The Progressive Corporation (PGR) 8-K/A Report on Loss Reserving Practices, dated June 2003

Exhibit 99

Report on Loss Reserving Practices, dated June 2003

REPORT ON LOSS RESERVING PRACTICES THE PROGRESSIVE CORPORATION JUNE 2003

Preface

Our desire for transparency demonstrates our belief that good decisions flow from clear information. With that in mind, we are releasing this report. Our goal is to help investors, analysts, employees and all interested stakeholders better understand how we reserve and improve their comprehension of our business and financial statements.

The focus of this report is to:

• Provide an in-depth perspective of how we view loss reserves

• Offer an updated summary of how loss reserve development can influence current year results

• Give readers an expanded overview of Incurred But Not Recorded (IBNR) reserves and how lag factors assist us in reserving for late reported claims

• Provide a detailed explanation of loss reserve segmentation and why it is important in our reserving process

The 2003 report enhances last year’s report in the following ways:

• Explanation of why accurate reserving is important for pricing, capital adequacy and investment strategy

• Expanded detail on 2002 process enhancements

• Two new case studies have been written and they are included in a separate Appendix. The case studies illustrate many of the concepts discussed in this report. See the “Appendix” to view these case studies or go to Progressive’s website @ progressive.com. This appendix includes exhibits and calculations used in typical reserve reviews, as well as discussions of the issues involved. Section VI is a loss reserve case study and Section VII is a loss adjustment expense reserve case study.

We welcome comments and criticism. Such feedback helps us clarify our thinking.

Comments and questions should be directed to: Kevin Rosenstein, Corporate Actuarial Manager or Sara Stehlik, Director, Risk Management at The Progressive Corporation, 6300 Wilson Mills Road, Mayfield Village, Ohio 44143 or e-mail kevin_rosenstein@progressive.com or sara_stehlik@progressive.com.

Table of Contents

Section I: About Progressive
	Our Business	1
	2002 Highlights	1
	Our Financial Objectives	2
	Reserve Data used by the Pricing Function	3
Section II: About Reserves
	Definition and Stated Goals	4
	Reserve Development	5
	External Reporting of Reserve Development	6
	Development Patterns	7
Section III: Types of Reserves
	Loss Reserves	8
	Case Reserves	8
	Average Reserves	9
	Inflation	9
	Threshold	9
	Adjuster Set Reserves	9
	Additional Needed Case Reserves	9
	Incurred But Not Recorded (IBNR) Reserves	11
	Loss Adjustment Expense (LAE) Reserves	13
	Defense and Cost Containment (DCC)	13
	Adjusting	13
	Salvage and Subrogation	13
	Involuntary Market Operating Loss Reserves	14
	Private Passenger Assigned Risk Plan	14
	Commercial Auto Insurance Plan	15
Section IV: Estimating Loss Reserves
	Segmentation of Reserves for Analysis	16
	Estimation Methods Defined	17
Section V: 2002 Process Enhancements
	Segmentation	19
	Analysis Process	21
	Data Alignment	21
	Modeling	23
	Communication	24
Terms and Definitions		25

Section I: About Progressive

Our Business:

Since the Progressive insurance organization began business in 1937, we have been innovators- growing into new markets and pioneering new ways to meet consumers’ needs. In 1956, Progressive Casualty Insurance Company was founded to be among the first specialty underwriters of nonstandard auto insurance. Today, the Progressive group of companies (collectively referred to as “Progressive” or the “Company”) provide drivers throughout the United States with competitive rates and 24-hour, in-person and online services, through 68 subsidiary companies and two affiliates of The Progressive Corporation.

We aspire to be recognized as the preeminent consumer franchise in auto insurance and other specialty property-casualty insurance throughout the United States. The Company’s personal lines consists predominately of insurance for automobiles, motorcycles, recreational vehicles and travel trailers. We also offer Commercial Auto liability, physical damage and related coverages to small businesses countrywide. Our business is either generated by the Company’s network of more than 30,000 independent agencies or is written directly by the Company via the telephone (1-800-PROGRESSIVE) or via the Internet (progressive.com).

2002 Business Highlights:

Overall, it was a great year. We became the third largest private passenger auto insurer in the country, net written premiums grew 30% and we attained a 7.6% underwriting margin. Our growth came at an opportunistic time: results benefited from price adequacy, which allowed us to assume market share from our competitors. We experienced favorable loss frequency, which provided our claims infrastructure room to breathe as they were pushed to increase and train a growing number of claims representatives and we encountered better than expected policy life retention, although now flattening, helping our rate of growth in 2002. We made no significant changes to prior year loss reserve estimates causing little influence on calendar year results, and the capital required to meet the demands of our growth was available at favorable rates.

As reported last year, expense management remains critical. Our Expense Ratio of 21.5% as of 12/31/02 remained generally flat over the prior year and we continue to focus on keeping our expenses low. We became more efficient with our advertising expenditures during the year, as spending was comparable to the prior year but yielded better results. We constantly push ourselves to identify and implement processing and electronic advancements making it easier and more efficient for our agents and customers to do business with us.

Page 1

Our Financial Objectives:

At Progressive, we measure ourselves against two specific goals designed to maximize the value of our firm. First, we aim to achieve a four percent underwriting profit on all products. Second, we aim to grow our business as fast as possible so long as doing so is consistent with our first goal and our ability to provide high quality service to our customers. We communicate these two corporate goals to every Progressive employee, and work together to achieve these goals. In 2002, our achievement of these two goals resulted in a Return on Shareholders Equity of 19.3%¹ or a comprehensive ROE of 20.5%².

At Progressive, we also identify and actively manage a wide variety of risks that could affect the achievement of these goals. We have classified these various risks as Underwriting (maintaining an adequate underwriting profit), Financing (maintaining adequate capital to support growth), and Investing (maintaining a liquid, diversified investment portfolio that withstands adverse market trends). We have also adopted a set of financial policies that specify how we intend to manage each of the risks we have identified.

Loss reserving is an activity that is central to the achievement of our goals. It involves estimating the magnitude and timing of future claim payments on past and current policies. These estimates take into account not only claims that are in the process of being settled, but also claims on accidents that have happened but have not yet been reported to the Company. At year-end 2002, Progressive’s estimated gross loss and LAE reserves amounted to $3.8 billion.

Accurate reserving is important for pricing, for capital adequacy, and for investment strategy.

Pricing. Estimates of future claim payments are essential for accurately measuring Progressive’s underwriting profit and for determining whether pricing changes are needed to achieve the target four percent underwriting profit. Reserve estimates that are too low can lead to the conclusion that pricing is adequate when it is not, so that we fail to achieve our underwriting target in future periods. Reserve estimates that are too high limit growth opportunities and establish a price umbrella for inefficient competitors.

Capital.Accurate reserve estimates facilitate proper capital management. When reserve estimates are accurate, there is less of a need to raise additional capital which costs the company money. Incorrectly estimating reserves can also distort our measurements of the capital available to support business growth, and lead us to grow too quickly or too slowly.

Investment strategy. Estimates of the magnitude and timing of future claim payments influence the formulation of an investment strategy for managing both liquidity and interest rate risk. With respect to liquidity, Progressive’s policy of investing predominantly in highly marketable securities enables it to withstand unanticipated cash outflows resulting from catastrophic events (e.g., hurricanes and earthquakes), as well as deviations from the estimated timing of future claim payments. In order to control interest rate risk, Progressive manages its investment portfolio to limit the difference between scheduled future cash inflows from invested assets and estimated future cash outflows to pay claims and other liabilities.

¹ Based on net income

² Use of comprehensive ROE is consistent with the Company’s policy to manage on a total return basis and reflects changes in unrealized gains and losses on securities held in our portfolio. For Progressive, Comprehensive ROE consists primarily of:
(net income + changes in unrealized security gains, net of tax) / average shareholders equity.
To review all components of Progressive’s Comprehensive ROE refer to our Consolidated Statement of Changes in Shareholders Equity in our 2002 Annual Report.

Page 2

Reserve Data used by the Pricing Function:

Unlike most industries, insurers do not know their costs until well after a sale has been made. One of the most important functions in our Company is setting rates or “pricing”. The goal for our pricing function is to properly evaluate future risks the company will assume but haven’t written. The pricing function is very different from the loss reserving function, although both functions use similar data to do their job. Typical information that the loss reserve analysts produce that is also used by the pricing organization to help determine future rate needs includes:

• Overall changes in the level of reserves (case, IBNR and total)

• History of claim development and selected ultimate losses by accident period

• Changes in claim closure rates

• Selected severity by accident period and resulting trends

• Selected frequency by accident period and resulting trends

• Changes in selected ultimate loss amounts over time

• Changes in case average reserves by age

• Changes in the rate of claims closed without payment (CWP rate)

• Changes in the level of average adjuster case reserves

As outlined in last year’s report, the role of the pricing function is to determine premium rates that are adequate to achieve our profitability goals without being excessive or unfairly discriminatory to our customers. When we price our programs, we do so without the prospect of reinsurance so pricing for catastrophe losses is embedded into establishing rate.

Our pricers use a cost-plus strategy, beginning with the projected ultimate loss and loss adjustment expense amount and then add estimated future organizational expenses to that amount plus a four percent profit load. As we collect information, we look at trend selections from prior reviews to determine if they are consistent with what we expected or if they have changed. When they have changed, we determine, by line coverage (bodily injury, property damage, collision, personal injury protection, etc.) if the change is driven by changes in frequency (number of claims per earned car year) or changes in severity (dollars of loss per claim). We also look at current claims management practices and consider whether growth or process changes have caused the claims adjustment process to be faster or slower than previous experience. New claims strategies clearly affect loss statistics and ultimately roll into trend statistics. Changes made by state insurance departments and changes in the underwriting process may also contribute to unexpected changes in the data we see.

The pricing function must also anticipate the rate approval process each state follows prior to implementation of proposed new rates. The pricer must understand rules, regulations and the lead time needed to obtain rate approval. In addition, the pricer is responsible for establishing premium rates by five different underwriting “tiers.” They are: Non Standard; Middle Market; Standard; Preferred and Ultra Preferred. Once the “rate need” or indication has been determined, pricers must then determine, subject to applicable legal requirements, whether to apply the rate need (or rate relief) evenly to each policy, to new business only, to renewal business only, or to execute an “off balance” rate revision making adjustments to one or several rating variables such as make, model and year of the vehicles we insure, or to apply the new rate to a specific group of insureds by age, geographic location, limits, line coverage, etc.

During 2002, we performed over 90 auto rate and program revisions.

Page 3

Section II: About Reserves

Definition and Stated Goals:

Loss reserves are the Company’s best estimate of amounts needed to pay unsettled claims. Our goal is to ensure that total reserves are adequate and develop with minimal variation. Exhibit 1 illustrates our performance against this goal.

Exhibit 1

Analysis of Loss and Loss Adjustment Expenses (LAE) Development
(not covered by report of independent accountants)

There is inherent uncertainty in the process of establishing property and casualty loss reserves. Reserves are liabilities established on the company’s balance sheet and are estimates, based on facts and circumstances then known, of the unpaid portion of what the company ultimately expects to pay out on claims. Reserves also include the estimated future payments to adjust claims. During 2002, the Company experienced exceptional growth creating additional uncertainty in estimating ultimate loss costs. The contributing factors of this uncertainty were changes in the Company’s limit mix and mix of business, as well as increases in claims staffing and changes in claims processes. To address this risk, the Company’s actuarial area expanded its scope of reserve reviews by approximately 50%.

Before we go any further into our explanation of how Progressive establishes loss reserves, a few key terms/concepts must be understood:

Severity:Severity measures the dollar amount of claims during a given accident period divided by the total number of claims. The result indicates the dollar amount of the average claim.

Frequency:Frequency measures the number of claims divided by the number of exposure counts. This provides a measure of the proportion of insureds who have a claim in a year.

Pure Premium:Loss dollars divided by the number of exposures. Pure premium is also equal to frequency multiplied by severity. The pure premium is equivalent to the loss component of the full policy premium.

Trend:Trend is the rate of change in either frequency or severity or both. Understanding trend from changes in the underlying historical data, especially abrupt changes, is important.

Calendar Year:When losses are paid or reserve changes are made, under GAAP and SAP (Statutory Accounting Principles), we account for those payments or reserve changes in the calendar year in which the payment or reserve change took place, regardless of when the

Page 4

accident happened. Prior calendar year results are not restated as new claim information comes to light.

Accident Year:Actuaries estimate reserves on an accident year basis. Reserve estimates and all subsequent revisions are assigned to the accident period that the loss took place. A period may be defined as monthly, quarterly, semiannual, annual, etc. Reserves, regardless of when they are established or changed, are assigned to that same accident period. Accident period results will change over time as further information becomes available.

The Triangle:The triangle is a tool used by actuaries to show how estimates of incurred and paid-to-date amounts have changed or “developed” over time. Development can be defined as changes to paid and incurred losses over subsequent evaluation dates.

Reserve Development:

Generally, there are three ways for reserve development to occur. They are:

• open claims developing favorably or unfavorably due to changes in adjuster estimates or changes to average reserves set by the Actuaries

• unreported claims emerging at a rate greater or less than anticipated

• claims being settled for more or less than the established reserve for that claim.

Ultimate paid loss and loss adjustment expenses may deviate, perhaps substantially, from point-in-time estimates of reserves contained in the Company’s financial statements. The Company’s actual claim payments may exceed its claim reserves causing the Company to incur losses the Company did not anticipate. Uncertainty arises from a number of factors, including the difficulty in predicting the rate of inflation and the rate and direction of changes in trend, ongoing interpretation of insurance policy provisions by courts, inconsistent decisions in lawsuits regarding coverage and expanded theories of liability. In addition, ongoing changes in claims settlement practices can lead to changes in loss payment patterns that are used to estimate reserve levels. As shown in Exhibit 2, reserve development created from claims that happened in prior years can influence current year earnings. In other words, during times when we under-reserve, we report a higher income level. Conversely, when we over-reserve, we report a lower income level.

Exhibit 2

Contribution of Prior Years Development on EPS

Page 5

A few items are worth mentioning about the unpredictability in establishing reserve liability. First, regardless of how close our initial accident year estimates are, we will never get them exactly right and there will always be development. Second, during years when we experience unfavorable reserve development, once we get through it, we’re more experienced and better at estimating future reserves. Third, reserve development on claims that settle more slowly (e.g. liability claims) can be highly variant and extremely difficult to evaluate. Fourth, loss reserves can only be established for events that have already occurred. P&C companies cannot establish a reserve for a hurricane, hail, flood or any other catastrophic event that “may” occur later in the year. We mentioned this in last year’s report, but feel it is important to mention again. Now that Progressive has adopted monthly external reporting of operating results, analysts, shareholders and employees should expect the Company to experience substantial fluctuations in monthly results when catastrophic events happen. This is because we cannot “slowly build” reserve liabilities for anticipated future events.

This is not to suggest that we do not plan for catastrophic events to happen. At some point, they will occur. As a result, for capital planning purposes, we model catastrophic claims scenarios to determine capital requirements in the event a catastrophic event happens. This information is then shared with our Treasury group who manages our capital needs now and in the future.

External Reporting of Reserve Development:

For reporting purposes, we feel it is important to disclose reserve changes monthly and reserve development quarterly.

