P&C Cat modelling: reinsurer’s standpoint Benoît Hugonin Director of Prudential affairs - SCOR Les innovations face au développement des catastrophes non-vie Tours, 19 March 2013
Impossible d’afficher l’image.
2
P&C Cat modelling: reinsurer’s standpoint
1 Attitude when CAT risk materializes
2 Providing diversification
3 Beyond model limitations
4 Beyond the non modelled
Impossible d’afficher l’image.
3
2011 was a year characterized by record Cat losses, exceeding $100 billiion1) for the insurance industry
Floods in Australia
Earthquake in New Zealand
Floods in Thailand
Earthquake and tsunami in Japan
2011 is the second most impacted year in terms of the frequency and level of nat cat losses
1) Source : AM Best report of 23 April 2012
Impossible d’afficher l’image.
4
In this volatile environment, reinsurers have fully played their shock-absorbing role
Financial disruptions
Regulatory & tax changes
Pandemic risks
Accounting rules
Solvency requirements
Protectionism menace
New risks
Expanding universe of shocks & ruptures
Nat cat, climate change
Anticipating and detecting early disruptions and discontinuities
Interpreting the various positive and negative news flows
Distinguishing potential adverse developments
Absorbing shocks
Respecting risk appetite and buffer capital
Optimizing ILS and retrocession protection
Enhancing financial flexibility Maximizing diversification
Identifying uncertainties 1
4
Assessing the potential impact of shocks on assets and/or liabilities
Using simulation models where necessary
Solving the “ambiguity” dilemma: change in statistical distributions or specific draws from a given distribution?
Using decision-making tools in uncertain environment models
Evaluating various alternative hedging strategies
Minimizing ex-post regrets
Preventing & hedging risks
Quantifying uncertainties 2
3
Source: SCOR
Impossible d’afficher l’image.
5
Reinsurers’ reactions when CAT risk occurs
After exceptional losses in 2011, some reinsurers have reconsidered their business models and have been retrenching
2 possible attitudes for reinsurers: Opportunistic approach Long-term commitment
Benefits of long-term commitment:
Providing capacity in tough times is an essential part of the value proposition of reinsurers over the long term
Pulling out of a market after a big Cat event doesn’t make sense. One event doesn’t usually make another one more likely!
Impossible d’afficher l’image.
6
Strong franchise means constant innovation
A joint development project with RMS, the Cat Platform: Streamlines portfolio accumulation processes and integrates in real-time with SCORs contract management
system Uses simulated Year Loss Table methodology, providing more accurate modelling of reinstatements,
aggregate terms and seasonality Is designed to accommodate SCOR’s internally developed cat models and 3rd party models in addition to RMS
views, as such or in blended solutions
SCOR has deployed v1.0 of its Cat Platform Technology
SCOR supports new open-architecture cat modelling initiative
SCOR is part of an industry initiative (OASIS) to develop an open-architecture cat modelling framework, which aims to: Support validation of the existing suite of vendor models by enabling access to alternative views of hazard /
vulnerability models Facilitate the development of new models for territories that are not currently supported by the main vendors,
with the Thailand Flood model an early candidate
Impossible d’afficher l’image.
7
P&C Cat modelling: reinsurer’s standpoint
1 Attitude when CAT risk materializes
2 Providing diversification
3 Beyond model limitations
4 Beyond the non modelled
Impossible d’afficher l’image.
8
High diversification facilitates better management of cat risk
Much has been said over the past few months against diversification:
"de-diversifying“ "di-worse-ification“ “diversi-fiction”…
We believe in the virtues of diversification, be it:
geographic diversification diversification across P&C lines of business… diversification between Life and Non-Life
Impossible d’afficher l’image.
9
Cat risks no longer always occur in the same place…
In Biblical times, plagues kept on striking the same country (Egypt) again and again: Water turned into blood Frogs Lice Flies Livestock diseases Unhealable boils Hail and thunder Locusts Darkness Death of the first-born of all
Egyptian humans and animals
Today, natural disasters do not always hit the same region
Source: Guy Carpenter, 2012
Impossible d’afficher l’image.
