0Proprietary & Confidential
From Geography to
(Re)Insurance:
Catastrophe Modelling
Presentation for Masaryk University
28th April 2020
Karel Vojak
karel.vojak@aon.com
Proprietary & Confidential 1
Content
Intro to insurance and reinsurance Section 1
Intro to catastrophe modelling Section 2
Few notes on modelling European windstorms Section 3
Final comments & takeaways Section 4
About Aon Section 5
Disclaimer Section 6
2
Insurance
&
Reinsurance
1
3Proprietary & Confidential
Hurricane Andrew 1992
Source: https://www.youtube.com/watch?v=2n8kJBcGaTM
Open the link below and watch the short video
4Proprietary & Confidential
Insurance and Reinsurance
Was this gentleman
insured?
Was the insurance company
reinsured – prepared to pay
his and all other claims?
Source: https://www.youtube.com/watch?v=2n8kJBcGaTM
Insurance
Reinsurance
5Proprietary & Confidential
Insurance and Reinsurance
Reinsurance = insurer’s insurance
Zajištění = pojištění pojišťoven
Owner
Insurance broker
(sometimes)
Insurance
company
Reinsurance
broker
(almost always)
Reinsurance
companies
Typical risk transfer: from owner (you) to reinsurance companies
The primary buyer of
insurance, insured
 You
 Me
 Tereza
 Viktor
 Marketa
 …
Intermediary between
insured owner and
insurance company,
helping owners with
their insurance
 Aon Risk Solutions
 Marsh
 Renomia
 …
Insurer
 Generali
 Kooperativa
 Allianz
 UNIQA
 …
Intermediary between
insurance and
reinsurance company,
helping insurers with
their reinsurance
 Aon Reinsurance
Solutions
 Guy Carpenter
 Willis Re
 …
Reinsurers
 Swiss Re
 Munich Re
 Scor
 TransRe
 …
6Proprietary & Confidential
Question: Hmm…but I thought insurers have enough money from the premium (pojistné) they
get from their clients (owners)?
Answer: Typically an insurance company pays single claims (car accident, family house on
fire,…) easily, but sometimes the loss can be so big that the collected premium may not be
enough for covering everything.
Q: So what kind of events do insurance companies worry about?
A: Large rare losses. Catastrophes.
Q: What catastrophes?
A: Mainly natural and man-made catastrophes causing large damage and financial losses
covered by insurance policies. Natural catastrophes are typically driving the need for
reinsurance. After a catastrophe, insurance companies need to pay a lot of money within very
short time. If they don’t have money to cover the claims, they bankrupt.
Q: Has it really happened that an insurance company bankrupt because of a natural
catastrophe?
A: Yes. Unfortunately. Sad example of the insurance industry unpreparedness for large losses
was hurricane Andrew in 1992. Several insurance companies went out of business leaving their
clients without money. Other insurers went into sever solvency issues as well.
Q: It was a fraud or why insurance companies didn’t have enough money?
A: No fraud. They thought this cannot happen. Several decades before Andrew, there was no
major hurricane making a landfall and causing such dramatic damage. Almost everybody
believed that the worst what can happen is a loss of about 5 billion USD I think. That’s what the
past insurance data told them. Andrew insured loss was more than 15 billion USD…
Q: So was there any lesson insurance industry learned from that catastrophe?
A: Yes. Andrew was a wake up call for more robust estimations of what can potentially happen
and what insurance industry needs to be prepared for. Andrew changed the (re)insurance
industry and it was the beginning of real catastrophe modelling.
Why Insurance Company Buys Reinsurance?
To be ready to pay large losses to all clients - owners
Q: OK, so insurers buy reinsurance from reinsurers to
prevent insolvency after an extreme rare loss. But I
still feel that big traditional insurers must have
collected a lot of money, so they should need no
reinsurance…
A: On long term, you are right. But it is more complex
within timescales of few years. There are other lines
of business, cost of capital one need to hold,
regulation, and many other things. However, the
main point is still that reinsurance allows insurers to
be ready for a big catastrophe loss.
Q: Reinsurers provide insurance for insurers. So
reinsurers must be way bigger companies than
insurers?