Since January 2003, we have been including the actuarial adjustments in the “Supplemental Information” section of the monthly earnings release. These adjustments are shown for the current month and year-to-date. For example, in the February 2003 earnings release, the monthly actuarial adjustment was $(3.3) million for the month, and $0.6 million year-to-date. This means that we increased our loss and LAE reserves during the month by $3.3 million as a result of regularly scheduled actuarial reviews. Each month, we generally review between 50 and 75 segments (as defined later in this report), which represent about 25 to 35 percent of our total reserves. We increase reserves for some segments and decrease reserves for other segments. The monthly change is the total for all reviews combined. A reserve increase is shown as a negative value on the earnings report because it is “unfavorable” to our earnings.

In addition, since the first quarter of 2002, we have been including “Actuarial Adjustments” to current and prior accident years in our “Additional Supplemental Information” section of the quarterly earnings release. We also report total prior accident years development (separated by “actuarial adjustments” and “all other development”). For example, in the first quarter 2003 earnings release, the year-to-date actuarial adjustments were $4.4 million (favorable to earnings). This means that we decreased our loss and LAE reserves during the first three months of 2003 by $4.4 million due to regularly scheduled actuarial reviews. The $4.4 million decrease consisted of an increase of $0.8 million for claims in the current accident year and a decrease of $5.2 million for claims in prior accident years.

The actuarial adjustment of $5.2 million reserve decrease that applies to claims in prior accident years is only part of the total development. The total prior accident years development as of March 31, 2003 was $39.0 million unfavorable. In other words, we now estimate that our reserves as of December 31, 2002 should have been $39.0 million higher than they were. As stated in the beginning of this section, favorable or unfavorable development is due to a combination of the following: changes to adjuster estimates and averages on open claims; actual and estimated emergence of claims that were unreported as of the prior year-end; and claims settling for more or less than the established reserves. In the first quarter of 2003, the unfavorable development has been largely attributed to unfavorable emergence of previously unreported claims (IBNR). This is due to more claims emerging than were anticipated, and/or higher severity than anticipated on these “late reported claims”.

Page 6

It is not inconsistent that we decreased reserves for prior accident years (per the actuarial reviews) at the same time that total development (for all causes all lines countrywide) was unfavorable. This situation can occur due to timing differences in the segments we review as well as relative changes and development by reserve component (loss case, loss IBNR, and LAE). Reserve changes made as a result of actuarial reviews are intended to keep our current reserve liability adequate. We change the reserves based upon our projections of expected future development.

Also note that the $39.0 million unfavorable prior accident years development as of March 31, 2003 is included in our current calendar year results. Our current calendar year incurred loss and LAE ratio of 66.7% is higher than our current accident year incurred loss and LAE ratio of 65.2%. The difference of 1.5 points reflects the $39.0 million unfavorable development through March 31, 2003.

Development Patterns:

Our experience has been that the longer a claim is expected to stay open, the more difficult it is to establish an accurate reserve at the time the accident is reported. Injury claims that are long-tailed (take longer to settle), require more data points and are more sensitive to changes in inflation than shorter-tailed (faster settling) property damage claims. Below, we have provided an updated exhibit using data as of 12/31/02 comparing the time it takes to settle a bodily injury liability claim vs. a shorter-tailed property damage liability claim.

Exhibit 3

Paid Development Patterns
Bodily Injury and Property Damage Claims

Page 7

Section III — Types of Reserves

Reserves are considered an operating liability on our balance sheet. At the end of 2002, we reported a $3.8 billion loss reserve liability on our balance sheet. We separate reserves into two categories: Losses and Loss Adjustment Expenses (LAE). While each of these two reserve categories are reported in the aggregate on the balance sheet, when we analyze loss reserves, we further break them into two distinct types of reserves — case and IBNR (Incurred But Not Recorded). In this section, we discuss these reserve types and how we evaluate them in order to get the total reserve balance as accurate as possible.

Exhibit 4 is a chart illustrating the types of reserves as a percent of our total reserve liability. As you can see, 82% of our reserve liability (Loss case + Loss IBNR) is set aside to pay claimants while 18% of our reserve liability (LAE case + LAE IBNR) is established to accommodate costs associated with adjusting those claims. These costs are described in more detail later in this section.

Exhibit 4

Reserve Distribution: Progressive Insurance (12/31/02)

Loss Reserves

1.     Case Reserves

Loss case reserves represented 64% of our total net carried reserves at 12/31/2002. Case reserves are used to pay claims that have already been reported by the claimant and recorded into Progressive’s systems, but have not yet been paid. We evaluate our total indicated case reserve need by record date or the date that the claim is recorded by the Company. We then establish:

• Average reserves for claims below a specified “threshold;”

• Inflation factor to apply to the average reserves each month;

• Adjuster-set reserves above the threshold;

• ANCR factors to allocate the additional needed case reserves above the threshold.

Page 8

For each open claim, the case reserve that is carried on the Company’s books is the average reserve if it is below the threshold. For claims above the threshold, the Company carries the adjuster set reserve (multiplied by the ANCR factor).

Average Reserves.For each open claim, the Company will identify and establish an average reserve regardless of the individual claim characteristics. The use of the average reserve allows claims personnel to concentrate their efforts on adjusting claims rather than merely accounting for them. For personal auto bodily injury claims (BI), all claims estimated to cost the Company less than $75,000 (our current threshold for personal auto bodily injury) are assigned an “average reserve.” Average reserves are determined by the Actuaries and are based upon analysis of claim data sorted by record date. Averages vary by “segment,” but are always determined by the actuaries. Once an average reserve is assigned to a claim, we monitor the “age” of a claim. The “age” of a claim is defined as the length of time from the accident date (i.e. the date the accident occurred) to the current accounting date. The more severe claims, such as bodily injury claims, tend to remain open longer than material damage claims. This creates a need for BI reserves to be periodically increased because of changes that may cause the claim to become more expensive at a future date due to medical treatments, litigation, etc. In other words, averages for bodily injury coverage increase as claims age.

Unlike bodily injury reserve averages, the averages for property damage, comprehensive and collision claims are not increased for age. These claims are easier to adjust and settle much more quickly, so we do not vary them by age.

There are several benefits to the use of actuarially determined average reserves. When a claim is first recorded by the Company, there may not be enough known about the claim for a claims representative to determine its severity. If it is an injury claim, research needs to be completed before a reasonable estimate can be made by the adjuster. If it is a physical damage claim, our claims representative can usually provide accurate on the spot estimates so the claim can be settled quickly.

Inflation.Average reserves for each segment are adjusted per the actuarial analysis in the month that the segment is reviewed. Since we do not review every segment every month, we apply an inflation factor to our average reserves in all other months. The inflation factor is based upon our projected severity trend from the most recent actuarial analysis for the segment under review.

Threshold.Our objective is to use the average reserve for claims when severity is at a reasonably stable, predictable level. However, the amount we pay for claims above a certain dollar threshold (which varies by coverage; for personal auto bodily injury the threshold is $75,000) can vary significantly depending on individual claim characteristics. For claims above the threshold we depend on the Adjuster-Set reserve rather than the Average Reserve.

Adjuster-Set Reserves. For claims in excess of the threshold, we have concluded that the adjuster-set amount provides a more accurate estimate of our ultimate liability because the claims representative has typically spent a great deal of time on these larger claims, and understands their unique claim characteristics. While only 2% of our total open claim count for personal auto bodily injury is above the established threshold of $75,000, these same claims represent about 20% of our total personal auto BI case reserve liability.

ANCR (Additional Needed Case Reserves).We also carry additional needed case reserves (ANCR) to cover expected future development of claims above the threshold. ANCR is a bulk reserve, and is mechanically allocated to lower levels by using ANCR factors. An ANCR factor is applied to bodily injury and uninsured/underinsured motorist bodily injury reserves in excess of the threshold. ANCR factors vary by segment and by age. ANCR factors decrease as the claims age, since more information is known about claims as they age, and the need for this bulk

Page 9

reserve diminishes. We determine the ANCR factors during our segment review process by analyzing the development of claims over the threshold.

Exhibits 5 and 6 illustrate the life of a large auto bodily injury claim. When the claim was originally reported, we assigned the actuarially determined “average reserve.” As the claim aged from the time it was reported in February through the end of September, the average reserve changed for the following reasons: 1. inflation; 2. actuarial reserve reviews; and 3. aging. Over this same period of time, the claims representative (red line) increased the reserve estimate multiple times as she learned more about the claim. Once the claims representative increased the reserve to just below the threshold of $75,000 (light blue line), the ANCR factor was applied. The resulting reserve amount exceeded the threshold of $75,000, so the booked reserve changed from an average reserve to an adjuster-set reserve. Note that the ANCR factor is mechanically applied to this claim in order to allocate the additional needed case reserves. It is not intended to improve the accuracy of individual claims.

Exhibit 5

Example of Case Reserving Over the
Life of a Large Bodily Injury Claim
Private Passenger Automobile
Policy Limit = 300,000
Threshold = 75,000
State XYZ
Inflation Factor = 6% per year
(Excludes Loss Adjustment Expense)

								Case
						Adjuster		Reserve
				Age of		Estimate of		Carried on
	Transaction			Claim		Claim		Company		Amount
Month-End	Date		Claim Activity	(months)*		Amount		Books		Paid		Explanation of Reserve Change
		1/5/01	Accident Occurs		1		N/A				0
Jan-01					1		N/A	IBNR			0	Aggregate amount based on factor of EP for Segment
		2/12/01	Claim is Reported		2		N/A				0	Claim is in "Pipeline"; Still IBNR
		2/15/01	Claim is Recorded		2		N/A				0	No estimate yet made by adjuster
Feb-01					2		N/A		5,829		0	Actuarially determined Average Reserve for 1-2 mo. Age group
Mar-01					3		N/A		7,121		0	Aging to 3-4 mo. Age group & Inflation
Apr-01					4		N/A		7,157		0	Inflation
		5/20/01	Adjuster sets up reserve		5		30,000				0	This is below the Effective Threshold**
May-01					5		30,000		8,391		0	Actuarial review & Aging to 5-6 mo. Age group
Jun-01					6		30,000		8,433		0	Inflation
July-01					7		30,000		9,789		0	Aging to 7-12 mo. Age group & Inflation
		8/10/01	Adjuster revises estimate		8		50,000				0	This is below the Effective Threshold**
Aug-01					8		50,000		10,250		0	Actuarial review revised Averages
Sep-01					9		50,000		10,301		0	Inflation
		10/25/01	Adjuster revises estimate		10		70,000				0	ANCR Factor for 7-12 mo. Age group = 1.25
Oct-01					10		70,000		87,500		0	(70,000) X (1.25) is greater than 75,000
Nov-01					11		70,000		87,500		0
Dec-01					12		70,000		87,500		0
Jan-02					13		70,000		80,500		0	ANCR Factor for 13-24 mo. Age group = 1.15
Feb-02					14		70,000		80,500		0	(70,000) X (1.15) is still greater than 75,000
		3/10/02	Adjuster revises estimate		15		80,000				0
Mar-02					15		80,000		92,000		0	(80,000) X (1.15)
Apr-02					16		80,000		92,000		0
May-02					17		80,000		88,000		0	Actuarial review revised ANCR Factor to 1.10
Jun-02					18		80,000		88,000		0	(80,000) X (1.10)
		7/15/02	Claim is Paid		19		90,000				90,000
Jul-02					19		N/A		0		90,000	Claim is Closed

* Age = [(# days since accident) / (30 days)] rounded up to the next whole number

** Effective threshold is (75,000) / (ANCR factor for age of claim)

Page 10

Exhibit 6

Example of Case Reserving Over the
Life of a Large Auto Bodily Injury Claim
Accident Occurred in January, 2001
Recorded in February, 2001
Claim Settled in July, 2002 for $90,000

2.     Incurred But Not Recorded (IBNR) Reserves

We establish a reserve for claims we are liable to pay but do not know about because they have not been reported by the claimants or recorded by the Company. IBNR Reserves are estimates of amounts needed to pay for these claims. At year end 2002, the loss IBNR reserves were 18% of our total net carried reserves.

We evaluate our IBNR reserve need by the same segmentation process we use for case reserves. We perform this analysis by sorting historical claims according to the time “lag” between the accident dates and the dates that these claims were recorded by our Company. If IBNR is reserved perfectly, then the dollar value of late reported claims will match the estimate made as of the balance sheet date.

Late reported claims are evaluated to determine the estimated ultimate losses for each accident quarter within each “lag” period. For example, “lag 1” consists of claims for which the accidents occurred during one quarter, but were not recorded until the next quarter. Similarly, “lag 2” consists of all claims for which the accidents occurred during one quarter, but they were recorded by the Company 2 quarters later. Since we need to determine the current reserve liability for claims that will be reported in the future, the estimated ultimate losses from these historical periods need to be trended forward. We use a trend based upon our estimates of both the frequency and severity of claims. The trended ultimate losses are divided by the earned car years (number of exposures) for those same respective periods, to determine the projected losses per exposure by lag period. Exhibit 7 below shows the percent of recorded features by record quarter lag for auto bodily injury as of 12/31/02. This exhibit shows that more than 80% of our claims are reported by the claimant and recorded in our systems by the end of the accident quarter (up to three months after the date of the accident).

Page 11

Exhibit 7

Countrywide Auto Bodily Injury
RECORDED FEATURE COUNT

Lag Quarters*	Incremental %			Cumulative %
Lag 0		84.1	%		84.1	%
Lag 1		11.7	%		95.8	%
Lags 2-3		2.2	%		98.1	%
Lags 4-6		1.1	%		99.2	%
Lags 7-9		0.5	%		99.7	%
Lags 10+		0.3	%		100.0	%

     * Record Quarter = Accident Quarter Plus Lag

IBNR reserves are to pay for claims that occurred in prior periods but have not yet been recorded. For example, for the accident quarter ending December 31, 2002, there will be claims reported one quarter later (for which we apply the trended lag 1 projection), two quarters later (for which we apply the trended lag 2 projection), and so on. We know how many exposures (earned car years) we had in the quarter ending December 31, 2002. Then we multiply this exposure count by the projected losses per exposure (pure premium) for the aggregate of all lag periods to derive the projected required IBNR reserves for the accident quarter ending December 31, 2002. We then convert that amount to a factor of earned premium, by dividing it by the premium earned during the quarter ending December 31, 2002. We go through this calculation for each accident quarter for approximately the past seven years (although for lag periods older than 7 quarters, we use annual projections). Refer to the case study in Section VI which shows a calculation of IBNR.

Our IBNR reserves are recorded at the end of each month (by segment) by applying those resulting IBNR factors to each 3-month-ending period of earned premium for the past 4 years. The largest IBNR factors (as derived from this analysis) are in the most recent accident periods. This makes sense, because the greatest IBNR reserve need is for the most recent accident periods. Also, as our premium volume increases, our IBNR reserves also increase. This allows our IBNR reserves to keep up with growth and inflation, as well as our increased exposure to claims.

During the first five months of 2003, the Company experienced adverse IBNR emergence. Consequently, we are assessing whether we would benefit by modifying the current methodology we use to determine required IBNR reserves. We are also reviewing the drivers of the unfavorable emergence, whether it is frequency or severity driven or if process changes in our claims organization have caused us to experience the unfavorable result.