10
The risk / return couple: optimization of the P&C and Life business mix
Diversification used for optimizing the return on risk taking
Risk / return curve based on the Life / Non-Life business mix
Risks (volatility of the portfolio)
Rat
e of
retu
rn
Minimized Risk taking
Optimized return on Risk taking
Life: ~50% of the
business mix
Life: ~75% of the
business mix
Life: 100% of the
business mix
Life: 0%
Sharpe ratio: Return per unit of risk
Market line
Intro | Internal Model | Diversification | Closing Remarks
Impossible d’afficher l’image.
11
Diversification benefit: writing different CAT-lines (Gross figures)
Strong diversification benefits can already be realized by writing a portfolio of CAT-lines (different Peril / Region)
26,8%
7,2%
16,1%
10,7% 12,6% 12,5%
14,1%
19,5%
0,3% 3,0%
0,6% 2,2% 1,4%
3,8%
0%
10%
20%
30%
CAT-line 1 CAT-line 2 CAT-line 3 CAT-line 4 CAT-line 5 CAT-line 6 CAT-line 7
one-
year
cap
ital
StandaloneDiversified
0%
20%
40%
60%
80%
100%
Sum of standalone one-yearcapital across CAT-lines
Sum of diversified one-yearcapital across CAT-lines
one-
year
cap
ital CAT-line 7
CAT-line 6CAT-line 5CAT-line 4CAT-line 3CAT-line 2CAT-line 1
30.8%
100% = sum of GROSS standalone one-year capital across all CAT lines
Intro | Internal Model | Diversification | Closing Remarks
Impossible d’afficher l’image.
12
Diversification benefit: writing different CAT-lines (Net figures)
Apart from the direct benefit from Reinsurance / Retrocessions additional diversification benefit is achieved through re-balancing CAT-lines to similar volume and volatility. The Scale is still based on the sum of the GROSS Stand-alone Capital to show effect of the Retro Protection.
12,8%
5,6%
7,8% 6,0%
10,7%
6,4%
10,4%
6,7%
0,3% 1,0% 0,3%
3,8%
0,7%
4,4%
0%
5%
10%
15%
CAT-line 1 CAT-line 2 CAT-line 3 CAT-line 4 CAT-line 5 CAT-line 6 CAT-line 7
one-
year
cap
ital
StandaloneDiversified
0%
20%
40%
60%
80%
100%
Sum of standalone one-yearcapital across CAT-lines
Sum of diversified one-yearcapital across CAT-lines
one-
year
cap
ital CAT-line 7
CAT-line 6CAT-line 5CAT-line 4CAT-line 3CAT-line 2CAT-line 1
59.7%
17.3%
100% = sum of GROSS standalone one-year capital across all CAT lines
Intro | Internal Model | Diversification | Closing Remarks
Impossible d’afficher l’image.
13
CAT tools while mature, are not yet comprehensive.
Don’t forget the «non-modelled» !
57,8%
87,8%
42,2%
12,2%
0,0%
10,0%
20,0%
30,0%
40,0%
50,0%
60,0%
70,0%
80,0%
90,0%
100,0%
Sum of standaloneone-year capitalacross CAT lines
Sum of diversifiedone-year capital
different CAT lines
CAT non-modelledCAT modelled
Impact on “modelled” & “non-modelled” CAT lines
Intro | Internal Model | Diversification | Closing Remarks
Impossible d’afficher l’image.
14
Diversification benefit (Net). with other lines of business & reserves
As expected material diversification benefits can be realized by additionally writing non CAT-lines. Scale is still based on GROSS for comparison to the two previous slides, showing the
combined effect of Retro and diversification with other lines.
