catastrophe modeling session reinsurance boot camp

48
Catastrophe Modeling Catastrophe Modeling Session Session Reinsurance Boot Camp Reinsurance Boot Camp August 10, 2009 August 10, 2009 Aleeza Cooperman Serafin Guy Carpenter & Co, LLC

Upload: wyome

Post on 03-Feb-2016

50 views

Category:

Documents


0 download

DESCRIPTION

August 10, 2009. Catastrophe Modeling Session Reinsurance Boot Camp. Aleeza Cooperman Serafin Guy Carpenter & Co, LLC. The Black Box. Cat Modeling. Presentation Outline. What are catastrophe models? How do catastrophe models work? Cat modeling process Understanding model output - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: Catastrophe Modeling Session Reinsurance Boot Camp

Catastrophe Modeling SessionCatastrophe Modeling SessionReinsurance Boot CampReinsurance Boot Camp

August 10, 2009August 10, 2009

Aleeza Cooperman Serafin

Guy Carpenter & Co, LLC

Page 2: Catastrophe Modeling Session Reinsurance Boot Camp

2

The Black Box

Page 3: Catastrophe Modeling Session Reinsurance Boot Camp

3

Cat Modeling

Page 4: Catastrophe Modeling Session Reinsurance Boot Camp

4

Presentation Outline

What are catastrophe models?

How do catastrophe models work?

Cat modeling process

Understanding model output

How is model output used?

Questions - throughout

Page 5: Catastrophe Modeling Session Reinsurance Boot Camp

What are Cat Models?

Page 6: Catastrophe Modeling Session Reinsurance Boot Camp

6

Catastrophe Modeling and Model Vendors

What?–A tool that quantifies risk

How?–Examines insured values that are exposed to catastrophic perils such as hurricanes, earthquakes and terrorism

Why?–Aids management decision making on

Pricing and underwriting Reinsurance buying Rating Agencies Portfolio management

Page 7: Catastrophe Modeling Session Reinsurance Boot Camp

7

Catastrophe Model Vendors

• Founded in 1987

• Pioneered the probabilistic catastrophe modeling technology

• Founded at Stanford University in 1988

• World's leading provider of products and services for the quantification and management of catastrophe risks.

• Grew in the 1990s, expanding services and perils covered.

• Founded in 1980s

• One of first catastrophe models in industry

Other models

• Most large reinsurers and other risk management companies have developed their own in-house models

Page 8: Catastrophe Modeling Session Reinsurance Boot Camp

8

Modeled Perils

Hurricane– Wind and rain– Demand Surge (Loss Amplification) and Storm Surge

Earthquake– Shake– Fire Following– Demand Surge and Sprinkler Leakage

Other wind

Winter storm

Terrorism

Flood (Europe)

Wildfire

Page 9: Catastrophe Modeling Session Reinsurance Boot Camp

9

Deterministic Model–Modeling using a single

discrete event–The event is assumed to

happen without regard to probability

–Commonly seen as recreations of historic events or single- hypothetical analysis

Probabilistic Model–Uses a series of

simulated events and accounts for the probability of those events over time

Types of Models

Page 10: Catastrophe Modeling Session Reinsurance Boot Camp

10

Modeled Lines of Business

Personal lines property

Commercial lines property

Industrial property

Builders Risk

Marine

Auto physical damage (Personal Auto)

Workers compensation

Lives at risk – Accident and Health

Page 11: Catastrophe Modeling Session Reinsurance Boot Camp

11

Building/Vessel/Vehicle

– Other structures

Contents

– Stock

– Machinery

– Inland marine

– Marine

Time Element

– Business Interruption

– Loss of Use

Head Count

Payroll

Modeled Coverages

Page 12: Catastrophe Modeling Session Reinsurance Boot Camp

12

FTP Site – used to transfer files to clients & markets

Transmittal Document – includes instructions for accessing the FTP site, lists what files are posted and explains what’s in them

EDM – RMS-specific database containing exposures

RDM – RMS-specific database containing analysis results

CEDE – AIR-specific database containing exposures

CLF – AIR-specific file containing detailed analysis results. Can be loaded into CATRADER in order to apply cat treaties.

Unicede – Text file containing aggregate (by county) exposure information by line of business, includes TIVs by county, no individual location detail. Used in AIR CATRADER (can be used in RMS) to perform aggregate analysis.

Post Import Summary (PISR) – RMS report summarizing exposures in a portfolio (TIV, count, geocoding, etc.)

