CAT Modelling - Actuaries Institute · Cat modelling is a bottomless pit of detail – The...
Transcript of CAT Modelling - Actuaries Institute · Cat modelling is a bottomless pit of detail – The...
CAT Modelling
Jeremy Waite Nicholas Miller
© Institute of Actuaries of Australia
This presentation has been prepared for the Actuaries Institute 2014 General Insurance Seminar. The Institute Council wishes it to be understood that opinions put forward herein are not necessarily those of the Institute and the Council is not
responsible for those opinions.
Team • Susan Ley, Munich Re, Convenor, • Anu Agarwal, IAG Re • Michael Barkhausen, Willis Re • Andrew Hulme, AON • Nicholas Miller, QBE • Jeremy Waite, JLT Re • Richard Yee, E&Y
Purpose of Note • To assist Members who are required to
provide commentary under APRA prudential standards or professional standards on estimates of liabilities or prescribed capital that rely to a material degree on the output of catastrophe models.
Purpose of Presentation • To broadly cover note topics
– Read the note for the detail.. – Read the references in the note for more detail… – Do a Phd in your specific peril engineering for even more detail….
• Then to go beyond the note…
– Aim to keep it practical and pragmatic – An RDS example – And some real approach examples from QBE and IAG
• Cat modelling is a bottomless pit of detail
– The construction and interrogation of cat models is not an actuarial pursuit – we are users of the outputs
– Applied common and business sense are needed to make catastrophe management decisions
– Actuaries can help in this process – AA’s are 3rd line of defence – Board is ultimately responsible
Actuaries and Cat models • Information Note is intended to cover what an Actuary may
consider in fulfilling the regulatory requirements as well as those of the professional standards:- – PS300 (IoAA) – GI Valuation – PS305 (IoAA) – FCR for GI – GPS320 (APRA) – Actuarial & Related Matters – GPS116 (APRA) - Capital Adequacy – ICRC
Ultimate responsibility for the prudential management of capital of a general insurer rests with its Board of directors.
What is a Cat model “Catastrophe models are developed by groups of scientists, engineers, mathematicians and actuaries working together to simulate catastrophic events. While most actuaries conceptually agree that catastrophe models may provide more realistic measures of catastrophic risk than those provided by analysing the latest twenty to fifty years of catastrophe losses, most actuaries are not experts in many of the underpinnings of these models.” Actuarial Standards Board (2011): ASOP 38.
Is the modelling appropriate • Not role of the Actuary to undertake “all or
any” of the investigations • How to assess • Considerations when using results • Data Quality & Governance
Vendor Model Availability in Australia
• How has Storm risk out of Sydney been considered? • How’s flood been modelled? • Are cyclones more frequent than in the past (RMS model 2006 vintage) • Any consideration given to ENSO and bushfire risk? • Be prepared to step away from the model
Vendor Cyclone Earthquake Bushfire Flood Storm Hail AIR Yes Yes Yes No No No
EQECAT Yes Yes No No No No Risk Frontiers Yes Yes Yes Yes No Yes
RMS Yes Yes No No Combined - Sydney only
Relevant Australian perils modelled?
Uncertainty - Causes?
Uncertainty from model input
Uncertainty from model Scope
Uncertainty from model Scope Uncertainty from model Scope
Uncertainty from model output
Uncertainty - Causes? “The uncertainty is not only due to the model, but real uncertainty in the phenomena which despite the best efforts of the brightest scientists/engineers and big budgets of the vendor firms can’t be quantified within the model”
What can go wrong with Models
• Bad Physics • Bad Assumptions • Bad Data • Bad Luck • Too many Black swans
Source: Stein, Geller, Liu 2012
Plate tectonics was only accepted in the early 1960’s by the geoscientific community
More Uncertainty - Causes?
Earthquake Epicentres in Australia 1841 to 2000
Source: McCue, K.F, 2001 – Earthquake Epicentres in Australia 1841 – 2000 and recent fault scarps, AGSO – Geoscience Australia, Canberra
• The Australian continent is completely within the Australian plate and consequently there are no major through going active faults like the Alpine fault in NZ or the San Andreas fault in Western USA.
