Summary

The I.I.I. has estimated the insured loss from hurricanes Charley, Frances, Ivan and Jeanne to be $22.5Bn. This and the other published estimates are likely too low. Holborn estimates that ultimate losses for US insurers will be well over $30Bn. This will be recognized over the next several months and will hurt the financial results that insurers and reinsurers book in upcoming quarters.

Published estimates of storm losses, especially for the later events, are based on catastrophe simulation model results. These models largely reflect the physical forces in each event considered in isolation. Considered together, the losses compound each other in several ways. The models cannot reflect that:

• Applying multiple deductibles will be practically and politically difficult. We believe hundreds of thousands of deductibles at 1% - 2% of building values will be waived, at an average cost of $3,000, providing an additional $2Bn in insured loss.


• Compounded damage follows from a storm striking an area already damaged by one or two prior storms. This includes increased structural damage, rain through damaged roofs and windows and wind-blown debris. Clearly these are major issues with these storms, although difficult to quantify. The modelers have not considered this scenario before and there are no prior data points.


• Demand surge levels reflect aggregate damage over all four events, and including within- deductible amounts. Also, Florida's economy is stronger in 2004 than in 1992 during Andrew, and there is less slack. We believe demand surge will be unprecedented and may add an extra 15% to modeled losses, or $3Bn.


• Time element coverages (living expenses, business interruption, loss of rents and extra expense) are aggravated by limited building materials, as in other large events, but also by weather interruptions. Repairs have been interrupted both by poor weather and by evacuations from following storms. We believe time element losses will be doubled, adding over $1Bn in loss.


• Loss adjustment expenses are not included in the models and will likely be over $2Bn, recognizing 2 million Homeowners claims in Florida, alone, and the cost of travel, lodging, meals and supplies for out-of-state and contracted adjusters.


• PCS and the I.I.I. omit many surplus lines companies which are notable writers of Florida Commercial business and some Personal business.


• US insurers also participate on the off-shore energy platforms and pipelines (over $800Mn in worldwide industry loss from Ivan) and the Caribbean (over $5Bn in loss, with US insurers involved at least in Puerto Rico on Frances and Jeanne).

The Florida Hurricane Catastrophe Fund (FHCF) covers Florida residential property in the industry loss levels between $4.5Bn and $19.5Bn. Cat Reinsurers have estimated that 27 companies are attaching the FHCF, based on their model results and the current published estimates. We believe most eligible companies will actually cede one or more losses to the FHCF. While no individual storm will reach $19.5Bn in Florida residential loss, we believe that several companies have exhausted their individual coverage. Reinsurance market placements cover losses for almost all companies above the FHCF. Reinsurers will assume more losses above the FHCF than they would estimate from current model results.

While some reinsurers may point to these losses as a reason to raise rates at January 1, 2005, virtually all are still reporting reasonable earnings for the year. Current reported losses may stabilize or slow the rate of decrease in rates for most buyers outside of Florida. We expect that the market will understand the entire loss situation late in 2004, and, while the jury is still out, rates for 2005 mid-year renewals may come under pressure. Companies that have ceded losses to the market will undoubtedly see rate increases in their renewals.

These losses will cause the Cat modelers to rethink some aspects of their models. In late 2005 and 2006, reinsurers will use updated models that will likely have higher US occurrence loss estimates, and especially higher total aggregate loss estimates. Rating agencies will pay increased attention to aggregate loss exposure, in addition to the questions they ask about per occurrence PMLs. This may increase demand for reinsurance coverage which would add rate pressure to the market.

Loss Estimates

US insurers sustained $30Bn to $40Bn in gross losses from the 2004 hurricane season. In constant dollars, this is significantly more than the $25Bn that they sustained in 1992 ($22Bn on hurricane Andrew in Florida and $3Bn on Iniki in Hawaii). In nominal dollars, this is by far the worst year for US natural catastrophes ever. The US loss of life is the worst since 1969, when hurricane Camille killed 256 people.