12,8%
5,6%
7,8% 6,0%
10,4%
6,4%
10,4%
4,2%
0,1% 0,8% 0,3% 1,9%
0,6% 2,1%
0%
5%
10%
15%
CAT-line 1 CAT-line 2 CAT-line 3 CAT-line 4 CAT-line 5 CAT-line 6 CAT-line 7
one-
year
cap
ital
Standalone
Diversified
0%
20%
40%
60%
80%
Sum of standalone one-yearcapital across CAT-lines
Sum of diversified one-yearcapital across CAT-lines
one-
year
cap
ital CAT-line 7
CAT-line 6CAT-line 5CAT-line 4CAT-line 3CAT-line 2CAT-line 1
59.7%
10.0%
100% = sum of GROSS standalone one-year capital across all CAT lines Note, diversification is across CAT and non-CAT lines, however, only the CAT-lines are displayed
Intro | Internal Model | Diversification | Closing Remarks
Impossible d’afficher l’image.
15
Diversification benefit: Illustration of the whole portfolio (Net)
Similarly, a global CAT portfolio aids in the diversification of the non-CAT lines. Scale is now the NET Standalone Capital of the whole portfolio, including
reserves
58,9%
41,1%
20,2%
8,8% 5,8%
1,1%
23,8%
48,0%
32,5%
6,6%
0,2% 3,6%
0,7%
13,7%
0%
10%
20%
30%
40%
50%
60%
Current UY Reserves CAT-line 1 CAT-line 2 Line of Business 1
Line ofBusiness 2
Line of Business 3
one
-yea
r cap
ital
StandaloneDiversified
100% = Standalone NET one-year capital Current UY + standalone NET one-year capital Reserves
Intro | Internal Model | Diversification | Closing Remarks
Impossible d’afficher l’image.
16
0100200300400500600700800900
-100%: -90%
-90% :-80%
-80% :-70%
-70% :-60%
-60% :-50%
-50% :-40%
-40% :-30%
-30% :-20%
-20% :-10%
-10% :0%
0% :10%
10% :20%
20% :30%
30% :40%
40% :50%
50% :60%
60% :70%
70% :80%
80% :90%
90% :100%
coun
ts o
f the
real
izat
ion
with
in th
e ra
nges
realization ranges
worst 1000 out of 100K realizations (1%) StandaloneDiversified
Negative outcome for the company positive outcome for the company
-100%= Most negative outcome for the company realized in the tail
100%= Most positive outcome for the company realized in the tail
How Diversification works: CAT-line 1 (Net figures) Worst 1% scenarios of Cat-line 1 (standalone), and Cat-Line 1 results in the worst 1%
scenarios of the whole portfolio (diversified)
There are still many scenarios where the contribution from a single CAT-line is profitable in the worst economic cases that drive the capital of the Non-Life company
Intro | Internal Model | Diversification | Closing Remarks
Impossible d’afficher l’image.
17
Diversification benefit: drivers for different Lines of Business (Net figures)
The drivers for & against diversification benefit are:
size/volume, volatility of the sub-portfolio as well as dependence with the rest of the portfolio
size of CAT-lines and LoBs is represented by the size of the spheres, respectively
Intro | Internal Model | Diversification | Closing Remarks
Impossible d’afficher l’image.
18
Cautionary preliminary conclusions
“The rising risk of extreme events has much in common with playing with
loaded dice.” - Pew Centre
Climate regimes: (El Nino, NAO, AMO) Climate change
Inaccurate dependencies
2010 & 2011 record years for natural hazards Increasingly frequent medium-severe to severe insured losses caused in
particular by “localized” natural catastrophes (tornadoes, wildfires, floods, hail storms, snow storms or frost)
Non-modelled or badly modelled losses
Impossible d’afficher l’image.
19
P&C Cat modelling: reinsurer’s standpoint
1 Attitude when CAT risk materializes
2 Providing diversification
3 Beyond model limitations
4 Beyond the non modelled
Impossible d’afficher l’image.