Catastrophe Modeling Terminology

Page 13: Catastrophe Modeling Session Reinsurance Boot Camp

How do Cat Models work?Understanding the Black Box

Page 14: Catastrophe Modeling Session Reinsurance Boot Camp

14

Hazard Module

Engineering Module

Financial Module

Defines the Event

Vulnerability of the Structure

Loss Calculation

Portfolio Definition

Insurer Location and Policy Inputs

The Four Catastrophe Model ComponentsThe Black Box

1

2

3

4

Page 15: Catastrophe Modeling Session Reinsurance Boot Camp

15

Module 1 – Portfolio DefinitionInputs

Formatted exposure data – Coverages– Terms– Risk characteristics– Reinsurance

Spatial Lookups– Geocoding– Hazard

Hurricane: Distance to Coast, Elevation Earthquake: Soil type

Portfolio Definition1

Hazard2

Engineering3

Financial4

Page 16: Catastrophe Modeling Session Reinsurance Boot Camp

16

Module 1 – Portfolio Definition Data Quality

Completeness

Correctness– Construction, occupancy, etc– Location information– Values

Valuation date– Current– Reflecting growth or reduction

Sources of uncertainty– Entry errors– Old records– Miscoding

Portfolio Definition1

Hazard2

Engineering3

Financial4

Page 17: Catastrophe Modeling Session Reinsurance Boot Camp

17

2

1

4

3

Individual risk locations

Geocoding – geographic recognition

Module 1 – Portfolio Definition Geocoding

Geocoding Resolution

County Worst

City

Zip Code

Street

Lat / Long Best

Portfolio Definition1

Hazard2

Engineering3

Financial4

Page 18: Catastrophe Modeling Session Reinsurance Boot Camp

18

Requirements– Geocoding: latitude and longitude coordinates

Based on address information – Geospatial information: environmental and/or physical factors that can

influence an event’s intensity at the site Soil conditions Topography and surface roughness Adjacent buildings

Module 2 – HazardPortfolio Definition1

Hazard2

Engineering3

Financial4 Generates the physical disturbance that is produced by an event

– Hurricane: Site Wind Speed– Earthquake: Ground Motion – Tornado/ Hail: Event Intensity

Page 19: Catastrophe Modeling Session Reinsurance Boot Camp

19

Module 2 – Hazard DefinitionHurricane Example

Portfolio Definition1

Hazard2

Engineering3

Financial4

Page 20: Catastrophe Modeling Session Reinsurance Boot Camp

20

Module 2 – Hazard DefinitionHurricane Example - Stochastic Database

Thousands of hypothetical events

Windstorm Parameters• Central Pressure• Radius to Max. Wind• Translational Speed• Wind Profile• Fill Rate• Terrain, etc.

Portfolio Definition1

Hazard2

Engineering3

Financial4

Page 21: Catastrophe Modeling Session Reinsurance Boot Camp

21

Module 2 – Hazard DefinitionHurricane Example - Event Rates

Last 100 years of historical data averages about 2.4 landfalling events per year

Traditional event probabilities distributed among thousands of storms

Portfolio Definition1

Hazard2

Engineering3

Financial4 Each stochastic event is assigned a rate – an annual frequency

Page 22: Catastrophe Modeling Session Reinsurance Boot Camp

22

Near-term hurricane frequency

Five year view (RMS)

More than three landfalling events per year

Module 2 – Hazard DefinitionHurricane Example - Event Rates

Portfolio Definition1

Hazard2

Engineering3

Financial4

Page 23: Catastrophe Modeling Session Reinsurance Boot Camp

23

2

1

4

3

Compute wind speed at each risk location

Vw = f(Pc, d, regional topography)

Distance (d)

Hurricane

Path

Module 2 – Hazard DefinitionHurricane Example - Calculate Site Windspeed

Portfolio Definition1

Hazard2

Engineering3

Financial4

Page 24: Catastrophe Modeling Session Reinsurance Boot Camp

24

• Frequency of earthquakes• Fault location• Fault geometry:

• length• depth• strike angle• dip angle

• Magnitude-recurrence• Soil type

Rupture length

Fault

Epicenter

Module 2 – Hazard DefinitionEarthquake Example - Site Ground Motion

Portfolio Definition1

Hazard2

Engineering3

Financial4

Page 25: Catastrophe Modeling Session Reinsurance Boot Camp

25

Major sources of uncertainty:– Limited historical data on events– Unknown atmospheric elements may not be recognized e.g.