Total Loss Reservesare the sum of case plus IBNR reserves. We also evaluate our total indicated loss reserve need by sorting and analyzing claims by accident date. This analysis is done by the same segmentations (and at the same time) as the evaluations of case and IBNR reserves, as discussed above. It is also discussed in more detail in the Case Study in Section VI. Since the total carried loss reserves are the sum of the carried case and IBNR reserves, it is reasonable to expect the total indicated reserve need to approximate the sum of the indicated

Page 12

case reserve need and the indicated IBNR reserve need. We compare these analyses for consistency and reasonableness in order to determine the most appropriate reserve changes.

Loss Adjustment Expense (LAE) Reserves

In addition to loss payments (which indemnify the claimants), the Company incurs expenses in the process of settling claims. Therefore, we need to estimate a reserve liability for “loss adjustment expenses.” At year-end 2002, the LAE Reserves were 18% of our total net carried reserves. There are two categories of loss adjustment expenses: “Defense and Cost Containment” (DCC), and “Adjusting.”

1.     “Defense and Cost Containment” (DCC) includes all defense and litigation-related expenses, as well as medical cost containment expenses. This category is comparable to (but not exactly the same as) what was called Allocated Loss Adjustment Expenses (ALAE) prior to the definition change by the NAIC (National Association of Insurance Commissioners) in 1998.

For DCC expense reserves, we evaluate our total indicated reserve need by sorting and analyzing these expenses by accident date. Through this analysis, we allocate the appropriate levels for our DCC case and IBNR reserves.

Most of the same mechanisms that are used to allocate loss case reserves are also used to allocate DCC case reserves. We determine average reserves by age within each segment. The older aged claims (based on the time between accident date and accounting date) carry higher averages. The averages are inflated each month (based upon a selected inflation factor). When the claims representative estimates the reserve to be over a certain threshold, we carry the adjuster-set reserve rather than the actuarially determined average reserve. Carried DCC IBNR reserves are calculated as a percentage of the carried loss IBNR reserves for each respective segment.

2.     “Adjusting”includes all claims adjusting expenses. This category is comparable to (but not exactly the same as) what was called Unallocated Loss Adjustment Expenses (ULAE) prior to the definition change by the NAIC in 1998.

For adjusting expense reserves, we evaluate our total indicated reserve need by comparing “adjusting expense” payments with loss payments over the past several calendar periods. The selected ratios are applied to the loss (Case and IBNR) reserves, and then modified to derive indicated “adjusting expense” reserves (split by case and IBNR). We derive our adjusting case reserves by applying the selected average adjusting expense reserve to each open claim. We derive our adjusting IBNR reserves by applying the selected IBNR factor to the carried loss IBNR reserves.

We carry both case and IBNR reserves for both categories of loss adjustment expenses. As stated above, we evaluate the LAE reserves by segment. Most of the LAE segments are reviewed two times per year. The evaluations of DCC expense reserves and adjusting expense reserves are performed independently of each other.

Salvage and Subrogation

Generally Accepted Accounting Principles (GAAP) and Statutory Accounting Principles (SAP) require that loss reserves be stated net of a deduction for estimated amounts that we expect to recover from salvage and subrogation. Therefore, anticipated salvage and subrogation reduce the loss reserve liability on the balance sheet.

Salvage and subrogation are similar, in that they are both recoveries which reduce the net claim amount. However, they are different types of transactions.

Page 13

Salvage:Progressive may take title to a vehicle when it has been declared a total loss. We then sell the vehicle remnants through a salvage dealer and the proceeds are referred to as salvage. Work is handled by either the branch claims offices or a specialized salvage recovery unit. Salvage is most relevant for collision claims.

Subrogation:When a Progressive insured is involved in an accident in which the other party is at fault, our insured may submit the claim to us. When we pay that claim, we obtain the insured’s right to recover damages from the at-fault party (usually the at-fault party’s insurance company). Work is handled by a specialized subrogation unit. Subrogation is most relevant for collision (for damage to vehicles) and personal injury protection claims (for injuries in no-fault states).

We do not treat salvage and subrogation as an independent type of reserve. As we collect salvage or subrogation from third parties, it reduces our net paid and incurred loss amount for that claim. When we evaluate our claim data to determine needed reserves, the evaluation is completed “net” of these recoveries. In other words, our estimated ultimate loss amounts are net of anticipated salvage and subrogation. Since claims are generally closed, and then reopened when the recoveries occur, the vast majority of our anticipated salvage and subrogation recoveries are included in the IBNR reserves.

Involuntary Market Operating Loss Reserves (Assigned Risk Reserves)

Progressive is required to participate in involuntary market plans. Below we discuss two examples of involuntary market plans.

1.     Private Passenger Assigned Risk Plan: State insurance regulations require us to participate in various assigned risk plans. Generally, assigned risk plans provide insurance to people who cannot find a carrier that will insure them voluntarily. We are obligated to accept assigned risk business based upon written premium volume and generally expect to incur an operating loss on these assignments in most states. This means that we expect our total losses and expenses to exceed the amount of premium we receive for these risks. Participation requirements in assigned risk plans differ from state to state. However, generally each insurance carrier is assigned a certain number of “state plan risks” based upon the company’s market share in that state and the type of coverage offered.

The reserves that we carry for the assigned risk plans comprised less that one percent of total net carried reserves at year-end 2002. However, since this is a unique type of exposure, we evaluate it separately, and segment it into several state/line coverage groups.

The process of determining the assigned risk reserve for a state is as follows:

• Determine our portion of the assigned risk pool by multiplying our projected market share by the estimated future size of the assigned risk pool in that state

• Reduce this by any state credits we receive for voluntarily writing policies that would have otherwise been considered assigned risk-type business

• The Assigned Risk Reserve = The estimated operating loss that we expect to incur from this business.

Page 14

2.     Commercial Auto Insurance Procedure (CAIP).Progressive is also required to share in the operating results of the involuntary commercial auto insurance plans, known as “Commercial Auto Insurance Procedure.” Due to the more complex nature of commercial business, these plans do not assign policies. Instead, they operate as a “Joint Underwriting Association” (JUA), and use a small number of servicing carriers who do not bear underwriting risk to service the business. Progressive is a servicing carrier in 26 states. The servicing carriers transfer the insurance risk, or “cede” 100 percent of the business, to the state CAIP pools. These state plans then retrocede, or further transfer, the operating results to all companies writing commercial automobile business in proportion to their share of the voluntary market for each respective state. If the plan generates an operating loss, we are assessed for our share of the loss. If the plan generates a profit, we benefit from our share of the profit.

Page 15

Section IV: Estimating Loss Reserves

For loss reserve review purposes, we divide our book of business into smaller pieces of data known as “segments.” During a segment review we usually apply six different estimation methods against the data in order to determine our ultimate required reserve balance. We discuss segmentation and provide more detail about the estimation methods below.

Segmentation of Reserves for Analysis

Segments are identified to allow us to review reserve needs at the finest level we can and provide us with the ability to identify and measure variances and trends in severity and frequency. They also allow process changes within states/regions to be reflected in the reviews. Each segment is required to have enough data to deliver reliable (credible) results. Our objective is to achieve adequacy in the reserve levels for each segment. This enhances the accuracy of our financial reporting, supports the income statements of our business units, and allows us to make better business decisions.

Most segments are defined by using a specific state/product/coverage grouping. Many states are large enough that we can look at the data by state while other states are so large that we have the ability to segment the state into regions within the state. There are also instances where the underlying data for the entire state is not large enough to provide credibility, so we combine states with similar frequency and severity loss characteristics and review them together. Below is a map showing how we segment our loss reserve reviews for personal auto bodily injury (Exhibit 8). As we continue to grow, we will continue to further segment our reviews at both the state and regional levels. In other words, the map below will probably look different by the time we publish this report because we continue to make changes and segment our business further for review purposes.

Exhibit 8: 2003 BI Loss Reserve Segments

Page 16

It is important to note that, with respect to bodily injury coverage, we analyze upper limits (personal auto policies with per-person limits greater than or equal to $100,000) separately from lower limits (less than $100,000). We do this because the data delivers different severity and frequency results by limit as well as different loss development patterns. Some products are even further segmented. For example, our Commercial Auto analysis separates specialty truck types from light local truck types. Regardless of how we group the data, each identified segment is reviewed annually, semiannually or quarterly depending on the size of the segment, the volatility within the segment, and other unique aspects of the individual segment.

Loss adjustment expense (LAE) reserves are analyzed using many of the same segmentation techniques as loss reserves; however, more states are aggregated when we review LAE. This is because the volume of LAE reserves is much less than that of loss reserves and grouping states together delivers more credibility in our results. All LAE segments are evaluated at least two times per year.

Estimation Methods Defined:

Our standard procedures are to review the results of six different estimation methods prior to determining if a reserve change is required. Three of the six methods use paid data and the other three methods use incurred data (payments plus changes in reserves). There are strengths and weaknesses to both data sets.

The three paid methods listed below (amount paid, average paid and Bornhuetter-Ferguson paid) all use paid loss data. The paid methods estimate growth and development of claims in an accident year by looking at the development of earlier accident years. This method assumes that past paid loss development is a predictor of future paid loss development. The primary strength of this method is that it removes the potential for distortions that may be created by including estimated data. The drawback to the paid method is that it does not always accurately project ultimate losses in the most recent periods under review since it also depends heavily on consistent claims closure or settlement practices. If these practices change, the results of the paid methods could be distorted. In addition, shifts in mix of business are not readily identified in paid development. This is especially important when the limit profile is changing.

The three incurred methods use paid losses plus loss reserves in each accident year and assume historical incurred loss development will be predictive of our future incurred loss development. The primary strength of incurred methods is that we can make use of our loss estimates for reported claims. This is especially critical when estimating the longer-tail “BI” type claims. The drawback to the incurred methods is that they depend heavily on consistent adjuster reserve estimates. The results of the incurred methods could be distorted if the adjuster reserve estimates are less or more accurate than in the past.

Interestingly, we saw changes in loss development patterns in both the paid and incurred methods during the past couple years. We partially attribute this to the high growth we experienced, increases in the average limits carried by our policyholders, and changes to our claims adjusting process. We also hired a significant number of new claims representatives countrywide. We saw changes in claims closure settlement practices (which affected our “paid” method results) as well as changes in the average level of adjuster set reserves (which affected our “incurred” method results). We identify these changes through our segmentation and our discussions with management. We adjust for these changes in our projections of losses. The case study in section VI includes more thorough explanations of how changes in closure rate affect paid loss development, and how changes in average adjuster set reserves affect incurred loss development.

Page 17

The six standard indication methods used are:

1.     Amount Paid: We take the total loss dollars paid by accident period and age and project them to an estimated ultimate level. We review total loss dollars only and base our future loss development projections largely on the development of prior periods.

2.     Average Paid: Paid severity by accident period is projected to an ultimate level, then multiplied by the estimated ultimate number of feature counts to be paid. Feature is defined in the terms and definitions section at the end of this document.

3.     Amount Incurred: We take the total dollars incurred by accident period and age and project them to an ultimate level. We review total dollars incurred only and base our future loss development projections largely on the development of prior periods.

4.     Average Incurred: Incurred severity by accident period is projected to an ultimate level, then multiplied by the estimated ultimate number of features to be paid

5.     Bornhuetter-Ferguson Paid:This method uses the paid loss development pattern. We use the development of paid losses to determine the percent unpaid. Then we apply the percent unpaid to the expected ultimate loss amount to arrive at the expected unpaid amount. The unpaid amount is added to actual losses paid-to-date.

6.     Bornhuetter-Ferguson Incurred:This method uses the incurred loss development pattern. We use the development of incurred losses to determine the percent unreported. Then we apply the percent unreported to the expected ultimate losses to arrive at the expected unreported amount. The unreported amount is added to actual losses incurred to date.

Page 18

Section V – Process Enhancements Introduced in 2002

During the past year, we implemented several enhancements into our loss and loss adjustment expense reserving process. These enhancements, summarized below, are described in more detail following this summary:

• Further Segmentation

• Upper and Lower Limits Auto BI

• Upper and Lower Limits Commercial Vehicles BI

• Upper and Lower Limits Auto and Motorcycle UMBI

• Upper and Lower Limits Commercial Vehicles UMBI

• Additional Commercial Vehicles BI Segmentation by State Grouping

• Specialty Commercial Vehicles

• Light Local Commercial Vehicles

• States and Regions

• Loss Adjustment Expenses – Auto BI/UMBI

• Commercial Auto Insurance Plan (CAIP)

• Analysis Process

• Data Alignment

• Increased Reserve Thresholds

• Auto UMBI

• Motorcycle BI and UMBI

• BI/PD IBNR Allocation

• Capped vs. Excess Losses for CV BI Segments

• Modeling

• Prior Selected Loss Development Factors (LDF’s)

• Reserve adequacy using alternative LDF’s

• Allocation to Accident Years

• Bornhuetter-Ferguson Methodology

• Commercial Vehicles Model

• LAE Model

• Communication

• Segment Review Meetings

• Summary Exhibits

• Decomposition (Decomp) Reports

Further Segmentation- We evaluate and set reserves by “segment.” Each segment should be large enough in volume that the data is reliable. It should also contain data that is homogeneous. The segment analysis will have more “credibility” as each of these conditions is satisfied to a greater degree.

We can estimate our reserve needs more accurately if we define our segments finely. As Progressive continues to grow, the additional growth provides the opportunity to further refine our analysis by segregating the data into more and more credible groupings. We believe that further segmentation provides us with the ability to predict reserve needs more accurately leading to better pricing decisions.

Page 19

In last year’s report, we stated that we divided our loss data into approximately 250 review segments. The number increased to approximately 300 segments by the end of 2002. Since we review each segment multiple times during the year, we completed about 775 segment reviews during 2002.