A: That’s not necessary. Typically an insurance
company is reinsured by several reinsurers, each
participating with some percentage on the deal. The
main point of insurance and reinsurance business is
diversification. Reinsurers write typically business
over the globe, so it is impossible for most natural
catastrophe to hit whole earth – whole business of a
reinsurer. On the other, many insurance companies
are more local and majority of their business can be
hit by a hurricane, flood,… Basically the risk transfer
throughout the industry from an homeowner to
reinsurers includes diversification in each step.
Q: Uff… many thanks
7Proprietary & Confidential
What Are the Main Perils to the Insurance Industry?
Worldwide:
 Hurricanes
 Floods
 Earthquakes (and
tsunami)
 EU Windstorms
 Hailstorms
 Tornadoes
 Wildfires
 Terrorism (man-
made)
Europewide:
 The continentwide
costliest peril
is windstorm, but
in some regions it
is flood (Central
Europe) or
earthquake (Italy,
Greece,…)
8Proprietary & Confidential
Brief History of Catastrophe Economic Losses
Economic annual losses
9Proprietary & Confidential
Brief History of Insured Catastrophe Losses
Insured annual losses with notes on selected major disasters driving the total loss that year
10Proprietary & Confidential
What Helps (Re)Insurers to Understand and Manage
Catastrophes They May Be Facing In (Near) Future?
Catastrophe Modelling
11
Intro to
Catastrophe
Modelling
2
12Proprietary & Confidential
Catastrophe Modelling – What Is It and Why Is It Used?
Why (re)insurance industry use catastrophe models?
 Statistical methods based on historical data do not provide reliable results when estimating impacts of rare extreme catastrophes. The industry needs
better estimations. Examples where pure statistics was not enough:
– Hurricane Andrew 1992 – (re)insurers believed this was impossible because nothing like that could have been seen in historical insurance loss data
over many years prior 1992.
– Haiti – 6 people died due to earthquakes there between years 1900 and 2009. In January 2010 more than 250000 people died in the earthquake.
Did a statistician have a chance to predict such catastrophe?
 Understanding the physics of natural catastrophes, modelling natural catastrophes from geo-science perspective, and linking this with
impact of the catastrophes (engineering and financial perspective) allows for more reliable estimations
– First catastrophe (cat) models were developed in late 80’s, but there was not much interest in the (re)insurance industry. Cat model results were
basically ignored. A cat model estimated that a loss from a hurricane in US can exceed tens of billion USD. (Re)insurers didn’t care that time. After
hurricane Andrew, it was clear that pure statistics is not enough and that climatology+statistics+claims can provide much better estimations.
 Because it provides better answers than other methods for estimating impacts of natural catastrophes
Catastrophe modelling is the process of using computer-assisted calculations to estimate losses from a catastrophic events.
Catastrophe model is a tool - a software providing information of what financial loss (e.g. EUR) from certain peril (storm, earthquake,
flood,…) an insurance company can expect with what probability.
Catastrophe modelling has two main aims in the (re)insurance industry:
 To help understand, evaluate, and quantify the natural catastrophe exposure and risk faced by a (re)insurance company
“Do I have enough cover?“ “Jsem dostatečně zajištěn?“ “What is my 1-in-200 years loss?” “What is return period of my largest loss in past?”
 To assist in determining appropriate risk mitigation strategies
“Is my reinsurance structure efficient?” “Do I get the right price?”
13Proprietary & Confidential
What Perils Are Modelled?
Today catastrophe models are available for most perils and regions
Example: Models developed by Impact Forecasting
14Proprietary & Confidential
Main Catastrophe Models Providers
Four main commercial catastrophe model vendors – firms developing and selling catastrophe models
 RMS (Risk Management
Solutions)
 World's largest catastrophe
modelling company
 Founded in 1988
 Headquartered in San
Francisco with several offices
around the world
 Clients include 85% of the top
40 global reinsurance
companies
 AIR (Applied Insurance
Research)
 Part of the Verisk Analytics
family of companies, a leading
data analytics provider
 Headquartered in Boston with
several offices around the
world
 Founded in 1987
 Previously EQECAT (acquired
by CoreLogic in 2013)
 CoreLogic is large data
provider as well as analytics
services
 EQECAT founded in 1994
 Headquartered in Los Angeles
with offices around the world
 Impact Forecasting – Aon
catastrophe modelling team
 Founded in 1996
 Models for peak risk zones
around the globe, as well as
for the emerging markets.