Hurricane cycles

Module 2 – Hazard DefinitionLimitations

Portfolio Definition1

Hazard2

Engineering3

Financial4

Page 26: Catastrophe Modeling Session Reinsurance Boot Camp

26

Data Required:– Value

What is the value of the insured property?– Occupancy

How is the property used?- Residential

Single or Multi-Family- Commercial

Mercantile or Industrial– Construction

How is the property constructed?- Frame, Masonry, Metal, etc.- Lowrise or Highrise

– Age When was the property built? What building codes apply?

Module 3 – Vulnerability DefinitionPortfolio Definition1

Hazard2

Engineering3

Financial4

Page 27: Catastrophe Modeling Session Reinsurance Boot Camp

27

Damaged Properties

1

4

Hurricane

Module 3 – Vulnerability

Building Damageability

0%

20%

40%

60%

80%

100%

70 90 110 130 150Wind Speed

Dam

ag

e

Const 1 Const 2 Const 3

Frame Construction

50%

Damage Rates are for illustration only and are not selected from any particular model

Portfolio Definition1

Hazard2

Engineering3

Financial4

Page 28: Catastrophe Modeling Session Reinsurance Boot Camp

28

Major sources of uncertainty:– Limited claims data– Improper coding of risk characteristics– Lack of understanding of structural behavior under severe loads

Module 3 – VulnerabilityLimitations

Portfolio Definition1

Hazard2

Engineering3

Financial4

Page 29: Catastrophe Modeling Session Reinsurance Boot Camp

29

Evaluates multiple financial perspectives–Ground up: damage prior to coverage limits and

deductibles–Gross: loss after deductibles, limits, attachment points–Net: loss after treaty cessions, facultative, etc.

Module 4 –Financial Perspectives

Decreasing loss levels

Portfolio Definition1

Hazard2

Engineering3

Financial4Calculates insured losses given the

damage level and user risk inputs

Page 30: Catastrophe Modeling Session Reinsurance Boot Camp

30

Major sources of uncertainty:– Limits versus Value at Risk– Insurance and reinsurance structures are applied to loss distribution

differently: Site-level loss Policy-level loss

Module 4 – Financial PerspectivesLimitations

Portfolio Definition1

Hazard2

Engineering3

Financial4

Page 31: Catastrophe Modeling Session Reinsurance Boot Camp

Catastrophe Modeling Process

Page 32: Catastrophe Modeling Session Reinsurance Boot Camp

32

The Catastrophe Modeling ProcessOverview

Determine project scope

Gather relevant data

Evaluate, verify and format data– Data quality checklist– Data assumptions document

Import

Run the model

Review the output

Extract detailed losses

Present results

Post analysis portfolio management

Page 33: Catastrophe Modeling Session Reinsurance Boot Camp

Understanding Model Output

Page 34: Catastrophe Modeling Session Reinsurance Boot Camp

34

Model OutputTerminology

Average Annual Loss (aka Pure Premium, aka Expected Loss): Long term average loss expected in any one year

OEP - Occurrence Exceeding Probability: Probability that a single occurrence will exceed a certain threshold

AEP - Aggregate Exceeding Probability: Probability that one or more occurrences will

combine in a year to exceed the threshold.

Return Period: Level of loss and the expected amount of time between recurrences.

Critical Prob. Return Period AEP Loss OEP Loss

0.10% 1,000 160 147

0.20% 500 144 134

0.40% 250 126 118

0.50% 200 120 112

1.00% 100 97 90

Pure Premium 8

Standard Deviation 18

Page 35: Catastrophe Modeling Session Reinsurance Boot Camp

35

Model OutputThe Event Loss Table

Sample Event Output:

Page 36: Catastrophe Modeling Session Reinsurance Boot Camp

36

Model OutputThe Event Loss Table – determining PMLs - OEP

Different levels of severity based on company appetite

Common to monitor portfolios “1-100 year” loss level

In the example, “100 year loss level” is saying that there is a 1% chance that there will be a single occurrence of $2.5 billion or greater in any given year

Event IDEvent Prob

Loss Accum Prob

437812 0.05% 4,400,000,000 0.05%

437830 0.10% 4,000,000,000 0.10%

438632 0.35% 3,750,000,000 0.45%

437676 0.30% 3,600,000,000 0.75%

438622 0.10% 2,750,000,000 0.85%

438489 0.15% 2,500,000,000 1.00%

438451 0.70% 2,400,000,000 1.70%

438351 0.80% 2,250,000,000 2.50%

438248 0.90% 2,240,000,000 3.40%

Page 37: Catastrophe Modeling Session Reinsurance Boot Camp

37

Model OutputThe Event Loss Table – AEP

AEP reflects year’s worth of events rather than a single event – i.e. “there is an X% chance that there will be a total of $XX billion or

greater losses in total in any given year”