20
Evolution of internal models
20
Marke
t Ri
sk
Cred
it Ri
sk
Insur
ance
Ri
sk
Oper
ation
al Ri
sk
Financial Instruments Portfolio Data
Internal Group Retro.
Total Risk
Market Risk
Credit Risk
Insurance Risk
Financial Instruments
Portfolio Data
Scenarios
Risk Factors Financial Instruments
Valuation Engine
Portfolio Data
Internal Group Retro.
Management Strategy
Distributional and Dependency Assumptions
Balance Sheet
Profit and Loss Distributional and Dependency
Assumptions
Valuation Model 1
Valuation Model 2
Risk Model 1
Risk Model 2
Valuation Model 3
Collection of sub models quantifying parts of the risks
Quantification of different risk types
in a portfolio
Risk types are combined to arrive at
the company’s total risk in the tails
Modelling of underlying risk drivers with full
probability distributions
1960s - 1980s 1993 - today
Value Protection Value Sustainment Value Creation
Management Strategy
Intro | Internal Model | Diversification | Closing Remarks
Impossible d’afficher l’image.
21
Dependencies - Data insufficient to reliably measure
21
• Domain of Risk Capital are the ‘tail’ or extreme scenarios (the 1/100 years for example)
• Historical data cannot capture the dependency at these extreme events, and sole reliance on it can be misleading
• Reliance on expert assessments, via an unbiased and thorough mechanism that digs into the sources of dependencies, is good complement.
X+Y
X Y
Assessment focus on the extreme scenarios and their causes:
“Suppose Y exceeds its 1-in-100 year threshold. What is the probability that also X exceeds its 1-in-100 year threshold?”
[ ])()( 99.099.0 YVaRYXVaRXP >>
Intro | Internal Model | Diversification | Closing Remarks
Impossible d’afficher l’image.
22
Calibration of P&C dependencies
The table represents a visual summary of the risk drivers documentation arisen from PrObEX
Risk drivers / Lines of Business
Natural Catastrophe
Man-Made Catastrophe Per Event Legal
Change Political Instability Economics Coinsurance
Model & parameter risks
Between LoBs
Agriculture
Auto
Aviation
Credit & Surety
Engineering
Liability
Marine
Property (Non CAT)
Space
Workers Comp.
Influence of risk drivers on LoB: high medium low
Inside LoB reading
Intro | Internal Model | Diversification | Closing Remarks
Impossible d’afficher l’image.
23
A couple of specific challenges to CAT (internal) models
Key requirements for internal models: 1. Reliable expert judgment / choice of assumptions 2. Data quality (appropriate, complete and accurate)
Additional challenges with CAT modelling
1. Use of external/vendor models 2. Rare events
Insurers and reinsurers are usually in different positions to face those challenges
1. Expertise on peak risks @ reinsurers > insurers 2. Data more directly available to insurers > reinsurers
Answers from a reinsurer
1. Robust documentation 2. Enhanced access to data through Cat Platform
Observations and Trends | Cat Platform Initiative | Fac Platform Initiative | Wrap-up and Q&A
Impossible d’afficher l’image.
24
What is a platform?
Most importantly, a platform is not A piece of bolt-on software that is unnecessarily complex and weighty A silo software that you simply use for “running a model” and “generating a report”
A platform can simply be defined as a place to launch application software and it may include some baseline hardware and software.
A successful platform in the context of re/insurance should be an agile and scalable business eco-system and have, amongst other features:
A comprehensive toolkit that allows robust execution and reduction of “model-ware” clutter Be designed to go beyond the conventional limits of modelling by leveraging advances in
analytic and technical computing capability Be able to drive out manual and repetitive processes that don’t add value Create integration and automation points to other business workflows to aid decisions Improve risk quantification through risk intelligence (e.g. explore uncertainties /sensitivities) Allow generation of internal view of risk with interoperability, blending and calibration of
multiple model sources
SCOR’s Platform initiatives include the Cat Platform and the Fac Platform.