 Headquartered in London with
offices around the world
including Prague
 Head of Impact Forecasting
holds a PhD in Physical
Geography from Charles
University in Prague
There are also other cat models developed by reinsurers, brokers, smaller modelling firms
15Proprietary & Confidential
How Cat Models Work
LOSS
Insured loss
calculation
Damage
Calculation
VULNERABILITY
HAZARD
Event Generation
Intensity Calculation
EXPOSURE
Risk Characterisation
Policy Conditions
 EXPOSURE – Insured portfolio; portfolio is a set of buildings
(example of portfolio: 1 million residential buildings of various
value and with addresses and 10 thousands industrial
buildings of various value with addresses in CZ)
– This an input to catastrophe model
 HAZARD – Simulated thousands of hypothetical – but
physically realistic – windstorms (or earthquakes, floods…
below windstorm used as an example)
– Each storm’s intensity expressed as wind speed field
(footprint, map)
– Each storm of certain frequency
– Several methods of generating hypothetical storms exist,
including use of reanalyses, global circulation models,
mesoscale weather forecast models, records from weather
stations, GIS,…
 VULNERABILITY – each storm’s intensity translated into
damage on the exposure (on each building within a portfolio)
– The vulnerability part uses functional relationship between
wind speed and a damage on a structure
– Probabilistic approach – uncertainty in the relationship
between wind speed and damage considered
 LOSS – Policy conditions covering the insured property
applied on the calculated damage and finally:
– Set of potential losses from each hypothetical storm on a
portfolio calculated to “Event Loss Table (ELT)”
– Final outcomes: Occurrence Exceedance Probability
(OEP), Annual Average Loss (AAL)
Cat model consists typically of 4 main parts
16Proprietary & Confidential
Probabilistic Approach
 Critical aspect of catastrophe modelling - uncertainties taken into account
 Certain local intensity of a natural catastrophe (wind speed, peak ground acceleration, flood depth, hail size,…) can cause various damage –
sometimes nothing happens to a building (0% damage), sometimes a window is broken, sometimes a roof is blown away, sometimes whole building
collapses (100% damage)
 Example below – an earthquake of certain intensity causes on average 30% damage on a building (Mean Damage Ratio = 30%) – the left graph.
However, this is an average for the given intensity and it could be that the loss can range from 0% to 100% damage and there is typically a distribution
around the mean loss allowing for loss calculations with uncertainties considered – the right graph.
Earthquake Intensity
MeanDamageRatio
Vulnerability function
30%
Loss
ProbabilityofLoss(%)
Damage ratio for a given intensity
Damage ratio distribution allowing for all the different
values of damage ratio surrounding the mean damage
Loss distribution found by multiplying damage ratio
distribution by replacement value
17Proprietary & Confidential
Cat Model Inputs
Input – Information about the portfolio – WHAT is in the portfolio and WHERE it is located
List of insured property (buildings, contents, cars,…)
 Example of a (fictitous) portfolio as input data (exposure) from an insurance company
Policy
(pojistka)
Location ID Country Postal Code Address Occupancy Construction Building
Value
Deductible
(spoluúčast)
Limit
000256 1 CZ 13000 Horní 17 Commercial
Reinforced
Concrete
3000000 1900 -
000256 2 CZ 77146 Dolní 6 Residential Masonry 200000 100 -
000257 3 SK 83104 Střední 127 Industrial Unknown 2500000 2% 1000000
…
… … … … … … … …
…
… … … … … … … … … …
… … … … … … … … … …
… … … … … … … … … …
… … … … … … … … … …
… … … … … … … … … …
0A7763 623991 CZ 26203 Podzemní 64 Residential Unknown 300000 150 -
18Proprietary & Confidential
Occurrence exceedance probability curve (OEP)
 Calculated from ELT
 Certain loss is exceeded with certain probability (usually expressed as return
period). This tell insurers what they need to be prepared for.