•	Upper and Lower Limits Auto BI — For personal auto bodily injury, we re-defined our two sets of average reserves for each state, based upon policy limits. We split our BI data so that the lower limits segments include claims for policies with per-person limits less than $100,000 per claim, while the upper limits segments include claims for policies with per-person limits greater than or equal to $100,000 per claim. In any particular state, the average severity for claims on policies with lower limits would be less than the severity on policies with higher limits. By analyzing the loss data for lower limits policies separately from upper limits policies, we can also see how losses develop differently. This allows us to make judgments regarding the frequency and severity of these segments separately, and set our reserves accordingly. This difference is clearer (and more reliable) in states where we have a larger volume of business.
	During 2001, we analyzed (and determined average reserves for) lower limits vs. upper limits in only a few of our largest states. We also defined the split at a lower limit ($25,000). During 2002, we increased the split to a new limit of $100,000 in every state. We did this because our limits profile has shown an increasing trend, and splitting the business at the new level gives us a more appropriate allocation of our reserves. At the time we implemented the change, we experienced an increase in reserves for upper limits business and decreased the reserves for lower limits business (without changing the overall reserve level). However, we expect future development on these claims to be minimized with this new allocation. In other words, the incurred loss amounts will better match claims when they are reported (and as they age), rather than developing at the time of settlement. Pricing decisions by limit and by tier will also be based upon more accurate loss information.
•	Upper and Lower Limits Commercial Vehicle BI — We separated our analysis for Commercial Vehicles (CV) Bodily Injury by upper limits vs. lower limits policies during the year also. However, the upper/lower limits split for CV is at the $500,000 BI/PD combined policy limit. This higher limit makes sense for CV since over 50 percent of our exposures are over the $500,000 limit. As with personal auto, this enhanced segmentation resulted in a shift of reserve dollars from lower limits policies to upper limits policies. The comments above regarding future development, better matching, and more accurate information for decision-making are also relevant to the CV limits split. In addition to the limits split, we also implemented several other segmentations for CV BI, described below.
•	Upper and Lower Limits Auto and Motorcycle UMBI — We are analyzing lower limits claims separately from upper limits claims for personal auto and motorcycle uninsured motorists bodily injury (UMBI). The split for UMBI is similar to BI, and it gives us a more appropriate allocation of our reserves. During 2001, we started implementing this enhancement in our larger states for personal auto. During 2002, we expanded it to the remainder of the states for auto, and also to motorcycles.
•	Upper and Lower Limits Commercial Vehicles UMBI — For commercial vehicles (CV) uninsured motorists bodily injury (UMBI), we separated our analysis (and average reserves) for upper limits vs. lower limits claims, split at the $100,000 per-person policy limit. As with personal auto, this enhanced segmentation has resulted in a shift of reserve dollars from lower limits policies to upper limits policies.
•	Additional Commercial Vehicle BI Segmentation – With our growth in commercial vehicles, it has become increasingly necessary to monitor the adequacy of our reserves. In addition to ensuring that our total reserve level is adequate, we must also ensure that the reserves are allocated appropriately in order to make better business decisions. In order to determine the

Page 20

most appropriate segmentation, we periodically analyze certain parameters of the CV BI data. During 2002, we determined that each of the following segments are credible for analysis (and for reserve allocation), since they are of sufficient volume, have similar severities, and have similar loss development patterns:

• Specialty Commercial Vehicles, which generally have a gross vehicle weight of at least 26,000 pounds. These include: tow trucks; local cartage (e.g. delivery vans and box trucks); and specialty trucks (e.g. dump trucks and flatbeds). These vehicles are separated into the following 5 segments, as defined by state and limit groupings:

• California (combined limits)

• Florida (lower limits vs. upper limits);

• All Other States (lower limits vs. upper limits).

• Light Local Commercial Vehicles, which generally have a gross vehicle weight under 26,000 pounds. These vehicles are separated into 10 segments, which include upper vs. lower limits segments for each of the following 5 groupings:

• California;

• Florida;

• Three state groupings for the remaining states, based on the frequency and severity of losses for each state

	We analyze each of the CV BI segments every three months, on a rotating basis. We continue to study ways to refine segmentation for Commercial Vehicles.
	Note that we also analyzed our CV segmentation for property damage coverage, but did not find significant enough differences between parameters of different groupings to justify creation of additional segments. Therefore, CV PD is analyzed countrywide, for limits combined. However, as with BI, we will continue to monitor this situation, and will refine the segments if we determine that it will improve our analysis and allocation of reserves.
•	States and Regions – We regularly review the reasonableness of our segmentations by state, and regions within each state. Some states are combined for certain coverages, if they are not credible for individual analysis, or if they are so similar that separation does not make sense. Some states have regions that are analyzed separately within the state, if they have enough credibility, and they are different enough to warrant it. During 2002, we added some segments as a result of this review process.
•	Loss Adjustment Expenses – Auto BI/UMBI – Previously, bodily injury liability coverage and uninsured motorists BI coverage had been combined in our analysis of loss adjustment expense reserves. In order to better allocate our LAE reserves, we now analyze BI separately from UMBI. In addition, for some segments, we analyze UMBI lower limits separately from UMBI upper limits.
•	Commercial Auto Insurance Plan (CAIP) – We enhanced our analysis of CAIP Loss and LAE reserves by realigning our segments.

Analysis Process — Data alignment– In addition to refining our segmentation of the data, we have also refined the alignment of the data or the way data is organized. Data alignment changes do not add to the number of segments that we review, rather they allow us to analyze the data (and allocate the reserves) from a different perspective, and to allocate the reserves more accurately.

Page 21

•	Increased Reserve Thresholds — When claim representatives set the reserve level above the threshold, the case reserve amount recorded on the company’s books is equal to the adjuster estimate, multiplied by an “ANCR Factor.” However, when the adjuster estimate is below the threshold, or if the adjuster has not yet made an estimate, the carried case reserve amount is equal to the average, as determined by our actuarial analyses.
	The dollar amount of the threshold determines how much of the carried case reserves utilize the averages vs. the adjuster estimates. When the threshold is higher, a greater proportion of the carried case reserves use the averages. However, in the month in which we increase the threshold, the averages in the case reserve tables will also be increased. This is because they apply to a group of claims with higher severity. When threshold changes are implemented, the total amount of carried reserves is unchanged. This is because a threshold change simply reallocates the reserves above or below the threshold.
	Threshold changes that we implemented in 2002 were:

• Auto UMBI threshold increased from $25,000 to $75,000

• Motorcycle BI and UMBI threshold increased from $25,000 to $75,000

•	BI/PD IBNR Allocation — As explained in an earlier section, our IBNR reserves are calculated by multiplying our IBNR factors (as determined by the segment analyses) by three-month-ending periods of historical earned premium. In 2002, we realigned our data so that the IBNR reserves are allocated more accurately between BI and PD.
•	Capped vs. Excess Losses for CV BI Segments — Our average reserve threshold for CV BI is $125,000. This threshold was implemented in 2001. However, in our analysis of losses by segment, we normally include claims that are both under and over the threshold. As explained above, we enhanced our analysis by separating the segments into upper and lower policy limits (split at $500,000). However, we further enhanced our analysis of each CV BI segment by capping each claim at $50,000. We are then able to analyze the development separately: for claims capped at $50,000, and for the excess portion of those larger claims (over $50,000).
	By definition, selected ultimate severity by accident period equals selected ultimate losses divided by selected ultimate claim counts. Judgment is involved in the selection of all three of these parameters. But the final selections must be reasonable. When the CV BI claims were analyzed in total, the frequency and severity of the larger losses distorted the data, which made reasonable selections more difficult. By separating the claims into capped and excess layers, we are better able to make reasonable selections. Final selections of ultimate losses for CV BI segments are the sum of the capped and excess loss selections.
	The development of the capped claims is more stable as are severity and loss ratio. So, the judgments made to select capped losses are based upon fairly standard methodologies. However, the excess portion of claims is more volatile, so these selections require a higher degree of judgment. We look at the proportion of claims that have exceeded the cap historically, and use that information in our subjective selection of excess counts for more recent accident periods. The excess severity is also the result of adding subjective insights to our analysis, by looking at the trends of the historical excess losses, knowledge of specific claims, and reasonableness of resulting parameters.
	There are also a few selected auto segments in which we separate the capped and excess losses for analysis. This is done for segments where there is a large variance in the severity of losses, the frequency and severity of large losses distorts the data, and the use of this methodology allows us to make better judgments.

Page 22

Analysis Process – Modeling– Our analysis model has been developed internally, and is continually being enhanced. These enhancements increase our efficiency, accuracy, allocation, flexibility, information, and/or communication. Some of the more significant enhancements incorporated into the model during 2002 have been:

• Prior Selected LDFs – Selections of paid and incurred loss development factors involve judgment. We need to determine future loss development patterns based upon historical development patterns. As part of that analytical process, it is helpful to know what we selected in prior analyses. We can now view the selected factors from the prior two analyses, in addition to the actual development. We can then compare how the losses have developed in relation to these prior projections. Our new selected development factors take that into consideration.

• Reserve adequacy using alternative LDFs – For each of the paid and incurred development methods, we select loss development factors. However, there are certain “default” selections that we built into the model to assist in our judgments. We calculate what the reserve adequacy would be if we used the loss development factors that result from use of the following defaults:

• Last Diagonal – to determine what the reserve adequacy would be if losses for each accident period (or record period) would develop over the next six calendar months in exactly the same way as they developed over the prior six calendar months;

• Second-Last Diagonal – to determine what the reserve adequacy would be if losses would develop over the next six months in exactly the same way as they developed over the six-month period that ended 6 months ago;

• Average of the Last 4 – to determine what the reserve adequacy would be if losses would develop over the next six months in exactly the same way as they developed over the average of the last four six-month-ending periods.

	Our selected loss development factors typically do not equal any of these default selections. However, they help us understand and explain our selections, in relation to how the losses have developed in recent history. In some cases, they give us a range of reasonableness. In other cases, they may help us to illustrate and explain why recent history is not indicative of future development.
•	Allocation to Accident Years – When we report earnings, it is important that we allocate reserves to the proper accident years. The most important separation is between the current accident year vs. prior accident years. This is required for Statutory and GAAP reporting. It is also important for management decision-making. Calendar year incurred losses include changes in reserves made during the current calendar year. However, these reserve changes may be for current year accidents or for prior year accidents.
	For each segment we review, we may decide to implement changes to both case and IBNR reserves. We built an enhancement into our analysis model so that we can specifically state how much of each of our reserve changes should be allocated to the current accident year vs. the prior accident years.
•	Bornhuetter-Ferguson Methodology – The Bornhuetter-Ferguson (BF) Methodology is a standard actuarial projection method, in which the projected ultimate losses combine the reported development with an expected loss ratio (or pure premium). We built some enhancements into our loss analysis model that make the BF method more relevant to the selections. We explain this in more detail in the case study in a later section.
•	Commercial Vehicles Model – The CV model is based upon the auto model, but there are several differences. The most significant difference is the addition of the capped and excess

Page 23

	projections, as explained above. However, as the CV analysis has evolved, other enhancements have been made to the CV model to accommodate the differences in this analysis.
•	LAE Model – Enhancements to this model include the addition of the “utilization ratio” parameter. Utilization ratio represents the proportion of claims for which we have incurred “Defense and Cost Containment” (ALAE) expenses. This is used to test the reasonableness of our selections.

Communication– Communication with the business units is a very important part of the loss reserving process.

•	Segment Review Meetings – We increased the number of meetings with the business units to discuss our analyses. We invite managers and analysts from product management, pricing and claims to these meetings. We share information regarding our reserve analysis: the effect of the reserve changes, and why selections were made. We solicit information from the business and claims units regarding process changes and trends that may be affecting the data, historically and/or in the future. We expect that this interaction will lead to better business decisions.
•	Summary Exhibits – We have a standard package of exhibits that summarize our segment analysis. These exhibits are explained in detail in the case study section. These summary exhibits are shared with the business units after each analysis is completed. They include summaries of the indicated and selected losses, counts, and severities by accident and record period, using several different projection methods. They also include information on average adjuster case reserves, closure rates, CWP rates, and resulting loss ratios and pure premiums. Actual average paid and average incurred loss triangles and loss development factors (including defaulted and selected factors) are also included in the package.
	This summary package is used as the basis for discussions in the segment review meetings. It has been enhanced to include more relevant and concise information.
•	Decomposition (Decomp) Reports – The Decomp Report was explained in detail in last year’s report. During 2002, the report was enhanced to match the Business Unit definitions. It is now produced monthly by distribution channel (Agent and Direct), by region within state, and by major line (auto, commercial vehicle, motorcycle, boat, etc.). This matches the Profit By Line (PBL) reports produced by our financial reporting group and helps the business units decompose the financial reserve changes into the following categories:

• Features that closed

• Features that opened (including reopened features)

• Changes in reserve averages on new features (due to loss reserving)

• Changes in reserve averages on open features (due to loss reserving)

• Inflationary impact on open features (inflation factor applied to average reserves)

• Aging of open features (changing age groupings)

• Changes from adjuster-set to average reserve (reserve amount changes from above threshold to below threshold)

• Changes from average reserve to adjuster-set (reserve amount changes from below threshold to above threshold)

• Changes in adjuster-set reserves (reserve amount stays above threshold)

• Changes due to resegmentation of data

Page 24

Terms and Definitions:

Accident Period Losses:Losses for each accident are assigned to the period in which the accident occurred. Accident periods used in our analysis are generally three months (accident quarter), six months (accident semester), or twelve months (accident year). Payments and reserve changes, regardless of when they are made, are assigned to that same period in which the accident occurred. As a result, accident period results will change over time as the losses develop, i.e. as the estimates of losses change due to payments and reserve changes for all accidents that occurred during that period.

Adjusting Expense:A component of loss adjustment expense. Adjusting expenses include all claims adjusting expenses, whether internal or external to the company. This category includes fees and salaries of those involved in a claim adjusting function, and other related expenses incurred in determination of coverage.

Bodily Injury:A liability coverage of the insurance policy which pays for the costs associated with injury to a person (third party), resulting from the negligent or intentional acts or omissions of the insured. These costs may include medical expenses, lost wages and non-economic losses.

Calendar Period Losses:Payments and reserve changes for claims which are recorded on the company’s financial records during the period in which the activity takes place, without regard to the period in which the accident occurred. Calendar period results do not change after the end of the period, even as new claim information develops.

Case Reserves:Estimates of amounts required to settle claims that have already been reported by claimants but have not yet been fully paid. Case reserves represent the largest portion of the reserves for automobile products. Progressive has two major categories of case reserves:

1.	Average Reserves: These are determined by the actuaries and are generally limited to losses estimated to be below a specified threshold. Use of the average reserves allows claims personnel to concentrate their efforts on adjusting claims rather than merely accounting for them. Our financial reserves include these average reserves when the adjuster estimate of the loss is below the threshold.
2.	Adjuster Set Reserves: These are the claims adjusters’ best estimate of how much specific claims will cost. Adjuster set reserves are included in our financial reserves only when they are above a specified threshold.

Catastrophe:A term applied to an incident, storm or series of related incidents resulting in losses which total more than $25 million in property damage for the insurance industry and affect a significant number of property and casualty insurance policyholders and insurers.

Cede:To transfer a risk from an insurance company to a reinsurance company.

Claim:A demand for payment of a policy benefit because of the occurrence of an insured event, such as damage, destruction, injury or death.

Claimant:One who submits a claim for an incurred loss.

Closed Without Payment (CWP):A claim that was reported and was originally set up with a reserve, but it did not require a payment and is now closed.

Closure Rate:The number of claims closed with payment at a specific evaluation date, divided by the ultimate number of claims to be paid for that accident period.

Page 25

Collision:A coverage of the insurance policy which indemnifies the insured when their automobile is damaged due to physical contact with another object (e.g. another automobile).

Combined Ratio:The sum of the underwriting expense ratio and the loss and loss adjustment expense ratio. A combined ratio less than 100% indicates an underwriting profit; a combined ratio in excess of 100% indicates an underwriting loss.
GAAP Combined Ratio =
(Incurred Losses + Incurred loss adjustment expenses + GAAP underwriting expenses) divided by (Earned Premium)

Comprehensive:A coverage in the automobile insurance policy that indemnifies the insured in the event of physical damage (other than collision) or theft of the insured vehicle. This includes perils such as fire, vandalism, hail, etc.

Credibility:A statistical measure of the ability to infer generalizations from a sample. Credibility increases as sample size increases and variability within the sample decreases.

Defense and Cost Containment Expense (DCC): A component of Loss adjustment expense. DCC expenses include defense, litigation and medical cost containment whether internal or external to the company. DCC expenses include but are not limited to accident investigation, surveillance, litigation management, and fees of attorneys and others if working in defense of a claim.

Earned Car Year:Insurance coverage on one automobile for 12 months = one earned car year. This is the unit of “Exposure” in auto insurance.