Uncertainty Australian EQ – real uncertainty
Realistic Disaster Scenarios Cyclone
Earthquake
Bushfire
Flood
Storm
Hail
• 1 In 200 PML not an event – What events of this scale are plausible
Hence RDS (used at Lloyds since June 1994)
• Easy to Understand
– Board and investor-friendly
• Qualitative and Quantitative Testing
– Reinsurance arrangements
– Capital impact
• Satisfies APRA stress and scenario test
– Vertical Requirement
– Horizontal Requirement
Realistic Disaster Scenarios A specific RDS event typically has the following details:- • A definition of the physical event, with a map showing
the footprint or storm-track; • The assumed industry insured loss (could be split by line
of business e.g. Property– Domestic, Commercial Industrial – or include other classes of business if material (e.g. Marine))
• Other details could be provided (e.g. where applicable, a catalogue of major infrastructure (i.e. ports) that may be affected by the event);
RDS - Example Cyclone
Earthquake
Bushfire
Flood
Storm
Hail
Peril Location RDS ID
Postcode at Centre Longditude Latitude Magnitude
153.14 -26.99 Category 3Event Description
Event Similar to
Loss Details
Industry Loss XYZ Gross Loss Most costly CRESTA ID
$13.743 Bn $351,977,876
XYZ Market share of Loss XYZ Reinsurance Recoveries Most costly CRESTA Name
2.561% -$ XYZ Market share of Exposure XYZ Net Loss Most costly CRESTA Gross Loss
1.922% $351,977,876
Event Details
Cyclone Brisbane 1
2
Brisbane
$173,653,234
Central Pressure 947hPa, South Westerly direction
Cyclone Daisy (1972, 965hPa)
4507
Low % Damage <<< Relative RIsk >>> High % Damage
• RDS example for a 1 in 200 cyclone (Australia return period) loss
• Market loss $13.7bn
• Event details and impact on XYZ insurer
• Here evidence of anti selection?
Dealing with Uncertainty in Models • Recognise importance • Identify sources • Assess consequences • Adjust or mitigate • Alternative insights • More than one model • Scenarios • Stress Testing • Workshops • Reverse Stress Testing Management Action
Dealing with Uncertainty Management Actions
• Avoid • Change risk profile –
portfolio • Introduce new rating
factors – avoid anti selection
• Change mix of business • Mitigate
• Use a larger PML than modelled
• Use conservative inputs • Have more capital than
minimums
• Reduce • Change cover – introduce sub
limits in direct policies • Have event excess in direct
policies • Increase deductibles
• Transfer • Buy reinsurance - VR • Buy aggregate cover - HR • Buy a longer return period
than APRA minimum • Add a $ load on the PML limit • Buy a parametric cover • Buy a cat bond
Does the note’s perspective help • Christchurch – Model driven PML Only
Unknown fault line (Hazard map) Model gave low probability (OEP curve) Liquefaction not included in model Etc…
• Christchurch – using alternate RDS Method IF Christchurch was a significant exposure AND a Christchurch RDS was selected,
then could have quantified RDS – impact of event of market return period size located at Christchurch on XYZ RDS- possible multiple large events in same location (not model driven) impact on
XYZ RDS vary damage (sense testing) impact on XYZ
Potentially leading to Action: Avoid, Reduce, Mitigate or Transfer
Would the note’s perspective help • Models do not replace common sense
– A fool with a tool is still a fool • Note aims to help wider Actuaries & Clients perspective
– There is real uncertainty and model measurable uncertainty • Alternate approaches can give wider perspective and lead to action
– RDS are accepted alternatives to model driven results • Stress and scenario can give perspective and open up understanding
– Can lead to alternative action • Actuaries are part of a process involving other expertise beyond their
own – Note aims to assist actuaries in carrying out their duties
QBE Example
QBE • Background to QBE ANZO
– Multi-line insurer – Wide geographic coverage – Global structure – License catastrophe models
Use of catastrophe models Data -consistent, accurate and complete
Selection of Model -fit for purpose -research, expert opinion
Results Analysis -gaps, limitations, sensitivity -experience, underwriting, other approaches
0
1
2
3
1900 1970 1990 2005
Year Built
0.0
0.5
1.0
1.5
2 5 10 15
No of Stories
0
0.5
1
1.5
Wood Masonry Concrete Steel
Construction
Alternative methods • Use of alternative
methods: – RDS – Experience,
internal models – Exposure analysis – Use of DFA tools
Questions