Demand Surge and Non-Modeled Losses

Post-event Loss Cost Inflation (aka Demand Surge)

Demand surge is a complex problem affecting losses in 2004:

• Volume of claims and damage (e.g. supply shortage, labor shortages)


• Close spacing of events in time & space


• Impact varies by type and class of construction and extent of damage

Repair Cost Estimate Fluctuation Jan 2004 - April 2005

Demand Surge Quantification: Key Observations

Repair cost increase ranges from 12%-55% after Charley until April 2005

Average increase ranges from 17%-38% over this time period

Timing of the repair has a lot to do with the demand surge impact

• Longer time to repair higher the demand surge


• Evidenced in the amount of reopened claims in 2004


• Expect commercial claims to be higher than residential since they take longer to close among other reasons

Analysis has focused on structure repair costs.

• Collateral damage & ALE/BI escalation expected to further impact losses

Demand Surge: Potential Impacts on 2005 Losses

Given the amount of repairs still underway in Florida it is doubtful that the key contributors to demand surge will scale back enough to appreciably reduce repair costs

State of Florida relaxing stance on out of state contractors could help speed up repairs and increase competition

Repair costs charted up to April of 2005 show demand surge has not decreased, but does appear to have leveled off

• Average for claims analyzed up to April 2005 is ~38%

Additional Loss Factors

Secondary Hazards

• Inland flooding


• Treefall


• Storm Surge (collapse)


• Tornado, Lightning


• Mold


• Power Loss, etc.

Geographic extent of loss

Insurance practices

• Claims settlement practices


• Custom policy wordings


• Off-premise outage coverage


• Loss-adjustment expenses

Isabel Example

Charley Footprint

Ivan Flood

Hurricane Isabel: Tree Related Claims

Highlights of Tree Damage Research following Hurricane Isabel

Key findings issued in December 2004 white paper:

• Evidence strongly indicates that the amount of tree fall that occurred was an unexpected factor linked to climate conditions over the preceding two years


• Our analysis indicates that these antecedent climate conditions are rare and the effect of this new variable is small on PMLs and annualized losses


• Extreme level of tree fall influenced by the low incidence of past hurricanes in the last 30 years in the area & the high density of trees in the area

Tree damage is non-modeled for areas outside the storm footprints

Extremity of the drought (3/02-9/02) and rain (3/03-9/03) conditions preceding Hurricane Isabel in Virginia Climate Division 2 (shown in lower right)— Average conditions are given by the 50th percentile, with lower values reflecting increasing levels of drought and higher values reflecting increasing levels of rain

Tree Damage in Hurricanes of 2004

Used the Urban Tree Density Index in 2004 Cat Response

Isabel research suggested that trees would be more of a factor in areas with a high urban tree density index

Tree Related Claims Analysis to Date - 2004

Analyzed combined personal lines data sets to compare frequency & severity of claims involving trees

Event	% Claims Involving Trees	Average Claim Amount Involving Trees
Charley	14%	$9600
Frances	13%	$8100
Ivan	22%	$9600
Jeanne	12%	$5400

Directionally it appears that Ivan being considerably higher in incidence of tree related claims validates link to urban tree density index

Area of further study as additional claims data sets are processed through the fall of 2005

Tropical Cyclone (TC) Inland Flooding

Growing loss source due to increasing exposure within or near flood prone areas and in areas subject to heavy rainfall events

NFIP reported 25 significant flood events associated with TCs between 1978-2003 (25% of the storms affecting the US in this period)

Event	Year	Maximum Event Rainfall	Total Insured Loss
Fran	1996	12”	$1.6 billion
Georges	1998	26”	$2.5 billion
Floyd	1999	19”	$1.8 billion
Allison	2001	36”	$2.5 billion

Hurricanes of 2004 resulted in $1.6 billion in losses to the NFI

• Biggest contributor being Ivan at $1billion

Key Parameters Impacting Rainfall

Track

Speed

Size

Shear

Topography

Extratropical Transition

Moisture

Complex interactions impacting rainfall amplitude and spatial effects

Hurricane Ivan (2004): Comparison of Gage Observations and Modeled Rainfall

TC Inland Flood Modeling: Next Steps

Evaluating various methods to link rainfall estimates to flood levels

• Because of the correlation between loss flood zones, will are first linking rain- flood by matching areas of rainfall beyond given threshold to flood zone areas


• Cat Response implementation for 2005 storm season

Claim data will be used to assess losses caused by inland flooding in historical events

Ongoing research on rainfall estimation methodology, hydrologic modeling, and linkage to US Hurricane stochastic storm set methodology (incorporation of Cat 0 storms)

Note on Non-modeled Loss Sources

Non-modeled losses a more pronounced issue for custom storm footprints issued by RMS that focus on a fine tuned representation of a wind field

For stochastic events more conservatism is built into the footprints

• Calibrated at an aggregate level


• To some extent this conservatism addresses non-modeled loss sources


• Conservatism is being evaluated as current non-modeled sources of loss are being analyzed for incorporation into modeling suite of products

Evaluating Methods: Hurricane Frances

Actual stochastic event IDs vs. post event footprint

Simulated stochastic profile vs. post event footprint

Simulated Stochastic Windfield Footprint

Post-Event Footprint

U.S. Industry modeled losses

Stochastic Windfield: $5.0 bn

Post-Event Footprint: $3.2 bn