Earned Premium:The portion of premium that has been exposed to loss. This is generally the proportion of written premium since the effective date of the policy.

Expense Ratio:Expenses associated with the cost of doing business excluding those expenses associated with losses, divided by the total earned premium.

Exposure:The basic rating unit underlying an insurance premium. The unit of exposure will vary based upon the characteristics of the insurance coverage involved. For automobile insurance, one automobile insured for a period of twelve months is a car year or one exposure.

Feature:The smallest divisible part of a claim. This is a loss on one coverage for one person. Often a claim will involve multiple features. It can involve multiple coverages, such as bodily injury (BI) and property damage (PD); or it can involve multiple claimants. However, each coverage for each person is one feature.

Frequency:The number of claim features divided by the number of exposures. This provides a measure of the proportion of insureds that have a claim in a year.

Incurred But Not Recorded (IBNR) Reserves:These are estimates, at a given evaluation date, of amounts that will be needed to settle claims that have already occurred but have not yet been reported by the claimant and recorded by the company.

Incurred Losses:The sum of payments and reserve changes for claims.

Loss Adjustment Expenses (LAE): Costs involved in an insurance company’s adjustment of losses. LAE = Defense and Cost Containment (DCC) Expenses + Adjusting Expenses.

Loss Ratio(Incurred Loss Ratio):
(Paid Losses + Changes in Loss Reserves) divided by (Earned Premium).

Page 26

Net Loss Reserves:“Net” indicates that we have deducted the expected reinsurance recoverable from the sum of case and IBNR reserves. Sometimes, net reserves refer to reserves that have been reduced for the expected salvage and subrogation recoveries.

Personal Injury Protection (PIP):Only in states with no-fault automobile insurance statutes, this coverage pays basic expenses for an insured that is injured in an automobile accident, regardless of fault. These first party benefits are paid by the insured’s insurance company. These covered expenses vary depending on the no-fault laws in the state. However, they generally include at least medical expenses.

Pure Premium:Loss dollars divided by the number of exposures. Pure premium is also equal to frequency times severity. The pure premium is equivalent to the loss component of the full policy premium.

Record Period Losses:Losses for each record period are assigned to the period in which the accident is recorded on the company’s books. Record periods used in our analysis are generally three months (record quarter), six months (record semester), or twelve months (record year). Payments and reserve changes, regardless of when they are made, are assigned to that same period in which the accident was recorded. As a result, record period results will change over time as the losses develop, i.e. as the estimates of losses change due to payments and reserve changes for all accidents that were recorded during that period.

Relativity:The magnitude by which a measure of a subset of data differs (expressed as a quotient) from the same measure for the entire set of data. For example, if the countrywide average paid BI claim was $4,500 and Florida’s average BI claim was $5,200, then Florida’s paid BI claim relativity would be 1.156.

Reopened Claim: A claim that was closed (with or without payment) but reopened again at a later date due to the discovery of additional information. We reserve for future reopened claims as IBNR.

Reserves:Estimates of our ultimate liability for losses and loss adjustment expenses for insured events that have occurred prior to a specific accounting date, but have not yet been paid by the Company. These estimates are influenced by many variables, such as trend/inflation on automobile costs, changes in claims practices, etc.

Salvage:This reserve is the amount of future recoveries we expect to collect for damaged property previously owned by the insured for which all legal rights have been assigned to the insurance company. Future salvage is carried as a bulk reserve and results in reduced IBNR. This is because it applies to features that have been closed.

Segment:A given combination of states, products, and coverages with reasonably similar loss characteristics grouped together when assessing reserve adequacy.

Severity:Loss dollars divided by number of claims. This indicates the dollar amount of the average claim.

Subrogation:An insurer may have the right to take legal action against a third party responsible for a loss to an insured for which a claim has been paid. Reserves for the future recoveries we expect to subrogate are included in our IBNR indications.

Threshold:The point above which the adjuster’s estimate of a claim is carried in our financial reserves, versus an average reserve being assigned by the system.

Trend (Exponential fit):Exponential fitted trends tell us the estimated average annual change in severity, frequency, or pure premium, by fitting an exponential curve to the selected values. These can use any number of data points. We generally use two-year or four-year fitted trends.

Page 27

Ultimate:The final selected amount, count, or ratio that we estimate by analyzing the data. For example, the selected ultimate loss amount for an accident period represents our estimate of the total cost of all claims for that accident period after they have all been paid and closed.

Page 28

Table of Contents

		��
Section VI: Loss Reserve Review
	Case Study	1
	Exhibits A – E	3
	Exhibit A – Summary of Accident Period Analysis	10
	Exhibit B – Accident Period Average Incurred Loss Development	23
	Exhibit C – Summary of Record Period Analysis	28
	Exhibit D – Summary of Estimated IBNR	31
	Exhibit E – IBNR Analysis	36
Section VII: Loss Adjustment Expense Reserve Review
	Case Study	40
	Exhibits DCC and ADJ	42
	Exhibit DCC – Defense and Cost Containment Expense Analysis	44
	Exhibit ADJ – Adjusting Expense Analysis	50

01-00102B(06-03) Copyright © Progressive Casualty Insurance Company. All Rights Reserved.

Section VI: A Case Study: Loss Reserve Review

The decisions we make regarding how to revise our reserves each month are based upon our segment reviews. Any or all of the following parameters may be revised as a result of our reviews.

• Case Reservescan be revised by changing:

• Average Reserves(applicable to open claims below theThresholdonly)

• ANCR Factors (applicable toAdjuster-Set Reserveson open claims above theThreshold)

• Inflation Factor(applicable to Average reserves in months following a review)

• IBNR Reservescan be revised by changing:

• IBNR Factors, applicable to trailing periods of earned premium

In this section, we present an example of a reserve review for losses for a sample segment. As discussed previously, most segments are defined as a state/product/coverage grouping. Loss reserve reviews do not include loss adjustment expenses (LAE).

Note that the data in this example is not from any specific segment and any similarity to a specific segment would be coincidental. Also, the selections made in this case study are not necessarily the same as those that would be made in an actual review. The results of this case study are also not intended to represent the actual results of the company.Our intent of this case study is to illustrate and discuss issues that we consider during an analysis. The calculations involved in the process will also be explained.

This case study will illustrate how we estimate the adequacy of our loss reserves by reviewing loss data organized in three different ways:

Type of Loss Reserve	Claims Data Organized by
Total (Case + IBNR) Case IBNR	Accident Period Record Period Record within Accident Period

By definition, the following identities are always true as of the designated evaluation date:

[Loss Reserve Adequacy] =
[Carried (or Held) Loss Reserves] –
[Required (or Indicated) Loss Reserves]

[Required (or Indicated) Loss Reserves] =
[Total Indicated (or Selected or Estimated) Ultimate Losses] –
[Total Paid Losses]

Carried reserves and paid losses are known statistics and reconcile with our financial records. There are no judgments involved in these amounts; they are known facts. Judgment is in the estimation of theUltimate Losses. As stated above, we make these estimations by accident

Page 1

period, record period, and record within accident period. Our job is to determine how losses will develop in the future using past development as a key indicator. In order to make reasonable selections, we look at several parameters, and also consider the business issues that underlie the data.

There are several exhibits that we include in our reviews to summarize our analysis. We also use these exhibits to discuss what we are seeing with management. The exhibits are presented on the next several pages. Following that, we provide an overview of each exhibit. Note that “rounding” in the exhibits, as well as the “order of calculation” may make some of the figures appear slightly out of balance.

Exhibit A– Summary ofAccident PeriodAnalysis

Exhibit B– Accident PeriodAverage Incurred LossDevelopment Triangle

Exhibit C– Summary ofRecord PeriodAnalysis

Exhibit D– Summary of EstimatedIBNR

Exhibit E(3 pages) –IBNRAnalysis Triangles

Page 2

Exhibit A

State XYZ Auto as of December 31, 2002

ACCIDENT PERIOD ANALYSIS

	(1)			(2)			(3)		(4)		(5)
Accident	Paid			Avg. Paid			BF Paid		Incurred		Avg. Incurred
Semesters	Method			Method			Method		Method		Method
Ending	Ult ($000)			Ult ($000)			Ult ($000)		Ult ($000)		Ult ($000)
PRIOR 3 yrs		35,427			35,385			35,427		35,395		35,347
Jun-1999		10,930			10,940			10,931		11,193		11,170
Dec-1999		13,257			13,163			13,254		13,249		13,273
Jun-2000		13,534			13,781			13,603		14,012		13,985
Dec-2000		9,962			9,868			9,954		10,324		10,304
Jun-2001		9,485			9,492			9,435		10,149		10,100
Dec-2001		7,187			6,928			8,001		8,181		8,129
Jun-2002		9,689			8,667			9,207		8,842		8,727
Dec-2002		11,020			12,069			10,285		9,665		9,673
Total		120,492			120,293			120,095		121,011		120,708
Paid Loss		93,602			93,602			93,602		93,602		93,602
Required Reserve		26,890			26,691			26,494		27,409		27,106
Held Reserve		28,038			28,038			28,038		28,038		28,038
Reserve Adequacy		1,148			1,347			1,545		629		932
Average Last 4		3,132			(2,025	)				3,261		3,835
Second Diagonal		2,865			(3,318	)				624		1,951
Last Diagonal		(7,001	)		(6,264	)				3,470		3,154

[Additional columns below]

[Continued from above table, first column(s) repeated]

	(6)		(7)		(8)			(9)
Accident	BF Incurred		Adj. inc. @		Pd. Loss @			Indicated
Semesters	Method		12/31/2002		12/31/2002			Ultimate
Ending	Ult ($000)		($000)		($000)			Loss ($000)
PRIOR 3 yrs		35,395		35,373		34,937			35,379
Jun-1999		11,193		11,111		10,434			11,186
Dec-1999		13,250		13,087		12,197			13,257
Jun-2000		14,141		13,738		11,955			14,046
Dec-2000		10,324		10,117		8,248			10,318
Jun-2001		10,127		9,888		7,014			10,125
Dec-2001		8,216		7,891		4,238			8,175
Jun-2002		8,846		8,529		3,221			8,805
Dec-2002		9,748		8,107		1,357			9,695
Total		121,241		117,840		93,602			120,987
Paid Loss		93,602							93,602
Required Reserve		27,639					Required Reserve		27,385
Held Reserve		28,038					Held Reserve		28,038
Reserve Adequacy		400					Reserve Adequacy		654
Average Last 4
Second Diagonal
Last Diagonal

	(10)			(11)			(12)		(13)			(14)			(15)
Accident							Avg. Adjuster
Semesters	Avg. Paid			Avg. Incurred			Case Reserves		Closure Rate			CWP Rate			Ultimate
Ending	Severity			Severity			@ 6 Months		@ 6 Months			@ 6 Months			CWP Rate
PRIOR 3 yrs		5,863			5,857
Jun-1999		5,796			5,918			4,207		33.7	%		26.3	%		37.9	%
Dec-1999		6,141			6,192			4,321		28.6	%		29.4	%		40.4	%
Jun-2000		6,342			6,436			5,341		26.1	%		27.6	%		41.3	%
Dec-2000		5,404			5,643			5,291		32.3	%		26.3	%		39.8	%
Jun-2001		6,280			6,682			5,462		30.8	%		30.7	%		41.8	%
Dec-2001		5,686			6,671			5,213		26.4	%		29.2	%		42.5	%
Jun-2002		7,449			7,501			4,606		23.5	%		32.4	%		47.2	%
Dec-2002		8,940			7,165			4,153		21.5	%		28.7	%		43.1	%
% Chg Dec-01 vs Jun-02		31.0	%		12.4	%
% Chg Jun-02 vs Dec-02		20.0	%		-4.5	%

[Additional columns below]

[Continued from above table, first column(s) repeated]

	(16)		(17)		(18)		(19)		(20)
Accident	Paid		Clsd.w/Pay		Incurred		Recorded		Indicated
Semesters	Counts		Counts		Counts		Counts		Ultimate
Ending	Method		Method		Method		Method		Counts
PRIOR 3 yrs		6,030		6,029		6,032		6,035		6,035
Jun-1999		1,885		1,886		1,888		1,887		1,888
Dec-1999		2,144		2,145		2,145		2,142		2,144
Jun-2000		2,132		2,134		2,175		2,171		2,173
Dec-2000		1,821		1,822		1,827		1,825		1,826
Jun-2001		1,483		1,486		1,514		1,509		1,512
Dec-2001		1,198		1,196		1,222		1,215		1,219
Jun-2002		1,080		1,074		1,179		1,148		1,164
Dec-2002		1,177		1,104		1,449		1,379		1,350
% Chg Dec-01 vs Jun-02		18,950		18,876		19,431		19,311		19,309
% Chg Jun-02 vs Dec-02

Accident	(21)			(22)			(23)			(24)			(25)
Semesters	Ultimate			Change in			Ultimate			Change in			Pure
Ending	Severity			Severity			Frequency			Frequency			Premium
PRIOR 3 yrs		5,862						3.22	%					189
Jun-1999		5,926						3.59	%					212
Dec-1999		6,185			4.4	%		3.41	%		-4.8	%		211
Jun-2000		6,464			4.5	%		3.46	%		1.5	%		224
Dec-2000		5,650			-12.6	%		3.24	%		-6.3	%		183
Jun-2001		6,699			18.6	%		2.97	%		-8.4	%		199
Dec-2001		6,709			0.2	%		2.56	%		-13.9	%		172
Jun-2002		7,568			12.8	%		2.60	%		1.6	%		197
Dec-2002		7,182			-5.1	%		2.59	%		-0.3	%		186
														196
4 point exp trend		6.8	%					-7.6	%
8 point exp trend		6.5	%					-10.5	%

[Additional columns below]

[Continued from above table, first column(s) repeated]

Accident	(26)			(27)		(28)		(29)			(30)
Semesters	Loss			Premium		Earned					Change In
Ending	Ratio			($000)		Exposures		Avg EP			Avg EP
PRIOR 3 yrs		49.1	%		72,054		187,526		384
Jun-1999		64.6	%		17,325		52,642		329
Dec-1999		70.7	%		18,744		62,827		298			-9.3	%
Jun-2000		79.5	%		17,670		62,734		282			-5.6	%
Dec-2000		65.9	%		15,652		56,287		278			-1.3	%
Jun-2001		68.6	%		14,749		50,881		290			4.2	%
Dec-2001		56.2	%		14,557		47,667		305			5.3	%
Jun-2002		61.9	%		14,233		44,804		318			4.0	%
Dec-2002		60.0	%		16,162		52,158		310			-2.5	%
		60.1	%		201,146		617,528
4 point exp trend									4.9	%
8 point exp trend									0.4	%