 Example below: an insurance company can expect loss of 93 million EUR or
more once in 200 years, according to the “yellow” cat model
-
20
40
60
80
100
120
0 100 200 300 400 500
LossEstimate[mEUR]
Return Period [Years]
Cat Model Outputs
93m
EUR
Event Loss Table (ELT)
 Core outcome from cat model and the basis for further analyses
and results
 List of all modelled hypothetical events (e.g. thousands of
storms) with corresponding losses on a portfolio and including
frequency of each storm.
 Example:
Annual Average Loss (AAL)
 Calculated from ELT
 Hypothetical long term average of annual loss for a company
from the analyzed peril. Example: 10.5 million EUR
Event ID Frequency Mean Loss Std Dev
1 0.0002298
358,401 252,189
2 0.0001925
1,565,878 840,450
3 0.0002601
1,511,418 814,073
4 0.0001567
45,862,410 9,305,069
5 0.0004021
300,263 218,313
… … … …
… … … …
… … … …
… … … …
18143 0.0001142
167,166 119,497
19
Few Notes on
EU Windstorm
Catastrophe
Modelling
3
20Proprietary & Confidential
European Windstorm
 The costliest natural peril in Europe
– Some major historical windstorms: 87J, 1990 storms (Daria, Herta, Vivian, Wiebke), Anatol and Lothar and Martin 1999, Erwin 2005, Kyrill 2007,
Klaus 2009, Xynthia 2010, Christian 2013
 European Windstorm = Extratropical cyclone = Winter windstorm
– Deep low pressure systems typically originating in west Atlantic and travelling towards Europe
– Sometimes an extratropical cyclone gets very deep and extreme pressure gradients lead to extreme wind speeds
– Rarely an extreme extratropical cyclone strikes the continent and has disastrous impact in many European countries – large regions can be affected
Orkanen over Danmark klokken 17:49 den 3. december 1999. Data:
NOAA / billedbehandling DMI.
21Proprietary & Confidential
Key Challenges – Examples in CZ
Questions
 How often a storm with “Kyrill” devasting effect hits CZ? What is Kyrill’s return period?
 Can it be worse? What is the worst case windstorm insurance companies need to be
prepared for? How much is the windstorm loss of return period 200 years?
 What is the return period of storms with impact similar to Herwart?
Windstorm is a Europe-wide perils affecting mainly western Europe.
However, do you remember major windstorms in the Czech Republic?
 Kyrill 2007
 Emma 2008
 Herwart 2017
 Sabine 2020
How to find answers?
 Limited historical data in CZ covering few decades (insurance losses or reliable wind
speed records) do not allow for robust estimations of the biggest extremes
 Cat models help – they include thousands of hypothetical storms to capture windstorm
climatology and include extremes never recorded in history
– Windstorm cat models are based on various sophisticated methods and data such
as global circulation models (GCM), reanalyses (e.g. ERA-Interim), specific
perturbation and interpolation techniques, etc.
– Even the models are based on relatively up-to-date science and data, there are
limitations (e.g. resolution of reanalyses and RCM); models are not perfect
– Cat modelling is still “young” discipline. Science drives improvements and critical
assessment of current tools. Looking at the topic from various angles is needed.
Example: Climate modelling using supercomputers and (vs.) historical climatology
Kyrill footprint – map of maximum wind gusts during the storm.
Source: AIR, https://www.air-
worldwide.com/blog/posts/2017/1/kyrill-the-winter-storm-that-
walloped-most-of-europe/
22Proprietary & Confidential
So what is the return period of Kyrill-size loss in CZ?
 Cat model A (primarily based on statistical perturbations of footprints derived from reanalysis) : X years
 Cat model B (primarily based GCM simulations, downscaled to finer resolution) : Y years
 More recent paper suggests that model A appears too optimistic/pessimistic and model B appears too optimistic/pessimistic
– Adjustments to model A/B is therefore recommended to better reflect the latest science
– According to the adjusted model A/B, Kyrill-size windstorm loss in CZ is of return period Z years
– Z is the final answer… until better model/approach/findings/data improves this estimate (your turn!)