We estimate the gross losses for US insurers by event to be:

	Published Estimates, Events Considered Separately		Estimated Ultimate, Events Considered Together
Tropical Storm Bonnie	$300Mn		$300Mn
Charley	6Bn – 7Bn		7Bn – 10Bn
Frances	3.5Bn – 4.5Bn		4Bn – 6Bn
Ivan On-Shore	4Bn – 6Bn		5Bn – 8Bn
Jeanne On-Shore	4Bn – 8Bn		8Bn – 14Bn
Caribbean, Puerto Rico, Platforms	500Mn – 2Bn		500Mn – 2Bn
Loss Adjustment	NA		2Bn – 2.5Bn
Total Gross	$20Bn – $25Bn		$30Bn – $40Bn

We estimate that the losses will fall by major line and state as follows:

Florida Residential*	$19Bn – $26Bn
Florida Other	5Bn – 7Bn
Adjacent States Homeowners	2.5Bn – 3.5Bn
Adjacent States Other	0.5Bn – 1.0Bn
Other States	0.5Bn – 1.0Bn
Off-Shore (US insurers only)	0.5Mn – 2Bn
Loss Adjustment	2.0Bn – 2.5Bn
Total Gross	$30Bn – $40Bn

Net Losses by Market Segment

To understand the impact on the market, it is important to see how the net Florida residential losses will fall. The Homeowners line in Florida has $4Bn in gross written premiums for voluntary business. This excludes Citizens Insurance Company. Florida is notable for the number of new specialist companies writing mostly Homeowners business. Many were formed only to depopulate the former JUA and are highly exposed in the Southeast part of the State. Those companies write almost one-fifth of the voluntary market.

	2003 DWP		Total Market Share		Voluntary Market Share
8 National Personal Companies	$2.10Bn		42%		54%
Regional and Other Multi-Line Companies	1.03Bn		21%		26%
20 Florida HO Specialists	0.75Bn		15%		19%
Citizens Insurance	1.10Bn		22%		0%
Florida HO Market	$5.0Bn		100%		100%

Losses do not strictly follow companies' market shares, especially because the southeast part of the state was largely spared by the storm tracks. Despite a 15% share, we believe that the 20 Homeowners specialists may only have as little as 10% of the Florida residential loss. Our estimates for the gross and net Florida Homeowners losses follow. (We have more confidence in our estimates for the entire market, and selected relatively wider ranges for the segments.)

	Insurers’ Gross Losses	FHCF Cessions	Reinsurance Market Cessions	Insurers’ Net Losses
National Personal Companies	$7 – 15	$1.0 – 3.5	$0.5 – 2.0	$7.0 – 10.0
Florida HO Specialists	2.5 – 5	0.5 – 2.0	0.5 – 2.5	2.0 – 3.5
Regional and Other Multi-Line	7 – 10	1.0 – 2.5	0.5 – 2.5	3.5 – 6.5
Citizens Insurance	1.5 – 3	0.5 – 1.5	0 – 0	1.5 – 2.5
Total Florida Homeowners ($Billions)	$18 – 30	$4 – 8	$2 – 6	$11 – 18

Even reflecting other lines and states, the net losses will not cause significant distress to US insurers based outside of Florida.

	Net Losses	12/31/03 PHS	Florida 2003 GWP	Countrywide 2003 GWP
National Personal Companies	$12 – 18	100+	$10.7	135.0
Florida HO Specialists	2.0 – 3.5	0.25	0.75	0.75
Regional and Other Multi-Line	5 – 8	250+	16.1	314.5
Citizens Insurance	1.5 – 2.5	1.8	1.1	1.1
US Primary Market	11 – 18	359.7	30.2	451.3
FHCF	4 – 8	5.5	1.5	1.5
Property Reinsurance Market	4 – 10	50+	2 – 3	10 – 20
Total Market ($Billions)	$30 – 40	400+	34+	460+

Financial Implications

Third quarter earnings have been down at many publicly traded insurers, but diversified companies have not had surplus declines attributable to these losses.

Citizens Insurance will run at a net loss and may exhaust its surplus. It will assess this deficit against the industry, with the assessments recouped by policyholder surcharges in 2005.