Page 3

Exhibit B

State XYZ Auto as of December 31, 2002

AVERAGE INCURRED LOSSES — ACCIDENT PERIOD ANALYSIS

Semiannual
Accident
Periods																																												Ultimate		Ultimate
Ending		1			2			3			4			5			6			7			8			9			10			11			12			13			14			Severity		Loss (000s)
	Jun-96		5,790			5,876			5,928			5,553			5,688			5,796			5,792			5,988			6,019			5,999			5,969			5,960			5,962			5,950			5,950		3,540
	Dec-96		5,365			5,961			5,385			5,730			5,636			5,514			5,782			5,928			5,884			5,970			5,939			5,981			5,981			5,969			5,969		4,262
	Jun-97		6,087			6,084			5,795			6,852			6,652			6,833			6,832			6,825			6,882			6,907			6,900			6,912			6,913			6,899			6,899		5,333
	Dec-97		5,031			5,470			5,558			5,623			5,774			5,974			6,084			6,102			6,139			6,230			6,160			6,170			6,171			6,159			6,159		7,151
	Jun-98		4,778			5,342			5,383			5,465			5,489			5,617			5,653			5,661			5,651			5,710			5,677			5,687			5,688			5,677			5,677		7,448
	Dec-98		4,153			4,765			4,971			4,988			5,030			4,974			5,078			5,124			5,118			5,174			5,145			5,154			5,154			5,144			5,144		7,613
	Jun-99		4,315			5,241			5,457			5,704			5,786			5,787			5,822			5,865			5,889			5,953			5,919			5,929			5,930			5,918			5,918		11,170
	Dec-99		4,830			5,839			5,985			5,975			6,088			6,058			6,068			6,136			6,161			6,228			6,193			6,204			6,205			6,192			6,192		13,273
	Jun-00		6,277			6,306			6,180			6,140			6,283			6,269			6,307			6,378			6,403			6,473			6,436			6,448			6,449			6,436			6,436		13,985
	Dec-00		5,440	��		5,411			5,274			5,440			5,456			5,497			5,530			5,592			5,614			5,675			5,643			5,653			5,654			5,643			5,643		10,304
	Jun-01		6,155			6,126			6,269			6,366			6,461			6,509			6,548			6,622			6,648			6,721			6,683			6,694			6,695			6,682			6,682		10,100
	Dec-01		5,657			5,850			6,189			6,356			6,451			6,499			6,538			6,611			6,638			6,710			6,672			6,684			6,685			6,671			6,671		8,129
	Jun-02		5,513			6,756			6,959			7,147			7,253			7,307			7,351			7,434			7,463			7,545			7,502			7,515			7,516			7,501			7,501		8,727
	Dec-02		5,289			6,453			6,647			6,826			6,928			6,979			7,021			7,100			7,129			7,206			7,166			7,178			7,179			7,165			7,165		9,673
			1-2			2-3			3-4			4-5			5-6			6-7			7-8			8-9			9-10			10-11			11-12			12-13			13-14
	Jun-96		1.015			1.009			0.937			1.024			1.019			0.999			1.034			1.005			0.997			0.995			0.998			1.000			0.998
	Dec-96		1.111			0.903			1.064			0.984			0.978			1.049			1.025			0.993			1.015			0.995			1.007			1.000
	Jun-97		1.000			0.953			1.182			0.971			1.027			1.000			0.999			1.008			1.004			0.999			1.002
	Dec-97		1.087			1.016			1.012			1.027			1.035			1.018			1.003			1.006			1.015			0.989
	Jun-98		1.118			1.008			1.015			1.004			1.023			1.006			1.001			0.998			1.010
	Dec-98		1.147			1.043			1.003			1.009			0.989			1.021			1.009			0.999
	Jun-99		1.215			1.041			1.045			1.014			1.000			1.006			1.007					Loss Development Factors									Adequacy
	Dec-99		1.209			1.025			0.998			1.019			0.995			1.002								Average Last 4										3,835
	Jun-00		1.005			0.980			0.993			1.023			0.998											Second Diagonal										1,951
	Dec-00		0.995			0.975			1.031			1.003														Last Diagonal										3,154
	Jun-01		0.995			1.023			1.016																	Selected Avg Inc Indication										932
	Dec-01		1.034			1.058																				Selected Ultimate Indication										654
	Jun-02		1.225
Straight Avg			1.089			1.003			1.027			1.008			1.007			1.013			1.011			1.002			1.008			0.994			1.002			1.000			0.998
Avg x HiLo			1.085			1.007			1.020			1.010			1.007			1.009			1.009			1.002			1.010			0.995			1.002
Avg Last 4			1.062			1.009			1.010			1.015			0.996			1.009			1.005			1.003			1.011			0.994
Fitted			1.063			1.022			1.012			1.008			1.006			1.004			1.003			1.003			1.002			1.002			1.002			1.002			1.001
Prior Sel @ 6 Mth			1.014			1.001			1.022			1.016			1.002			1.008			1.003			1.004			1.007			0.997			1.001			1.002			1.000
Prior Sel @ 3 Mth			1.130			1.030			1.007			1.021			1.007			1.011			1.009			1.006			0.997			1.006			0.998			1.000			1.000
Select			1.220			1.030			1.027			1.015			1.007			1.006			1.011			1.004			1.011			0.994			1.002			1.000			0.998		Tail
Cumulative			1.355			1.110			1.078			1.050			1.034			1.027			1.020			1.009			1.005			0.994			1.000			0.998			0.998			1.000
		Dec-02			Jun-02			Dec-01			Jun-01			Dec-00			Jun-00			Dec-99			Jun-99			Dec-98			Jun-98			Dec-97			Jun-97			Dec-96			Jun-96
Ult. Severity			7,165			7,501			6,671			6,682			5,643			6,436			6,192			5,918			5,144			5,677			6,159			6,899			5,969			5,950
Ult. Counts			1,350			1,164			1,219			1,512			1,826			2,173			2,144			1,888			1,480			1,312			1,161			773			714			595
Ultimate Loss			9,673			8,727			8,129			10,100			10,304			13,985			13,273			11,170			7,613			7,448			7,151			5,333			4,262			3,540
Ultimate LR			59.9	%		61.3	%		55.8	%		68.5	%		65.8	%		79.1	%		70.8	%		64.5	%		46.2	%		48.1	%		54.5	%		53.3	%		48.4	%		43.3	%
Ultimate PP			185			195			171			198			183			223			211			212			168			185			215			213			191			167

Exhibit C

State XYZ Auto as of December 31, 2002

RECORD PERIOD ANALYSIS

			(1)			(2)			(3)			(4)			(5)
Record			Paid			Avg. Paid			BF Paid			Incurred			Avg. Incurred
Semesters			Method			Method			Method Ult			Method			Method
Ending			Ult($000)			Ult ($000)			($000)			Ult ($000)			Ult($000)
PRIOR 3 yrs				34,764			34,720			34,763			34,729			34,727
Jun-1999				9,723			9,720			9,723			9,934			9,944
Dec- 1999				12,772			12,662			12,769			12,658			12,724
Jun-2000				14,159			14,183			14,160			14,656			14,692
Dec-2000				10,455			10,318			10,445			10,588			10,658
Jun-2001				10,108			9.948			9,919			10,923			10,955
D9C-2001				7,063			6,813			7,818			8,067			10,067
Jun-2002				8,484			8,264			8,764			8,584			8,727
Dec-2002				11,901			12,037			10,437			9,486			9,161
		Total		119,430			118,664			118,798			119,627			119,656
	Paid Loss			94,961			94,961			94,961			94,961			94,961
Required Reserve				24,468			23,703			23,837			24,666			24,694
Held Reserve				23,587			23,587			23,587			23,587			23,587
Reserve Adequacy				(882	)		(116	)		(250	)		(1,079	)		(1,108	)
Average Last 4				1,021			(2,991	)					559			1,378
Second Diagonal				1,284			(3,280	)					(1,435	)		242
Last Diagonal				(8,089	)		(7,008	)					1,646			1,614

[Additional columns below]

[Continued from above table, first column(s) repeated]

			(6)			(7)		(8)			(9)
Record			BF Incurred			Adj. inc. @		Pd. Loss @			Indicated
Semesters			Method			12/31/2002		12/31/2002			Ultimate
Ending			Ult ($000)			($000)		($000)			Loss ($000)
PRIOR 3 yrs				34,729			34,672		34,324			34,729
Jun-1999				9,933			9,867		9,368			9,937
Dec- 1999				12,659			12,573		11,966			12,681
Jun-2000				14,649			14,440		12,747			14,666
Dec-2000				10,586			10,482		8,918			10,611
Jun-2001				10,905			10,802		7,770			10,928
D9C-2001				8,077			7,995		4,535			8,071
Jun-2002				8,582			8,771		3,565			8,631
Dec-2002				9,487			9,597		1,768			9,323
		Total		119,609			119,199		94,961			119,577
	Paid Loss			94,961
Required Reserve				24,647						Required Reserve		24,615
Held Reserve				23,587						Held Reserve		23,587
Reserve Adequacy				(1,061	)					Reserve Adequacy		(1,029	)
Average Last 4
Second Diagonal
Last Diagonal

	(10)		(11)		(12)		(13)			(14)
Record Semesters	Avg. Paid		Avg. Incurred		Ultimate		Change In			Paid Counts
Ending	Severity		Severity		Severity		Severity			Method
PRIOR 3 yrs		5,867		5,868		5,868					5,914
Jun-1999		5,287		5,409		5,405					1,835
Dec- 1999		6,262		6,293		6,271		16.0	%		2,021
Jun-2000		6,438		6,669		6,657		6.2	%		2,172
Dec-2000		5,371		5,548		5,524		-17.0	%		1,904
Jun-2001		6,173		6,798		6,781		22.8	%		1,577
Dec-2001		5,605		6,637		6,640		-2.1	%		1,185
Jun-2002		7,118		7,517		7,434		12.0	%		1,075
Dec-2002		9,259		7,047		7,172		-3.5	%		1,230
											18,918

[Additional columns below]

[Continued from above table, first column(s) repeated]

	(15)		(16)		(17)		(18)
	Clsd. w/Pay		Incurred		Recored		ndicated
Record Semesters	Counts		Counts		Counts		Ultimate
Ending	Method		Method		Method		Counts
PRIOR 3 yrs		5,915		5,919		5,918		5,918
Jun-1999		1,837		1,839		1,838		1,839
Dec- 1999		2,023		2,024		2,020		2,022
Jun-2000		2,174		2,205		2,201		2,203
Dec-2000		1,905		1,922		1,920		1,921
Jun-2001		1,581		1,614		1,609		1,612
Dec-2001		1,182		1,219		1,212		1,216
Jun-2002		1,066		1,177		1,145		1,161
Dec-2002		1,135		1,412		1,373		1,300
		18,818		19,331		19,236		19,191

Page 5

Exhibit D

State XYZ Auto as of December 31, 2002

SUMMARY OF ESTIMATED IBNR

(1)	(2)		(3)	(4)		(5)		(6)
	Calculated		Quarterly
Prior Review	PP using		Rec w/n Acc	Total
Future	6 month		Periods	Future		Earned		Earned
Pure Premium	Emerged		Endjng	Pure Prem		Exposures		Premium
1.17		0.89	Sep-1998		0.60		22,103		8,156,777
1.65		1.22	Dec-1998		0.78		23,265		8,307,946
2.12		0.87	Mar-1999		0.98		24,674		8,417,123
2.43		1.05	Jun-1999		1.16		27,968		8,907,753
2.74		1.70	Sep-1999		1.35		30,751		9,331,069
3.05		1.80	Dec-1999		1.54		32,076		9,413,188
3.36		1.93	Mar-2000		1.73		31,817		9,094,404
3.80		2.10	Jun-2000		2.15		30,918		8,575,229
4.24		2.68	Sep-2000		2.58		29,011		7,995,863
4.69		3.13	Dec-2000		3.01		27,276		7,655,772
5.14		3.62	Mar-2001		3.44		26,502		7,425,622
5.69		4.11	Jun-2001		4.00		24,379		7,323,851
6.81		5.14	Sep-2001		4.78		24,375		7,367,666
7.58		5.64	Dec-2001		5.47		23,292		7,189,243
8.95		6.33	Mar-2002		6.59		22,123		7,035,903
11.31		9.00	Jun-2002		8.92		22,681		7,197,385
15.82		13.83	Sep-2002		11.74		25,217		7,724,378
35.65		34.05	Dec-2002		33.52		26,942		8,437,252
		2,560,917					475,370		145,556,425

[Additional columns below]

[Continued from above table]

(7)		(8)			(9)			(10)		(11)			(12)		(13)
								Avg IBNR					Avg IBNR
		Required			Current			Emerged		Required			Emerged		Required
Required		IBNR			IBNR			Since		IBNR			Since		IBNR
IBNR		Factors			Factors			Jun-2002		Factors			Jun-2002		Factors
	13,163		0.2	%					6,110		0.2	%		6,734		0.2	%
	18,249		0.2	%					6,110		0.3	%		6,734		0.3	%
	24,058		0.3	%		1.1	%		17,913		0.5	%		9,949		0.4	%
	32,482		0.4	%		1.1	%		17,913		0.6	%		9,949		0.5	%
	41,499		0.4	%		1.1	%		17,913		0.6	%		9,949		0.6	%
	49,381		0.5	%		1.1	%		17,913		0.7	%		9,949		0.6	%
	55,086		0.6	%		2.1	%		30,074		0.9	%		30,963		0.9	%
	66,559		0.8	%		2.1	%		30,074		1.1	%		30,963		1.1	%
	74,802		0.9	%		2.1	%		30,074		1.3	%		30,963		1.3	%
	82,052		1.1	%		2.1	%		30,074		1.5	%		30,963		1.5	%
	91,225		1.2	%		3.1	%		45,060		1.8	%		35,787		1.7	%
	97,579		1.3	%		3.0	%		39,863		1.9	%		37,362		1.8	%
	116,551		1.6	%		4.1	%		37,814		2.1	%		41,819		2.1	%
	127,483		1.8	%		4.5	%		82,033		2.9	%		77,573		2.9	%
	145,689		2.1	%		4.9	%		160,243		4.3	%		121,408		3.8	%
	202,219		2.8	%		5.7	%		570,118		10.7	%		552,718		10.5	%
	295,929		3.8	%		6.9	%
	903,099		10.7	%		11.6	%
	2,437,105				Total				1,139,299					1,043,783
Zero Runoff								Six Mth Runoff #1					Six Mth Runoff #2
	2,437				Indicated IBNR (000s)				2,377					2,282
	4,452				Carried IBNR (000s)				4,196					4,196
	2,015				Adequacy (000s)				1,819					1,914

