adjustments to cat modeling cas seminar on cat sean devlin september 18, 2006
TRANSCRIPT
Adjustments to Cat Modeling
CAS Seminar on CatSean DevlinSeptember 18, 2006
Slide 2
TCNA Adjustments - Climate
Options on Using Climate Forecasts
Find no credibility in the forecasts
Believe that the forecasts are directionally correct
Believe completely in the multi-year forecasts
Believe completely in the single year forecasts
Slide 3
TCNA Adjustments - Climate
Option 1 - Find no credibility in the forecasts
Use the a vendor model based on long term climate
Adjust the loss curve down of a vendor model that has increased frequency/severity
Use own model
A blend of the above
Slide 4
TCNA Adjustments - Climate
Option 2- Believe that the forecasts are directionally correct
Credibility weighting between models in option 1 and a model with frequency adjustments
Adjust a long-term model for frequency/severity
Adjust long-term version of a vendor model
Adjust own model for frequency/severity
Combination of the above
Slide 5
TCNA Adjustments - Climate
Option 3 - Believe completely in the multi-year forecasts
Implement a vendor model with a multi-year view
Make frequency/severity adjustments to a long term vendor model
Adjust own model
Blend of the above
Slide 6
TCNA Adjustments - Climate
Option 4 - Believe completely in the single year forecasts
Implement seasonal forecast version for a vendor model
Adjust vendor model for frequency/severity
Adjust internal model for frequency/severity
Combination of the above
Slide 7
TCNA Adjustments – Frequency/Severity
Adjust whole curve equally
Ignores shape change
Treats all regions equally
Adjust whole curve by return period/region
0%
50%
100%
150%
200%
250%
ST/LT 1
ST/LT 2
Slide 8
Modeled Perils – Other Adjustments
Actual vs. Modeled – look for biases (Macro/Micro)
Other Biases in modeling
Exposure Changes / Missing Exposure/ITV Issues
LAE
Fair plans/pools/assessments
Demand Surge
Pre Event
Post Event
Slide 9
Unmodeled Exposure
Tornado/Hail
Winter Storm
Wildfire
Flood
Terrorism
Fire Following
Other
Slide 10
Unmodeled Perils
Tornado Hail National writers tend not to include TO
exposures Models are improving, but not quite there yet Significant exposure
Frequency: TX Severity:
2003: 3.2B – 12th largest 2001: 2.2B – 15th largest 2002: 1.7B – 21st largest
Methodology Experience and exposure Rate Compare to peer companies with more data Compare experience data to ISO wind history Weight methods
Slide 11
Unmodeled Perils
Winter storm Not insignificant peril in some areas, esp.
low layers 1994: 100M, 175M, 800M, 105M 1993: 1.75B – 18th largest 1996: 600M, 110M, 90M, 395M 2003: 1.6B # of occurrences in a cluster????? Possible Understatement of PCS data
Methodology Degree considered in models Evaluate past event return period(s) Adjust loss for today’s exposure Fit curve to events
Slide 12
Unmodeled Perils
Wildfire Not just CA Oakland Fires: 1.7B – 19th largest Development of land should increase
freq/severity Two main loss drivers
Brush clearance – mandated by code Roof type (wood shake vs. tiled)
Methodology Degree considered in models Evaluate past event return period(s), if
possible Incorporate Risk management, esp. changes No loss history - not necessarily no exposure
Slide 13
Unmodeled Perils
Flood Less frequent Development of land should increase frequency Methodology
Degree considered in models Evaluate past event return period(s),if possible No loss history – not necessarily no exposure
Terrorism Modeled by vendor model? Scope? Adjustments needed
Take-up rate – current/future Future of TRIA – exposure in 2006 Other – depends on data
Slide 14
Unmodeled Perils
Fire Following No EQ coverage = No loss potential?
NO!!!!! Model reflective of FF exposure on EQ
policies? Severity adjustment of event needed, if
Some policies are EQ, some are FF only Only EQ was modeled
Methodology Degree considered in models Compare to peer companies for FF only Default Loadings for unmodeled FF Multiplicative Loadings on EQ runs
Slide 15
Unmodeled Perils
Other Perils Expected the unexpected Examples: Blackout caused unexpected
losses Methodology
Blanket load Exclusions, Named Perils in contract Develop default loads/methodology for an
complete list of perils
Slide 16
Summary
Don’t trust the Black Box Understand the weakness/strengths of
model Know which perils/losses were modeled Perform reasonability checks Add in loads to include ALL perils Reflect the prospective exposure