The FHCF has depleted its surplus, as well. Even if the surplus is not exhausted, the FHCF has the authority to issue bonds guaranteeing its claims-paying ability for the rest of this season and for 2005. FHCF bonds will also be serviced by assessments recoupable against future policyholder premiums.

With bonding, the FHCF is probably able to renew its full capacity next year, perhaps somewhat less. But it cannot commit now to full "following season" capacity for 2006.

There will be distress among the new Homeowners specialist companies. Several will fail and most will need to recapitalize or re-trench. These companies were thinly capitalized and highly exposed. This is partly because much of their business was taken from the former JUA and wind pool, having been declined by traditional insurers. A typical representative of these companies has a 1% market share and thus has a pro-forma financial profile like this:

Based on market shares, the FHCF provides a company like this with reinsurance of $135Mn in excess of $40Mn (Specific company layers depend on their spread by zipcode.) All of these companies chose 90% coverage (and the larger layer of coverage which was an option in 2004 only).

The Florida Insurance Department required these companies to protect themselves against their 1-in- 100 year loss. Thus, the reinsurance program for a typical one of these companies was structured like this:

With a single $250Mn occurrence, this company would sustain a $17.25Mn net loss, and be (just) able to meet its obligations and pay for reinstatements on its open market reinsurance placements.

These companies in total sustained gross losses of $2.5Bn to $5.0Bn, or an average of $125Mn to $250Mn each. Depending on their spreads within Florida, companies have their individual largest losses on different storms. Our typical company has at least one loss over $50Mn. Most of these new companies attached the FHCF on at least one event. Some will exhaust it.

Sustaining two or more retentions, and paying to reinstate the first layer, at least once, will be a significant cost to these companies. For our typical company, two retentions and one reinstatement, and 5% co-participations, would cost at least $20Mn and more likely $25Mn. Surplus would decline by $3Mn to $6Mn after tax. Gross premium to surplus would rise from 3:1 to 4:1 or even 5:1. Resulting RBC levels will require regulatory attention in 2005. These figures are for average companies. Companies with significantly better results are not so stressed. The companies with significantly worse results failed.

Assumed Reinsurance

Regional companies have meaningful losses to their reinsurance programs, especially from commercial lines and outside of Florida. National companies tend to purchase Catastrophe programs that do not cover Florida losses below the FHCF. We do not believe that the largest national companies exceeded their FHCF protection. Our estimate of $4Bn - $8Bn of net ceded loss to the reinsurance market reflects the market's overall participation on Property Cat covers, as well as estimates for Per Risk, proportional and finite covers. This does not reflect direct losses when the business was written as surplus lines or binding authority carriers.

Recognizing the strong profits in the first half, the reinsurance market is still marginally profitable for the first nine months. They have reported $8Bn in pre-tax losses on these four storms and the two largest Japanese typhoons. But the first half profits for Bermuda reinsurers alone was over $9Bn. Property Cat specialists should expect an unprofitable year, especially those writing Per Risk, Pro-Rata and Facultative, as well as Cat. Early estimates are for losses on North American assumed Cat books of between 40% and 120% of premiums. These are also too low. Our analyses of on-shore property reinsurers confirms this.

Accident Year	Net Earned Premiums	Estimated Ultimate Loss Ratio	Average Renewal ROL Increase/Decrease
2000	$4.0Bn	73%	-10%
2001	$4.2Bn	153%	+10%
(2001 ex 9/11)		(73%)
2002	$4.9Bn	40%	+35%
2003	$5.2Bn	45%	-5% – -10%

Our estimates for 2004 were for an excellent year, except for these losses:

2004 Loss Ratio Scenario
With Normal Cats	50%
With No 3rd Quarter Cats	30%
+ Hurricane Losses	75% – 125%
2004 Revised Estimate	105% – 155%

Market Trends

A few reinsurers are already pointing to $25Bn in hurricane losses as a reason to raise Catastrophe rates; however, most do not recognize the full cost as yet, and they will add resolve as they learn more. For renewing accounts with meaningful losses some form of payback will be sought, and often achieved. As reinsurers examine their mounting aggregate losses, they will re-evaluate their overall Florida and Southeastern portfolios and this will have an impact on prices in the affected region.