Page 6

Exhibit E

Page 1

State XYZ Auto as of December 31, 2002

Quarterly
Rec w/n Ace INCURRED LOSSES QUARTERLY LAG 1 — IBNR ANALYSIS

Periods
Ending		0		1		2		3		4		5
	Sep-98		496,686		518,026		470,878		478,747		444,503		444,244
	Dec-98		563,065		602,907		574,026		582,766		533,957		521,264
	Mar-99		701,628		724,404		683,572		722,382		756,639		873,140
	Jun-99		926,112		992,882		936,052		1,074,640		1,040,663		990,419
	Sep-99		762,137		742,380		710,141		696,527		614,613		598,305
	Dec-99		1,403,988		1,429,926		1,328,895		1,282,623		1,223,390		1,217,221
	Mar-00		1,153,489		1,091,243		941,836		889,421		866,636		858,661
	Jun-00		1,273,559		1,032,051		895,131		827,142		819,180		787,061
	Sep-00		950,170		902,377		796,609		748,899		725,726		705,040
	Dec-00		945,192		904,793		884,895		836,910		796,854		766,037
	Mar-01		905,649		794,721		860,296		820,179		805,707		803,143
	Jun-01		473,607		400,046		393,187		374,912		353,414		326,284
	Sep-01		523,663		444,461		382,911		458,488		390,422
	Dec-01		772,190		700,711		604,762		577,760
	Mar-02		553,743		496,540		466,983
	Jun-02		559,263		621,759
	Sep-02		673,829
			0-1		1-2		2-3		3-4		4-5		5-6
	Sep-98		1.043		0.909		1.017		0.928		0.999		0.977
	Dec-98		1.071		0.952		1.015		0.916		0.976		0.997
	Mar-99		1.032		0.944		1.057		1.047		1.154		0.973
	Jun-99		1.072		0.943		1.148		0.968		0.952		0.891
	Sep-99		0.974		0.957		0.981		0.882		0.973		0.982
	Dec-99		1.018		0.929		0.965		0.954		0.995		1.004
	Mar-00		0.946		0.863		0.944		0.974		0.991		1.018
	Jun-00		0.810		0.867		0.924		0.990		0.961		0.942
	Sep-00		0.950		0.883		0.940		0.969		0.971		0.991
	Dec-00		0.957		0.978		0.946		0.952		0.961		0.978
	Mar-01		0.878		1.083		0.953		0.982		0.997		1.036
	Jun-01		0.845		0.983		0.954		0.943		0.923
	Sep-01		0.849		0.862		1.197		0.852
	Dec-01		0.907		0.863		0.955
	Mar-02		0.897		0.940
	Jun-02		1.112
Straight Avg			0.963		0.919		0.973		0.954		0.970		0.999
AvgxHiLo			0.963		0.918		0.971		0.951		0.968		0.990
Wtd Avg All			0.957		0.921		0.975		0.959		0.980		0.988
Avg Last 8			0.924		0.932		0.977		0.952		0.972		0.980
WtAvg.8			0.926		0.931		0.961		0.959		0.977		0.980
Avg Last 4			0.941		0.912		1.015		0.932		0.963		0.987
WtAvg.4			0.940		0.905		0.996		0.942		0.970		0.987
Select			0.957		0.921		0.977		0.952		0.972		0.980
Cumulative			0.782		0.817		0.887		0.908		0.954		0.982
Ultimate Loss			527,186		508,232		414,268		524,747		372,461		320,368

[Additional columns below]

[Continued from above table, first column(s) repeated]

Periods
Ending		6		7		8		Ultimate
	Sep-98		434,161		423,421		421,999		402,689
	Dec-98		519,834		510,883		508,050		496,017
	Mar-99		849,821		858,810		866,299		841,951
	Jun-99		882,542		936,217		922,570		893,998
	Sep-99		587,628		580,404		575,320		559,296
	Dec-99		1,222,239		1,219,462		1,209,300		1,182,188
	Mar-00		874,015		836,479		848,368		834,987
	Jun-00		741,515		750,787		725,020		687,127
	Sep-00		698,794		692,021		674,166		674,166
	Dec-00		749,239		745,902				745,902
	Mar-01		832359						833,632
	Jun-01								320,368
	Sep-01								372,461
	Dec-01								524,747
	Mar-02								414,268
	Jun-02								508,232
	Sep-02								527,186
			6-7		7-8		8-9
	Sep-98		0.975		0.997		0.973
	Dec-98		0.983		0.994		0.990
	Mar-99		1.011		1.009		0.973
	Jun-99		1.061		0.985		0.970
	Sep-99		0.988		0.991		0.992
	Dec-99		0.998		0.992		0.991
	Mar-00		0.957		1.014		0.990
	Jun-00		1.013		0.966		0.948
	Sep-00		0.990		0.974
	Dec-00		0.996
	Mar-01
	Jun-01
	Sep-01
	Dec-01
	Mar-02
	Jun-02
Straight Avg			0.997		1.008		1.009
AvgxHiLo			0.996		1.004		1.001
Wtd Avg All			1.000		1.000		0.998
Avg Last 8			1.002		0.991		0.978
WtAvg.8			1.002		0.991		0.979
Avg Last 4			0.989		0.986		0.980
WtAvg.4			0.987		0.988		0.981
Select			1.002		1.000		1.000
Cumulative			1.002		1.000		1.000
Ultimate Loss			833,632		745,902		674,166

Page 7

Exhibit E

Page 2

State XYZ Auto as of December 31, 2002

Quarterly
Rec w/n Ace		INCURRED LOSSES QUARTERLY LAG 0-7 EMERGENCE - IBNR ANALYSIS
Periods
Ending		0		1		2		3		4		5			6		7
	Sep-98		2,618,475		402,689		137,687		20,484		35,110		5,112			24,319		22,365
	Dec-98		3,340,158		496,017		226,614		96,699		34,548		22,026			19,872		9,514
	Mar-99		3,930,469		841,951		89,797		359,298		79,781		80,585			26,712		17,874
	Jun-99		4,092,147		893,998		255,026		144,934		117,008		50,989			28,301		19,937
	Sep-99		5,225,296		559,296		185,020		191,422		36,711		59,504			107,033		21,442
	Dec-99		4,830,247		1,182,188		250,533		281,475		42,245		32,341			16,295		10,619
	Mar-00		6,405,308		834,987		128,883		112,328		34,996		119,103			89,602		6,863
	Jun-00		4,924,350		687,127		134,681		169,119		91,889		(58,356	)		81,906		14,844
	Sep-00		4,186,260		674,166		139,948		115,327		17,039		14,286			13,298		21,573
	Dec-00		4,028,433		745,902		69,329		30,049		35,935		15,788			17,576		7,043
	Mar-01		4,370,268		833,632		189,272		37,264		24,374		16,504			29,907		15,153
	Jun-01		3,946,033		320,368		77,327		52,106		13,905		23,755			16,108
	Sep-01		3,375,799		372,461		34,944		53,028		29,579		8,235
	Dec-01		3,252,471		524,747		40,514		48,554		33,479
	Mar-02		3,751,012		414,268		106,068		54,175
	Jun-02		3,283,120		508,232		61,886
	Sep-02		3,631,092		527,186
	Dec-02		3,209,628
Avg 3 Emergence			3,374,613		483,229		69,489		51,919		25,654		16,165			21,197		14,590

Quarterly
Rec w/n Ace	INFLATED INCURRED LOSSES QUARTERLY LAG 0-7 EMERGENCE - IBNR ANALYSIS
Periods
Ending	0		1		2		3		4		5			6		7
Sep-98		3,093,429		471,089		159,503		23,498		39,883		5,750			27,089		24,669
Dec-98		3,907,512		574,608		259,958		109,845		38,862		24,535			21,919		10,392
Mar-99		4,553,228		965,836		102,005		404,161		88,867		88,887			29,176		19,333
Jun-99		4,694,268		1,015,535		286,870		161,440		129,062		55,693			30,610		21,353
Sep-99		5,935,663		629,132		206,092		211,142		40,098		64,360			114,637		22,741
Dec-99		5,433,371		1,316,826		276,343		307,443		45,692		34,639			17,282		11,153
Mar-00		7,134,798		921,007		140,773		121,494		37,482		126,320			94,104		7,138
Jun-00		5,431,656		750,520		145,671		181,134		97,457		(61,288	)		85,182		15,287
Sep-00		4,572,474		729,178		149,891		122,315		17,895		14,857			13,695		22,000
Dec-00		4,357,153		798,896		73,530		31,559		37,372		16,259			17,924		7,112
Mar-01		4,680,760		884,147		198,782		38,755		25,102		16,831			30,202		15,153
Jun-01		4,185,147		336,466		80,420		53,661		14,180		23,989			16,108
Sep-01		3,545,425		387,359		35,987		54,078		29,870		8,235
Dec-01		3,382,570		540,412		41,316		49,032		33,479
Mar-02		3,862,989		422,472		107,113		54,175
Jun-02		3,348,139		513,240		61,886
Sep-02		3,666,871		527,186
Dec-02		3,209,628

Page 8

Exhibit E

Page 3

State XYZ Auto as of December 31, 2002

Quarterly
Rec w/n Acc
Periods	INFLATED INCURRED LOSSES QUARTERLY LAG 0-7 PURE PREMIUMS - IBNR ANALYSIS
Ending	0		1		2		3		4		5			6		7
Mar-98		182.752		23.334		5.715		2.948		4.443		1.243			1.445		0.330
Jun-98		177.307		21.565		5.858		6.062		2.925		0.391			0.623		0.636
Sep-98		139.957		21.314		7.216		1.063		1.804		0.260			1.226		1.116
Dec-98		167.954		24.698		11.174		4.721		1.670		1.055			0.942		0.447
Mar-99		184.538		39.144		4.134		16.380		3.602		3.602			1.182		0.784
Jun-99		167.842		36.310		10.257		5.772		4.615		1.991			1.094		0.763
Sep-99		193.022		20.459		6.702		6.866		1.304		2.093			3.728		0.740
Dec-99		169.390		41.053		8.615		9.585		1.424		1.080			0.539		0.348
Mar-00		224.248		28.947		4.425		3.819		1.178		3.970			2.958		0.224
Jun-00		175.682		24.275		4.712		5.859		3.152		(1.982	)		2.755		0.494
Sep-00		157.611		25.134		5.167		4.216		0.617		0.512			0.472		0.758
Dec-00		159.741		29.289		2.696		1.157		1.370		0.596			0.657		0.261
Mar-01		176.618		33.361		7.501		1.462		0.947		0.635			1.140		0.572
Jun-01		171.668		13.801		3.299		2.201		0.582		0.984			0.661
Sep-01		145.453		15.892		1.476		2.219		1.225		0.338
Dec-01		145.225		23.202		1.774		2.105		1.437
Mar-02		174.614		19.096		4.842		2.449
Jun-02		147.619		22.629		2.729
Sep-02		145.414		20.906
Dec-02		119.133
Straight Avg		166.289		25.495		5.461		4.640		2.019		1.118			1.387		0.575
AvgxHiLo		165.689		25.268		5.352		4.096		1.936		1.137			1.269		0.557
Avg Last 8		153.218		22.272		3.685		2.708		1.314		0.767			1.614		0.520
Avg Last 4		146.695		21.458		2.705		2.243		1.048		0.638			0.732		0.521
Select		146.695		21.458		2.705		2.243		1.048		0.638			0.732		0.521

Quarterly
Rec w/n Acc																			Total
Periods	FUTURE PURE PREMIUMS BY QUARTERLY LAG, INFLATED																			Future
Ending	0	1		2		3			4		5		6		7		8-27		Pure Prem
Sep-98																		0.596		0.60
Dec-98																		0.784		0.78
Mar-99																		0.975		0.98
Jun-99																		1.161		1.16
Sep-99																		1.350		1.35
Dec-99																		1.540		1.54
Mar-00																		1.731		1.73
Jun-00																		2.153		2.15
Sep-00																		2.578		2.58
Dec-00																		3.008		3.01
Mar-01																		3.442		3.44
Jun-01																0.526		3.476		4.00
Sep-01														0.740		0.532		3.510		4.78
Dec-01												0.645		0.747		0.537		3.545		5.47
Mar-02										1.058		0.651		0.754		0.542		3.580		6.59
Jun-02							2.266			1.069		0.657		0.762		0.548		3.615		8.92
Sep-02					2.732		2.288			1.079		0.664		0.769		0.553		3.651		11.74
Dec-02			21.670		2.759		2.310			1.090		0.670		0.777		0.558		3.687		33.52
	Inflation rate used in IBNR calculation						4.0	%

Page 9

Exhibit A – Summary of Accident Period Analysis

This exhibit summarizes our accident period Analysis for this segment. Since it is an accident period analysis, theclaims are sorted and analyzed by accident date. We use 6-month accident periods (i.e. accident semesters) for this analysis. Each accident semester represents claims that occurred (or are expected to occur) during the 6-month period ending at the end of the designated month (in the left-hand column of the exhibit).

An accident period analysismeasures the adequacy of our total reserves (Case + IBNR).In other words, the estimated ultimate losses for each accident period include losses for claims that have already been reported to the company PLUS losses for claims that have not yet been reported.

What follows is an explanation of the information on Exhibit A:

• COLUMNS (1) – (6):Estimated ultimate losses resulting from six different sets of projections, along with the resulting required reserves and reserve adequacy for each respective method, as well as the adequacy that would have resulted using 3 different types of “default” selections for each method

• COLUMNS (7) and (8):Cumulative adjuster-incurred losses and paid losses as of the evaluation date of 12/31/2002

• COLUMN (9):Indicated ultimate losses which have been judgmentally selected by the actuaries considering all information obtained during the analysis, along with the resulting required reserves and reserve adequacy

• COLUMNS (10) and (11):Estimated ultimate average paid and average incurred severities, based upon the projections of average paid and average incurred losses

• COLUMN (12):Average adjuster case reserves, as of the first evaluation point (i.e. the evaluation date is the end-date of each respective accident semester)

• COLUMN (13):Closure Rate = [(Claims Closed with Pay) evaluated at 6 months development] divided by [Ultimate Selected Claim Count]

• COLUMNS (14) and (15):CWP Rate (i.e. percentage of reported claims which are closed without payment), as of the first evaluation point, and estimated to ultimate

• COLUMNS (16) – (19):Estimated ultimate counts resulting from four different sets of projections

• COLUMN (20):Indicated ultimate counts which have been judgmentally selected, considering all of the information obtained during the analysis

• COLUMNS (21) and (22):Indicated ultimate severities which result from the other selections (of losses and counts) in the process, along with the change from period to period, and the 4-point and 8-point fitted exponential trends. Fitted exponential trends tell us the estimated average annual change in severity (or frequency) considering our selections over the past two years (4 points) and four years (8 points)

• COLUMNS (23) and (24):Indicated ultimate frequencies which result from the selected ultimate counts, along with the change from period to period, and the 4-point and 8-point exponential trends

Page 10

• COLUMNS (25) and (26):The pure premiums and loss ratios which result from the selected ultimate losses

• COLUMNS (27) – (30):Earned premium and earned exposures, which are used in some of the other calculations, along with average earned premium, and changes in average earned premium

Page 11

The following chart contains columns (1) through (6) of Exhibit A, and will be explained in more detail below.

											(5) per
	(1)			(2)			(3)		(4)		Exhibit B		(6)
Accident	Paid			Avg. Paid			BF Paid		Incurred		Avg. Incurred		BF Incurred
Semesters	Method			Method			Method		Method		Method		Method
Ending	Ult ($000)			Ult ($000)			Ult ($000)		Ult ($000)		Ult ($000)		Ult ($000)
PRIOR 3 yrs		35,427			35,385			35,427		35,395		35,347		35,395
Jun-1999		10,930			10,940			10,931		11,193		11,170		11,193
Dec-1999		13,257			13,163			13,254		13,249		13,273		13,250
Jun-2000		13,534			13,781			13,603		14,012		13,985		14,141
Dec-2000		9,962			9,868			9,954		10,324		10,304		10,324
Jun-2001		9,485			9,492			9,435		10,149		10,100		10,127
Dec-2001		7,187			6,928			8,001		8,181		8,129		8,216
Jun-2002		9,689			8,667			9,207		8,842		8,727		8,846
Dec-2002		11,020			12,069			10,285		9,665		9,673		9,748
Tot Ult Loss		120,492			120,293			120,095		121,011		120,708		121,241
Tot Paid Loss		93,602			93,602			93,602		93,602		93,602		93,602
Required Reserves		26,890			26,691			26,494		27,409		27,106		27,639
Held Reserves		28,038			28,038			28,038		28,038		28,038		28,038
Reserve Adequacy		1,148			1,347			1,545		629		932		400
Avg Last 4		3,132			(2,025	)				3,261		3,835
2nd Diag		2,865			(3,318	)				624		1,951
Last Diag		(7,001	)		(6,264	)				3,470		3,154

We use six sets of projections in most of our segment analyses. There are other approaches built into our model that we use occasionally, when conditions warrant their use. However, we typically arrive at our indications using projections from all six of these data sets:paid losses; average paid losses; incurred losses; average incurred losses; and Bornhuetter-Ferguson expected loss ratio method using both paid and incurred loss development. Exhibit B goes into more detail regarding our selection process using the average incurred loss projection methodology. (Thus, there is a “box” around Column (5)). However, this discussion will focus more on the merits of each type of projection, the thought-processes behind the projections and the relationships between various components.

