Presentation to CARE Conference Boston, MA June 15th, 2000
Overview What models are there? What are they good for? What aren’t they good for? Four examples of model-inspired reinsurance structures Pitfalls 1
Using Cat Models What models do we use? EQECat US Wind US Quake World Cat RMS IRAS 3.7 4.1 being tested AIR Catrader CLASIC Dames & Moore Tempest Re proprietary 2
What models do we use? 3
What are they good for? Underwriting/Pricing Consulting with clients Portfolio/Capital Management 1 4
Underwriting with Cat Models Which perils are best suited to models? Hurricane Earthquake Flood Extra tropical storm 5
Underwriting with Cat Models Which perils or issues are not? Winter storm Tornado/Hail Freeze Fire - Brush & Fire Following Volcanic eruption Riots, etc Business Interruption “Cross territorial” clash “Non landfall” events (Floyd) 1 6
How do we use them? Assess the data Run the models simultaneously Collate the output Evaluate on differences LOB Data Region Peril Cat Management credit 7
Loss Pick 8
Loss Pick 9
ROL of 17.75% would produce sufficient ROE (15%) Pricing ROE = (Price*.9-0.11)/0.45+.04 ROL of 17.75% would produce sufficient ROE (15%) 10
Limit Example Florida Coverage to attach at a defined PML - e.g. 30 years Limit to track 100 year PML Price is simply a ROL applied to calculated limit ($40mm) 11
Limit Example Florida Limit and retention drop due to decrease in exposure ROL now applied to a limit of $25mm excess $15mm No more exposed than before, just lower down in dollar terms 12
Limit Example Florida Now limit has increased beyond year one $62mm excess $38mm 13
Limit Example Florida Limit is usually capped ($65mm) 14
Limit Example Florida Even if 100 year pml is higher than $65mm excess $46mm Results in more retained net 15
Limit Example Florida For Florida covers, a major complication is FHCF Cover becomes a “Wrap” around FHCF Warrant a minimum FHCF cover amount 16
Pricing Example California Earthquake Start up Residential EQ writer Only needs limit as exposed on a month-to-month basis Calculation: Rate*(250 yr PML - attachment)*(GU Loss*probability)/10.5 17
Swap Example Trade limit of coverage at pre-agreed EL Requires transparent modeling Like perils, or even diverse perils Florida wind for Japan wind Capital efficient Could charge loss cost, or identical premium Real benefit is release of capital 18
Structure Example Models are not credible for some perils Occurrence structure (hours clause) is not best suited to all perils Not all clients fit into a model-designed solution Examples Other Wind Winter Weather 19
Structure Example Client wanted all ww claims to be classified as one occurrence - models can’t do that Occurrence definition for cat cover would need to be broadened Strip out ww claims & simulate trended historical experience Add in other perils on more traditional hours clauses, using models Capital selection is tricky 20
Pitfalls Blocking and tackling is still required Coverage A only Unicede Exposure base as of when? Commercial exposures are the same as Personal Line exposures Event Set is incomplete Model myopia 21
Pitfalls 250 Yr PML 22
Pitfalls 250 Yr PML 22
Conclusion The old way of underwriting is no good 23
Conclusion The old way of underwriting is no good Models know best 23
Conclusion The old way of underwriting is no good Models know best Models don’t know everything 23
Presentation to CARE Conference Boston, MA June 15th, 2000