Page 38: Catastrophe Modeling Session Reinsurance Boot Camp

38

Model OutputThe Event Loss Table – determining average annual loss

Average annual loss is the weighted average of the event losses and their likelihood of occurring

A company should collect at least $91million in CAT premium to cover its average annual expected loss for the peril and portfolio being modeled

Event IDEvent Prob

Loss

437812 0.05% 4,400,000,000

437830 0.10% 4,000,000,000

438632 0.35% 3,750,000,000

437676 0.30% 3,600,000,000

438622 0.10% 2,750,000,000

438489 0.15% 2,500,000,000

438451 0.70% 2,400,000,000

438351 0.80% 2,250,000,000

438248 0.90% 2,240,000,000

Sum Product of Event Probability

and Loss

=

$91M

Page 39: Catastrophe Modeling Session Reinsurance Boot Camp

39

Average Annual LossProperties

AAL used to determine loss drivers:– Territory

Zip code County State Rating territory

– Source Risk location Policy Product line Producer

– Characteristics Construction class Occupancy

Page 40: Catastrophe Modeling Session Reinsurance Boot Camp

40

Understanding Model Uncertainty

Primary Uncertainty - Uncertainty in the occurrence of an event

Secondary Uncertainty - Uncertainty in the loss level–Range of possible loss levels

“Inherent” uncertainty–Uncertainty in the vulnerability (damage) driven by:

Insufficient historical data (infrequent) Poor quality data Translating data from one region to the next (San

Francisco 1906)

Page 41: Catastrophe Modeling Session Reinsurance Boot Camp

How is Catastrophe Model Output Used?

Page 42: Catastrophe Modeling Session Reinsurance Boot Camp

43

Catastrophe Model OutputPortfolio Management - Monitoring Loss/Premium Ratio in RML

Excluding 685 policies

from portfolio produces

an optimal RML/Premium ratio

Risk Managed Layer (RML): a range of loss levels from the EP Curve that the company wants to manage

Page 43: Catastrophe Modeling Session Reinsurance Boot Camp

44

Index Range # ZipCodes

Top 10 ZipCodes: ZipCode Index

County Name ZipCode IndexMiami-Dade 33149 1.0000

Miami-Dade 33109 0.9946

Miami-Dade 33131 0.8472

Martin 34958 0.8230

Miami-Dade 33231 0.8195

Miami-Dade 33121 0.7317

Miami-Dade 33159 0.7283

Miami-Dade 33119 0.7237

Miami-Dade 33245 0.6882

Martin 34996 0.6754

Catastrophe Model OutputGradient Map – Zip Code Index

Identifies how geographic areas are correlated to show growth/reduction opportunities

Reveals the most critical geographic areas contributing loss to the RML

Shows relative contribution to RML losses by Zip Code.

Page 44: Catastrophe Modeling Session Reinsurance Boot Camp

45

Earthquake

Severe WeatherHurricaneWildfire

Tornado/HailFlood

Catastrophe Model OutputReal-time event monitoring

Page 45: Catastrophe Modeling Session Reinsurance Boot Camp

46

ConclusionsCatastrophe Model Benefits and Shortcomings

Values– Valuable risk measure– Encourage better data tracking– Create marketplace advantages– Innovation

Dangers– Over-reliance– Misuse– Errors

Page 46: Catastrophe Modeling Session Reinsurance Boot Camp

47

Questions?

Page 47: Catastrophe Modeling Session Reinsurance Boot Camp

48

Disclaimer

The data and analysis provided by Guy Carpenter herein or in connection herewith are provided “as is”, without warranty of any kind whether express or implied. Neither Guy Carpenter, its affiliates nor their officers, directors, agents, modelers, or subcontractors (collectively, “Providers”) guarantee or warrant the correctness, completeness, currentness, merchantability, or fitness for a particular purpose of such data and analysis. In no event will any Provider be liable for loss of profits or any other indirect, special, incidental and/or consequential damage of any kind howsoever incurred or designated, arising from any use of the data and analysis provided herein or in connection herewith.

Page 48: Catastrophe Modeling Session Reinsurance Boot Camp

www.guycarp.com