Kyrill
Emma
Vivian +
Wiebke
Capella
Source: Brazdil R. et al. (2018): Windstorms
and forest disturbances in the Czech Lands:
1801-2015. Agricultural and Forest
Meteorology. vol. 250-251. p 47-63.
Example of recent findings with direct impact on catastrophe modelling:
23
Comments &
Takeaways
4
24Proprietary & Confidential
Final Comments and Takeaways
Actuarial
Mathematics
And Statistics
Climatology,
Seismology,
Hydrology,
and Other Natural
Sciences…
Structural
Engineering
Data
Data
Data
Data
Programming
and IT
Catastrophe modelling is multi-disciplinary
 Cat models are tools to estimate impacts of catastrophes
 Cat models are embedded in the (re)insurance industry
 Cat models are based on knowledge and findings from various
disciplines including natural sciences, actuarial mathematics,
engineering,…
– The multi-disciplinary nature together with major role of
natural sciences … anyone feels the link with geography?
 Don’t forget that “model” is synonym to something imperfect –
critical thinking is always an advantage
… and requires excellence in each discipline
 From geography to (re)insurance
Question: Karle, you studied physical geography and now you
are pojišťovák. How that happen?
KV: Via catastrophe modelling. Catastrophe modelling is the
link between (re)insurance and geography.
25
About Aon
5
26Proprietary & Confidential
About Aon
27Proprietary & Confidential
Aon’s Portfolio of Solutions
28Proprietary & Confidential
Aon Reinsurance Solutions
29
Disclaimer
6
Proprietary & Confidential 30
Disclaimer
Legal Disclaimer
©Aon UK Limited (for itself and on behalf of each subsidiary company of Aon plc) (“Aon’s Reinsurance Solutions”) reserves all rights to the content of this
report (“Report”). This Report is for distribution to Aon’s Reinsurance Solutions and the organisation to which it was originally delivered only. Copies may be
made by that organisation for its own internal purposes but this Report may not be distributed in whole or in part to any third party without both (i) the prior
written consent of Aon’s Reinsurance Solutions. and (ii) the third party having first signed a “recipient of report” letter in a form acceptable to Aon’s
Reinsurance Solutions. Aon’s Reinsurance Solutions cannot accept any liability to any third party to whom this Report is disclosed, whether disclosed in
compliance with the preceding sentence of otherwise. To the extent this Report expresses any recommendation or assessment on any aspect of risk, the
recipient acknowledges that any such recommendation or assessment is an expression of Aon’s Reinsurance Solutions’ opinion only, and is not a statement
of fact. Any decision to rely on any such recommendation or assessment of risk is entirely the responsibility of the recipient. Aon’s Reinsurance Solutions will
not in any event be responsible for any losses that may be incurred by any party as a result of any reliance placed on any such opinion. The recipient
acknowledges that this Report does not replace the need for the recipient to undertake its own assessment. The recipient acknowledges that in preparing
this Report Aon’s Reinsurance Solutions may have based analysis on data provided by the recipient and/or from third party sources. This data may have
been subjected to mathematical and/or empirical analysis and modelling. Aon’s Reinsurance Solutions has not verified, and accepts no responsibility for, the
accuracy or completeness of any such data. In addition, the recipient acknowledges that any form of mathematical and/or empirical analysis and modelling
(including that used in the preparation of this Report) may produce results which differ from actual events or losses. The Aon’s Reinsurance Solutions
analysis has been undertaken from the perspective of a reinsurance broker. Consequently this Report does not constitute an opinion of reserving levels or
accounting treatment. This Report does not constitute any form of legal, accounting, taxation, regulatory or actuarial advice.
This presentation was prepared for Masaryk University for educational purposes only
Thank
you!
Aon plc (NYSE:AON) is a leading global professional services firm
providing a broad range of risk, retirement and health solutions. Our
50.000 colleagues in 120 countries empower results for clients by using
proprietary data and analytics to deliver insights that reduce volatility and
improve performance.
© Aon plc 2020. All rights reserved.