Adding a Cat Load to Property Reinsurance Pricing

1
Adding a Cat Load to Property Reinsurance Pricing

• One Reinsurers Approach
• June 1, 2005 - CAGNY

2
Agenda

• Early Disclaimers
• Property Reinsurance Pricing Laying the
  Groundwork before adding a Cat Load
• What do you do with Cat Modeling Input and
  Output?
• How do you incorporate a Cat Load into Cash Flow
  Modeling?
• Can you judge a company by its Cat Modeling?
• Questions/Comments

3
Early Disclaimers

• Scope of discussion
• Not HOW to run cat models
• Rather, analyzing inputs and outputs
• Focus on RMS
• Types of treaties
• Per Risk
• Quota Shares
• Endurance in N.A. doesnt price pure cat treaties
• More ways to skin the cat than presented here
• Comments and suggestions welcome!

4
Property Reinsurance Pricing Getting the ball
rolling

• Analyze cat vs. non-cat separately
• Exposure rate
• PSOLD, Loss to Value Curves, etc.
• Use gross non-cat loss ratios
• Experience rate
• Both non-cat and cat only basis
• Consider including some cats in non-cat analysis
• Hurricanes w/significant flood (Floyd, Allison)
• Tornado and hail events
• Once non-cat burn is selected, add cat load
• Monte Carlo Simulation models are used to value
  any loss sensitive features.

5
Examining your EDM Avoiding Garbage in, Garbage
out

• EDM Content
• Perils
• Regions
• Examine Post Import Summary
• of locations with
• street address
• construction code
• occupancy code
• Compare to prior years Summary
• Compare TIVs with limits profile
• How old is the EDM?

6
Trending the EDM prior to modeling

• Average exposure date 6 months prior to EDM
  date stamp
• Example Date Stamp 12/31/2004
• EDM has policies in force at 12/31/2004
• These policies incept 1/1/2004 - 12/31/2004
• 7/1/2004 is average exposure date
• Trend TIVs to prospective treaty period
• Average prospective date of loss trend to
  date
• Damage curve based on property values at time of
  loss

7
Dealing with your Output What do you do with
your results?

• Treaty cat loss ratio
• (Modeled treaty cat loss) / (Inforce on-leveled
  premium)
• Onlevel consistent with EDM date stamp
• Note not PROSPECTIVE Subject Premium!
• Ratio would be too low if real growth in
  portfolio.
• Example 2004 EDM produces losses of 2M
• 2004 WP 20M
• 2005 WP 35M due to expansive growth
• Cat loss ratio 2M / 20M
• On-level for rate changes.
• Otherwise, ratio too low if there were rate
  decreases
• Example 2004 EDM produces losses of 3M
• 2004 WP 30M
• Onlevel 2004 premium at 2005 rates 25M
• Cat loss ratio 3M / 25M
• Adjust for any part of Subject Premium not
  covered by cat model (e.g. International)

8
What happens if you only get aggregate cat
modeling data for a per risk treaty?

• Suppose client unable to provide EDM
• If Unicede file (aggregate data) available, run
  Catrader to get gross losses
• Use gross cat loss ratio in exposure rating model
  
• Allow property curves to layer gross cat losses
• We reselect curves that give more weight to wind
• There may be other methods to consider, but since
  we are more of an RMS company, this is what we do.

9
Examining your Cat Experience

• Take a longer time horizon
• Example may choose 5 year average for non-cat,
  but all year average for cat
• Has the book shifted?
• More coastal exposure?
• Change in management?
• Other?

10
How do you choose between Cat Experience and Cat
Modeling Results?

• Shifts in the book
• Has management changed the books direction?
• Limits shifting up or down
• More or less cat exposed
• Changes in terms and conditions
• Loss data quality
• EDM data quality
• Validity of Cat Model for these exposures
  policies
• Agreement of modeled results with recent
  experience
• How much weight would you EVER give to cat
  experience anyway?

11
Loss Sensitive Features Why including a Cat
Distribution matters

• If you model all your property exposure using
  just one distribution, you are likely missing the
  inherent volatility in the cat you are
  subsequently understating the value that the loss
  sensitive feature could have. This could lead you
  to make a decision that you may one day regret.
• And that day usually happens between August and
  November, in places like Florida.

12
Example

• Assumptions
• Subject Premium 50M
• Total Loss Ratio 60
• Non-cat Loss Ratio 30
• Cat Loss Ratio 30
• Ceding Commission 27.5
• Brokerage 1
• Profit Commission 30 after 20
• One year deal no deficit/credit carryforwards
  considered

13
What your results look like if you use a
lognormal to model all losses together

• Assume a mean of 60 with a CV of 15

14
Modeling the Cat and Non-Cat separately -
Assumptions

• Assume a non-cat mean of 30 with a CV of 10, a
  cat mean of 30 and a cat distribution from RMSs
  AEP curve.

15
What your results look like if you model the Cat
and Non-Cat separately

• Using the assumptions on the previous page

16
Can you judge a company by its Cat Modeling?

• Meeting the companys cat modeler can clarify
• Companys pricing of property business
• How company assesses cat risk
• How much company values data quality
• How well company can monitor and control its book
• Understanding what the client deems important can
  give you great insight over whether they are
  someone you even want to reinsure.
• Any reinsurer has finite cat capacity so must
  rank clients to reflect differing levels of
  quality in making underwriting decisions.

17
The Spanish Inquisition Cat Style

• Do you run Riskbrowser pre-binding or
  post-binding?
• Do you run all regions for all perils?
• How diligent are you about capturing street
  address? Construction code? Occupancy code?
• Do you turn on demand surge? Storm surge?
• What about secondary uncertainty?
• How do you think about capital allocation?
• How often do you roll up your portfolio?
• How often do you inspect insured locations?
• Do you use an external source to help keep up
  with proper valuations?
• Do you really know the values of those 25,000
  locations on that large schedule of properties?

18
Some definitions

• Primary uncertainty
• Whether or not an event will occur, and if an
  event does occur, which event it will be.
• Secondary uncertainty
• Uncertainty in the size of loss, given that a
  specific event has occurred.
• Demand Surge
• Increases in claims costs following a major
  event, due to economic, social, and operational
  factors in the post-event environment.
• Storm surge
• Rising ocean water levels along hurricane
  coastlines that can cause widespread flooding.

19
(No Transcript)