Observations and Trends | Cat Platform Initiative | Fac Platform Initiative | Wrap-up and Q&A
Impossible d’afficher l’image.
25
The accidental cat modelling ecosystem
Cat models have become embedded within organisations and facilitate key functions : Cat pricing Portfolio management including post event response Inwards risk management and retro placement Calculation of capital adequacy Regulatory compliance and rating agency reporting
Off-the-shelf cat models have provided incredible value by enabling:
Improved Risk quantification of natural and man-made disasters Increased discipline across the entire risk transfer chain Ease of risk transfer through the “currency” of model outputs Robust risk management analytics Data creation to support the overall ERM process
Over the last 20 years or so...
Observations and Trends | Cat Platform Initiative | Fac Platform Initiative | Wrap-up and Q&A
Impossible d’afficher l’image.
26
The accidental cat modelling ecosystem
Business leaders, regulators and practitioners are continuously raising the bar… Conventional model packages are creating real operational challenges for businesses Analytical appetite of users are no longer being fulfilled, challenges around:
Model performance/execution capability (behind what is needed and what is possible) Core financials are modelled well but more complex T&C’s on the perimeter are often not Business requirements around data governance (reporting/downstream) are increasing
across the entire risk transfer chain Models are generally fine (…although always evolving) for peak perils but not for
secondary, non-peak and un-modelled perils
Model inflexibility has forced companies develop suites of bespoke and complex ‘model interfacing tools’. These accidental modelling ecosystems have evolved over time but there are issues: Significant overhead to keep bespoke “model-ware” aligned High risk strategy and often this introduces key person risk Reached a saturation point where the value proposition is diminishing
The current challenges...
Observations and Trends | Cat Platform Initiative | Fac Platform Initiative | Wrap-up and Q&A
Impossible d’afficher l’image.
27
Cat Platform – An ambitious development project
Platform capability to handle a reinsurers complex cat book including advanced analytics tools including: Cedant exposure analytics Cedant and portfolio modelling Dynamic portfolio risk aggregation an follow-up Portfolio management and optimization
Real-time integration with other production systems:
Contract management system Pricing Capital Model
Increased operational efficiency by driving out manual processes
Delivering interactive access and value to a broader internal audience beyond cat modelling team
Phased development with releases timed to deliver business value in time
State-of-the-art comprehensive analytical framework...
Observations and Trends | Cat Platform Initiative | Fac Platform Initiative | Wrap-up and Q&A
Impossible d’afficher l’image.
28
Cat Platform - Functional features overview
28
Observations and Trends | Cat Platform Initiative | Fac Platform Initiative | Wrap-up and Q&A
Impossible d’afficher l’image.
29
P&C Cat modelling: reinsurer’s standpoint
1 Attitude when CAT risk materializes
2 Providing diversification
3 Beyond model limitations
4 Beyond the non modelled
Impossible d’afficher l’image.
30
Non-peak, non-modelled is no excuse for non-management
2011 Cat events were unusual: Regional pattern:
Asia-Pacific skew Non-Peak: over 50% of insured costs
Perils: Earthquakes/floods rather than wind Non-modelled: 28% of insured costs
Thai floods Australian floods
The lack of model can apply :
to the event itself: Copenhagen floods to secondary perils: Japanese tsunami following earthquake to the consequences for insureds: CBI/Supply chain disruptions
following Thai floods
Impossible d’afficher l’image.
31
2011 : ‘Annus Horribilis’ for Catastrophe Losses…
Impact on the industry / society: over $100bn insured losses; Economic losses: ~ $350bn Mortality: ~27,000
Unusual (?) peril-regional pattern Asia Pacific skew Earthquake / Flood dominated
Non-Peak over 50% Everything apart from Tohoku,
Irene Non-modelled $30bn (28%)
Thai flood Australia flood
‘Surprises’ Japan: Mag, Tsunami, Fukushima US: Tornadoes NZ: Frequency, aftershocks,
liquefaction Thai/Japan: CBI / Supply Chain
A record year:
Intro | Blind Spots| Closing Gaps | Paradigm Shift | Closing Remarks
Impossible d’afficher l’image.