Insurers exposed in the Southeast may seek to place more reinsurance coverage in 2005. This will include more "sideways" cover, driven by this year's frequency of losses. There may also be further reviews at both the top- and bottom-end of programs, in response to reductions in policyholders' surplus and the weakened capital base of the FHCF. This increased demand will cause Property Catastrophe reinsurance rates to increase in the Southeast, even for "clean" accounts. Other factors pushing up Property Catastrophe reinsurance rates include the psychological impact from the year's unprecedented frequency of severe events, and the "wake up call" that losses could have been much, much worse if either Miami or Tampa had been involved. There is often a spillover when accounts priced for payback raise the prices for clean accounts.

Most retrocessions are placed on an "Industry Loss Warranty" basis with retentions higher than these losses, although some "second event" covers will have losses. We do not expect much loss to be ceded to the retrocessional catastrophe market. The cost of reinsurers' own protections will increase, but not significantly, as there will be no sharp loss of capacity and new capital is presently entering the retrocessional market through hedge fund transactions.

Modelers will reconsider their hurricane frequency assumptions. The overall assumed frequency of hurricanes may not change, since 2004 was not a remarkably active year for the entire Atlantic. However, we believe that they will rethink:

• The chance of a hurricane making landfall in the US.


• The chances of multiple US losses in the same year.


• Whether tracks of subsequent events are correlated.

These changes will not happen in time to significantly influence any 1/1/05 renewals and not all 2005 midyear renewals. But new models will be used by almost all reinsurers by 1/1/06. This will have an intermediate-to-long term affect on pricing and reinsurance buying strategies.

Outside of the Southeast, the basic business fundamentals remain. Capacity is greater than demand. Underwriting year results are excellent. We expect the hurricane losses will have only a moderate impact on other states' pricing.

Florida will see continued upheaval. Multi-state writers will have reduced appetites for Florida, especially if proposals to limit windstorm deductibles to an annual aggregate basis are approved. Some companies will withdraw or look to non-renew some business. Premiums will rise and be compounded with FHCF, Citizens Insurance and Guarantee Fund assessments. This will only increase the political pressure in 2005.

Exhibits

I. Track Maps

II. Details on the Storms

III. Company Loss Estimates

IV. Market Share Analysis

EXHIBIT I: 2004 Land-Falling Hurricane Tracks

EXHIBIT II: Details on the Storms

Bonnie

Tropical Storm Bonnie came ashore in Apalachicola, Florida on August 12th. Winds were 55mph, and damage was light. Leon County with 239,000 people saw the largest amount of exposure.

Charley

Charley suddenly gained strength and veered to the right on the afternoon of August 13th and made landfall at Cayo Costa, in Charlotte County, Florida. Prior to landfall, the core of the storm passed over Sanibel and Captiva Islands. The landfall point is 80 miles south of Tampa and 30 miles north of Fort Myers. Peak winds were 140mph. The eye was one of the smallest ever seen. The tight eye, deep central pressure and high winds kept Charley better organized than the models would suggest. Significant damage continued inland as far as Orlando. Charley left the state near Daytona as a "1", gained strength over the ocean and landed in North Carolina as a "2" on the 15th.

The most affected counties were: Charlotte, Desoto, Flagler, Hardee, Henry, Lee, Orange, Polk, Sarasota, Seminole and Volusia. The population in these eleven counties is 3.2Mn.

Frances

Frances landed in east central Florida late at night on September 4th as a "2" with sustained 105mph winds. The center came ashore at 1 AM on the 5th near Stuart, which is about 30 miles north of Palm Beach and 30 miles south of Ft. Pierce. The storm was very wide, with a 75 mile wide "eye", but that width means that the storm's strength was not tightly packed around the center. Because of the width and slow movement, storm-strength winds were felt for at least 18 hours as the storm passed overhead.

Tide was near maximum at 1 AM, so surge and coastal flooding were significant, but largely uninsured.

