 2015Venice, Italy 4-7 October
Building Digitized and Field Verified: Nadi, Fiji
Specific location of buildings is necessary for urban scale studies and for perils with localized effects
20
BRUSSELS, 20-21 October www.ferma.eu
FORUM 2015Venice, Italy 4-7 October
Clusters of Buildings Within Delineated Polygons: Appropriate for Regional Studies and Perils without (or with Limited) Localized Effects
21
BRUSSELS, 20-21 October www.ferma.eu
FORUM 2015Venice, Italy 4-7 October
#22
1. Identify Study
Region
3. Define and Overlay
Exposure
• 5. Estimate Losses
4. Model Damage
2. Characterize
Hazard
The risk modeling approach is conceptually simple
22
BRUSSELS, 20-21 October www.ferma.eu
FORUM 2015Venice, Italy 4-7 October
Mean Damage Function for California Wood Frame Buildings of given Vintage based on Claims Data (1994
M6.7 Northridge Earthquake) • Company
• Company
23
BRUSSELS, 20-21 October www.ferma.eu
FORUM 2015Venice, Italy 4-7 October
#24
1. Identify Study
Region
3. Define and Overlay
Exposure
5. Estimate Losses
4. Model Damage
2. Characterize
Hazard
The risk modeling approach is conceptually simple
24
BRUSSELS, 20-21 October www.ferma.eu
FORUM 2015Venice, Italy 4-7 October
Catastrophe Models Provide a Wide Range of
OutputsExceedance Probability (EP) Curve - Occurrence
0%1%2%3%4%5%6%7%8%9%
10%
0 50 100 150 200 250 300 350 400Loss Amount ($ millions)
Exc
eeda
nce
Pro
babi
lity
0
50
100
150
200
250
300
350
10 20 50 100 250 500 1,000
Estimated Return Period
Loss
Am
ount
($m
illio
ns)
25
BRUSSELS, 20-21 October www.ferma.eu
FORUM 2015Venice, Italy 4-7 October
Standard Output is the Loss Annual Exceedance Probability CurveExceedance Probability (EP) Curve - Occurrence
0%1%2%3%4%5%6%7%8%9%
10%
0 50 100 150 200 250 300 350 400Loss Amount ($ millions)
Exc
eeda
nce
Pro
babi
lity
0
50
100
150
200
250
300
350
10 20 50 100 250 500 1,000
Estimated Return PeriodLo
ss A
mou
nt ($
mill
ions
)
The Average Annual Loss (AAL) is the mean, or average, loss over a long period of time
CAUTION!
26
BRUSSELS, 20-21 October www.ferma.eu
FORUM 2015Venice, Italy 4-7 October
How Do Parametric Insurance and Catastrophe Bonds Fit in the Cat Risk
Modeling Framework?• HAZARD
•ENGINEERING
• FINANCIAL
Intensity Calculation
Exposure Information
Damage Estimation
Event Generation
Policy Conditions
Landfall Location & Cent. Press.
Local Wind Speed
Insured Loss1st Generation Parametric
2nd Generation Parametric
Indemnity
• Cat-in-a-Box• Carvill Hurricane
Index
• Measured Wind Speed at Concentrations of Exposure
• Indemnity bonds
Insured Loss Calculations
Industry Loss• ILW• PCS Bonds
Modeled Loss & Hybrids• Modeled Loss bonds• Modeled Loss/PCS hybrids• Modeled Loss/Parametric
hybrids 27
BRUSSELS, 20-21 October www.ferma.eu
FORUM 2015Venice, Italy 4-7 October
Beyond Economic Losses:Integrated Risk
• Khazai et al.( 2011)
RISK can be further factored by social, economic, political factors that influence vulnerability to obtain a risk index.
28
Institute for Advanced Study IUSS PaviaWorkshop
Prof. Paolo Bazzurro
IUSS PaviaInstitute for Advanced StudyPalazzo del BrolettoPiazza della Vittoria, 1527100 Pavia – ITALY
Web: www.iusspavia.itEmail: [email protected]