32
…against a backdrop of higher expectations
Regulatory Solvency II and equivalent Internal model, ORSA, Pillar III…
Ratings Agencies Cat Questionnaires/Surveys ERM
Investors / Analysts Transparency, confidence
Clients Capacity Price stability Continuity
Management Effective controls Better information for strategic
decision making Protect Balance Sheet and Earnings
Stakeholder needs:
Intro | Blind Spots| Closing Gaps | Paradigm Shift | Closing Remarks
Impossible d’afficher l’image.
33
An incomplete vendor modelled view of global cat risk
Flood models (primary hazard) Thailand, Australia, India, China,
US, Brazil, Central Eastern Europe, Saudi Arabia…
2ndry perils Tsunami…everywhere Storm surge other than US, UK Fire-following earthquakes
Hail, Bushfire Correlations between perils
Large scale weather patterns El Nino/La Nina NAO, AMO
Lines of business Non-property CBI / Supply Chain
Not available…yet
Intro | Blind Spots| Closing Gaps | Paradigm Shift | Closing Remarks
Impossible d’afficher l’image.
34
Some ways to bridge the gaps…
Tracking exposure Availability Reliability Confidentiality Complex contracts
Extending the model framework Bolt on risk profiles Tsunami, Fire-following earthquake Correlate non-, badly-modelled LoBs
In-house models Not easy, but sometimes necessary Simple models, leveraging 3rd party
research Pricing and capital analysis
Coherence, consistency One internal model
Scenarios – thinking the unthinkable
…to a comprehensive view
0%10%20%30%40%50%60%70%80%90%
100%
2 5 10 15 20 50 75 100 200 250 500 1000
Model Pareto
TVaR contribution by Return Period
Intro | Blind Spots| Closing Gaps | Paradigm Shift | Closing Remarks
Impossible d’afficher l’image.
35
An Oasis in the desert?
An ‘open’ framework for cat modelling
Prototype models under
construction: UK Flood European Wind North Africa
Earthquake Cascadia Tsunami
New cat model initiative ‘Plug and Play’
Intro | Blind Spots| Closing Gaps | Paradigm Shift | Closing Remarks
Impossible d’afficher l’image.
36
Some closing comments (1/2)
Vendor cat models have served the industry well for 20 years, and will continue to for the next 20 …but, material gaps remain.
Insurers and reinsurers need more and better models, quicker than the ‘big 3’ can deliver
The cat modelling eco-system needs to and will get bigger
Intro | Blind Spots| Closing Gaps | Paradigm Shift | Closing Remarks
Impossible d’afficher l’image.
37
Some closing comments (2/2)
But, risk that regulators somehow slow down that process of growth Delays with Solvency II
Growing skepticism around the use of internal models for regulatory purposes, for the time being in the banking world: “In surveying the failings of financial authorities, both here and abroad, one can certainly identify
some specific characteristics of pre-crisis regulation that look today to have been significantly misguided, rather than the advances they were formerly thought to be. So, for example, regulators became prone to place too much confidence in the capacity of firms to measure and manage their risks. […] a supervisory approach, which relies on a more opaque, firm-specific process of watching over banks' own risk-management and compliance systems. […] standardized measures serve as a floor to guard against the potential for models-based capital measures to understate capital needs under some circumstances. They are also substantially less opaque than, for example, the advanced internal ratings-based approach of Basel II, and thus would provide more comparable measures that are also more amenable to international monitoring.”
Daniel Tarullo, member of the Board of Governors of the Fed, 22 Feb 2013 Growing skepticism around diversification benefits
Joint Forum reports Insurance risk vs financial risk: not the same (in)dependence in stressed situations.
Intro | Blind Spots| Closing Gaps | Paradigm Shift | Closing Remarks