Hurricane force winds (75+ mph) were reported from south of Lauderdale to north of Cocoa Beach, a distance of 200 miles. Hurricane winds were felt inland from Orlando to Tampa on the Gulf Coast. Damage was widespread but not locally as extreme from Charley. Wind patterns were very symmetrical because of the slow movement, so wind damage was fairly evenly spread and not focused on the right side.

News reports did not describe any buildings blowing apart as in Charley, but mentioned power outages, wave action, flood, treelimb strikes and very extensive rain. Rainfall over a foot was common. Frances exited as a tropical storm, regained strength over the Gulf and came onshore again in the Panhandle as a "1."

The most effected counties were in the central part of the state: Brevard, Glades, Highlands, Indian River, Martin, Okeechobee, Osceola, Palm Beach, Polk and St. Lucie. The population in these ten counties is 2.8Mn.

Ivan

Ivan came ashore shortly after 3 AM (NY time) September 16th, passing over Mobile Bay. The city of Mobile was hit by the left side of the eyewall. Winds were down to 130mph at landfall, which is the top of the "3" range. Pressure was 943 mB, still just in the "4" range. Movement was due north at 13 mph.

Storm surge was 10 to 16 feet and coastal waves were up to 25 feet. Total rainfall was again about one foot. Tide was about average at landfall and rising. Tornados broke out in several locations between Pensacola and Tallahassee.

Tropical storm force winds extended 290 miles and reached from New Orleans past the Tampa area. No wind damage occurred in these two cities from Ivan. Hurricane force winds were felt out to 115 miles on the right side, which would include Fort Walton and Panama City, Florida. Pensacola saw very strong winds with the eye passing within 30 miles. Gulfport, Biloxi and Pascagoula, MS also sustained hurricane force winds. Over 1 million people live near these six cities. The most effected counties were Baldwin and Mobile, Alabama; Plaquemines and St. Bernard, Louisiana; Harrison and Jackson, Mississippi; and Escambia, Okaloosa, Santa Rosa and Walton, Florida. These ten counties have a population of 1.7Mn.

Jeanne

Jeanne made landfall late at night on September 25th, at Hutchinson's Island, Florida. That is about 5 miles north from Stuart, where Frances came in, and about 35 miles north of Palm Beach. The track was west - northwest at 13 mph. The tide level was medium and falling, moderating the wave damage. Storm surge was 4-6 feet.

Sustained winds were 120 mph, which made Jeanne a "3". Jeanne was stronger but somewhat smaller than Frances. Frances was a wider than average storm, and both were much wider than Charley, which was unusually focused. Areas that saw both Frances and Jeanne report that Jeanne was more intense and damaging.

The most effected counties were: Brevard, Highlands, Indian River, Okeechobee, Orange, Osceola and St. Lucie. The population in these seven counties is 2.0Mn.

Overview

In total, 29 counties with 7.6Mn people were significantly exposed to these storms. Eight Florida counties: Brevard, Highlands, Indian River, Okeechobee, Orange, Osceola, Polk and St. Lucie, with 2.5 million, people saw two major storms.

Florida's five largest Counties with over 4.8Mn people were largely spared. Part of Palm Beach was exposed to Frances, and Broward, Duval, Hillsborough and Dade did not experience hurricane force winds in 2004. Differences of 50 miles or less in the storm tracks could have multiplied the losses on any of these storms.

EXHIBIT III: Company Loss Estimates as of October 18th ($ Millions)

EXHIBIT IV: Market Statistics

Description	Amount (000's
Florida Total DWP HOMP	$3,848,066
Florida Total DWP CMP*	$1,838,335
Florida Total DWP CMP (NL)	$1,088,575
Florida Total DWP CMP (L)	$749,760
Florida Total DWP FOMP	$21,013
Florida Total DWP Allied	$782,459
Alabama Total DWP HOMP	$818,856
Alabama Total DWP FOMP	$447,535
Mississippi Total DWP HOMP	$476,292
Mississippi Total DWP FOMP	$11,095
Countrywide Total DWP CMP*	$31,464,820
Countrywide Total DWP CMP (NL)	$19,042,178
Countrywide Total DWP CMP (L)	$12,422,642
Alabama Total DWP All Lines	$5,676,778