Note that the paid, average paid, incurred and average incurred projections all use a similar actuarial technique to estimate ultimate losses. As illustrated in Exhibit B, we organize the data into a triangular format, and project ultimate values by selecting development factors for each evaluation interval based upon historical patterns and judgment. In the exhibits and explanations, we refer to each of these sets of projections as “methods” even though they use basically the same actuarial method with different data sets.

Page 12

Estimated ultimate losses are generatedfor the past 7 accident years (by accident semester) for each of the six methods. These ultimate losses are shown on the exhibit for each of the past 8 accident semesters (4 years), and then the prior 3 accident years combined. Required reserves and reserve adequacy are then calculated (and shown in bold print below the total ultimate losses) for each method by using the identities stated at the beginning of this section:

Required Reserves = Total Ultimate Losses – Total Paid Losses
Reserve Adequacy = Held Reserves – Required Reserves

Below the reserve adequacy for each method, we show the adequacy that would have resulted from the application of three different types of“default” factor selectionsfor each method. Exhibit B shows more details behind these calculations. Exhibit A summarizes the results. “Avg Last 4” is the adequacy that would result if we had selected loss development factors equal to the average of the last 4 development factors at each development point. “2^nd Diag” and “Last Diag” are the adequacies that would result if we had selected loss development factors equal to those on each of the last 2 diagonals of the development triangle. The last diagonal represents the development (payments and/or adjuster case reserve changes) during the most recent 6 calendar months for each accident semester. The 2^nd-last diagonal represents the development during the 6-month period that ended 6 months ago.

(Paid and Incurred) Loss Development vs. Average (Paid and Incurred) Loss Development:

When we make our projections of ultimate losses, we need to consider trends in the frequency and severity of claims and consider the underlying influences on the historical changes in frequency and severity. The dollars of paid and incurred losses would be expected to increase in magnitude as our premium dollars and exposures increase. In the development of paid and incurred dollars, we observe these increases over time, but do not necessarily know whether they are due to increases in severity of claims, increases in the volume of business or a mixture of both. On the other hand,by looking at the development of average paid and average incurred losses, we are able to focus upon changes in severity over time. Therefore, we tend to rely more heavily on the development of average paid and average incurred losses (summarized in columns (2) and (5)) than that of the pure paid and incurred loss dollars (summarized in columns (1) and (4)).

• Each data point in theAverage Paid Lossdevelopment triangle =
[paid loss dollars] divided by [paid count]
where paid count = Claim features (closed or open) with loss payment

• Each data point in theAverage Incurred Lossdevelopment triangle =
[incurred loss dollars] divided by [incurred count]
where incurred count = Claim features closed with loss payment + open features

     TheUltimate Lossesfor the Average Incurred Method (column (5) exhibit A) are calculated for each accident semester as:

     [Ultimate Average Incurred Severity (column (11)] MULTIPLIED BY

     [Indicated Ultimate Counts (column (20)]

The ultimate average incurred severities are derived from the projections of average incurred losses, as shown in Exhibit B. The indicated ultimate counts are selected from the 4 methods of projected counts, as described later in this section. Similar calculations are relevant for the average paid method. These calculations are illustrated in the following chart, an excerpt from the relevant columns of Exhibit A:

Page 13

	(10)		(20)		(2)=(10)X(20)		(11)		(20)		(5)=(11)X(20)
Accident			Indicated		Avg. Paid		[Per Exh B]		Indicated		Avg. Incr
Semesters	Avg. Paid		Ultimate		Method		Avg. Incr		Ultimate		Method
Ending	Severity		Counts		Ult ($000)		Severity		Counts		Ult ($000)
PRIOR 3 yrs		5,863		6,035		35,385		5,857		6,035		35,347
Jun-1999		5,796		1,888		10,940		5,918		1,888		11,170
Dec-1999		6,141		2,144		13,163		6,192		2,144		13,273
Jun-2000		6,342		2,173		13,781		6,436		2,173		13,985
Dec-2000		5,404		1,826		9,868		5,643		1,826		10,304
Jun-2001		6,280		1,512		9,492		6,682		1,512		10,100
Dec-2001		5,686		1,219		6,928		6,671		1,219		8,129
Jun-2002		7,449		1,164		8,667		7,501		1,164		8,727
Dec-2002		8,940		1,350		12,069		7,165		1,350		9,673

Paid (and Average Paid) Losses- The development of paid losses is influenced by the rate at which claims arerecorded, the rate at which the claims arepaid and settled, and theseverityof the claims. Injury claims (BI, PIP, and UMBI) will tend to have more variability in development and a longer payment period than property/physical damage claims (Comprehensive, Collision, and Property Damage).

Therecording of claimscan be influenced by the time it takes for the claimant to report the claim, and the time it takes for the company to record the claim. The time it takes for the claimant to report the claim can be influenced by external forces, such as laws and regulations in the state, the legal environment, and the economy. The time it takes for the company to record the claim can be influenced by the adequacy of claim staffing, and changes in claim processing.

Some or all of the same items as mentioned for claim reporting and recording can also influence the rate at which claimsare paid and settled. In addition,severity of claims tends to be related to the rate of payment of claims. Smaller claims tend to settle more quickly, and larger claims tend to settle more slowly. As a result of this relationship, we consider the closure rate when making our judgments regarding paid (and average paid) loss development.

As stated above:Closure Rate =
[Claims Closed with Pay] divided by [Ultimate Selected Claim Count]

We look at this ratio to see if there is a change in the rate of claim closure, which may be impacting the paid loss development (historically and in the future). Column (13) of Exhibit A shows this closure rate at the first evaluation point for each accident period. We also look at further development points for the same reason, but it is the first development point (i.e. 6 months), which tends to be the most informative. This is because the closure rate tends to vary more when claims are less mature. Greater variability in the closure rate causes greater distortions in the development of paid losses.

Page 14

The following section from Exhibit A (as well as the underlying data) illustrates this point.

	(Data)		(20)		(13)
Accident	Features		Indicated		=(Data)/(20)
Semesters	Closed w/ Pay		Ultimate		Closure Rate
Ending	@ 6 Months		Counts		@ 6 Months
Jun-1999		636		1,888		33.7	%
Dec-1999		613		2,144		28.6	%
Jun-2000		568		2,173		26.1	%
Dec-2000		589		1,826		32.3	%
Jun-2001		466		1,512		30.8	%
Dec-2001		322		1,219		26.4	%
Jun-2002		273		1,164		23.5	%
Dec-2002		290		1,350		21.5	%

The closure rate has been decreasing for the past 4 consecutive accident semesters. This will tend todistort our historical paid (and average paid) loss development. The current paid losses will therefore not be expected to develop similarly to the historical paid losses. If a standard paid development projection is applied blindly, the resulting indication will likely not be reasonable.

Assuming that thelower severity claims are settled first, the trend seen in the closure rate would imply that the claims that have been paid in the most recent accident periods have a lower average severity (at the 6-month evaluation point) than those in the past. In addition, the future development on these losses may be understated if historical development patterns would be applied. Therefore, the ultimate losses may be understated, the required reserves may be understated, and thereserve adequacy may be overstated.

There is an actuarial approach known as theBerquist-Shermanmethod, which can be used to adjust historical paid losses for these distortions. A standard application of this methodology would assume that the current closure rate would apply going forward. Historical paid losses would be adjusted to reflect the development that would have been expected to occur if the current closure rates had been relevant in the past. We may utilize this methodology for some segments, but we would also supplement it with judgment.

Wediscuss the closure rate pattern with our claims management organizationto determine what may be causing the change in closure rate (e.g. process changes, staffing changes, or excessive growth). We will consider whether the trend is expected to continue or reverse, or whether we are now at the level that is expected to remain consistent. We consider this information in our selections for future development of paid (and average paid) losses.

With this specific segment,some of the hypotheses stated above are not necessarily true. In fact, application of the paid and average paid development factors from the most recent 6-month period (i.e. the “last diagonal”, as shown inredat the bottom of columns (1) and (2) above) would result in lower reserve adequacy. That is because, even though the closure rate is the lowest in history at the 6-month evaluation point, and even though it also tends to be low at further evaluation points, the rate of decrease is slowing.

Upon further review, the vast majority of the reserve inadequacy that results from the last diagonal of the paid methods is due to the most recent accident semester. For this period, even though the closure rate is lower than history, the average paid loss is higher than history. This is

Page 15

a time when it is especially helpful todiscuss these issues with management, to get additional information that may help in the analysis. It is quite possible that there are process changes or specific claims that may help to explain this development, and help us to make better judgments. This type of volatility in paid development also indicates that it may be preferable to give more credibility to the incurred projection in making our final selections of indicated ultimate losses.

Incurred (and Average Incurred) Losses– Incurred losses that we use in our analysis consist of paid losses plusadjuster-set case reserves. Recall from the section on Types of Reserves that the case reserve carried on the company’s financial records for each open claim is the (inflated) average reserve if it is below the threshold, or the [(ANCR factor) times the (adjuster-set reserve)] if it is greater than or equal to the threshold. However, when we analyze the development of incurred losses in order to make our projections, it is more meaningful to use the adjuster-set reserves for all claims, regardless of amount.

When a claim is reported, the case reserve is set according to the average in the case tables, as determined by the actuarial reviews. However,when the adjuster has enough information about the claim in order to make a reasonable estimate of its value, the adjuster may enter an adjuster-set reserve amount into the claims system. The adjuster then revises this amount when additional information is known about the claim. These adjustments can be done any time until the claim is settled.

It is appropriate for the financial records to carry the average amount (as long as it is below the threshold), since that is reasonably accurate at aggregate levels, and the averages are reviewed and inflated regularly. However, in order for theactuarial analysis of the claim development to consider all of the information available about the development of each claim, it is appropriate to use theadjuster-set case reservesfor this analysis.

Since incurred losses include case reserves on each claim,each claim carries an incurred loss estimate as soon as it is reported.Also, case reserves areadjusted whenever additional information is knownabout the claim. For these reasons, the incurred (and average incurred) losses are usually more reliable than the paid (and average paid) losses for projection of ultimate losses. We feel this is especially true when we have volatile closure rates (and therefore volatile paid loss development), as we do with this segment.

Since paid losses are a component of incurred losses, the same processes and forces influence the development of incurred losses. However, the case reserve estimation that is added to the Paid Loss amounts helps to mitigate some of the distortions causingincurred loss development to be more stable than paid loss development.At the same time, the case reserve estimation adds another type of uncertainty to the process. As with paid losses, injury claims (BI, PIP, and UMBI) will tend to have more variability in development and a longer development period than property/physical damage claims (comprehensive, collision, and property damage). Since liability claims will often involve lawsuits, those claims are more difficult for the adjusters to make estimates of the ultimate amount needed to fully settle a claim.

If the claim adjusting process and external influences were the same in every state, and if there were no changes in processes, personnel, or external forces over time, then projection of incurred losses would be fairly straightforward. However, change does occur and we need toanalyze our case reserving process in order to make reasonable projections of incurred losses.

One of the parameters we consider in the analysis is“Average Adjuster Case Reserves.”We expect an inflationary trend over time in the average adjuster case reserves. We also know that the changes we make to the averages in the case tables will have an impact on the adjuster reserves. This is because, up until the time the adjuster makes an estimate, the adjuster-set reserve is equal to the average reserve. At the 6-month development point, we expect approximately 50% of our BI liability claims will still carry the average. But as the claims age, the adjusters will set reserves on a greater proportion of the claims.

Page 16

We look at this parameter to see if there is achange in the adjuster activity that may be affecting the incurred loss development and the incurred severities. Column (12) of Exhibit A shows the average adjuster case reserve at the first evaluation point for each accident period. We also look at further evaluation points for the same reason, but the first evaluation point (i.e. 6 months) tends to be the most informative.

Also note that the averageadjuster case reserve amount is influenced by the closure ratediscussed earlier. Assume (as with this segment) that the closure rate is lower than history. Also assume that the lower severity claims close earlier than the higher severity claims. One could conclude then that the bulk of claim inventory includes a mixture of the higher severity claims AND the medium severity claims. Thus, the average adjuster case reserves would be expected to be lower than history.

The following excerpt from Exhibit A is also used to illustrate this point.

	(12)		(13)
Accident	Avg. Adjuster
Semesters	Case Reserves		Closure Rate
Ending	@ 6 Months		@ 6 Months
Jun-1999		4,207		33.7	%
Dec-1999		4,321		28.6	%
Jun-2000		5,341		26.1	%
Dec-2000		5,291		32.3	%
Jun-2001		5,462		30.8	%
Dec-2001		5,213		26.4	%
Jun-2002		4,606		23.5	%
Dec-2002		4,153		21.5	%

This data potentially confirms the hypothesis, which is that adecreasing closure rate will lead to decreasing average adjuster case reserves. However, there could also be other reasons for the decrease in these average adjuster case reserve amounts. Several possibilities are as follows:

• We may have madechanges to the averagesin the case tables that caused part of the decrease.

• There may have beenprocess changescausing:

• Adjusters to leave claims at the case table average for a longer period of time before assigning their own estimates

• Adjusters may be estimating the value of the claims differently

• Higher severity claims may be settling more quickly

• There may beexternal(legal, regulatory, or environmental) forces causing severity of open claims (or all claims) to decrease

Wediscuss the adjuster reserving patterns with managementto determine what may be causing this trend, whether it is expected to continue or reverse, or whether we are now at an expected level. We consider this information in our selections for future development of incurred (and average incurred) losses.

Page 17

If adjuster estimates are lower than history for similar claims, we select higher development factors to project ultimate losses. There is also aBerquist-Shermanapproach for incurred losses, which can be used to adjust historical incurred losses for changes in the strength of case reserves. A standard application of this approach would adjust the case reserve portion of the incurred losses to a level consistent with a selected trend in the average case reserves. We utilize this methodology for some segments, and supplement it with judgment.

Based upon the reserve adequacies shown in columns (4) and (5) of Exhibit A, we selected factors for the incurred methods that are somewhat higher, on average, than the default development factors from the recent diagonals.

Bornhuetter-Ferguson Method– The “BF” method (columns (3) and (6) on exhibit A for Paid and Incurred, respectively) is an actuarial methodology thatsmoothes the projected ultimate losses between a pure incurred or paid development method and an expected loss ratio.We usually utilize an estimated expected loss ratio on the four latest accident semesters only. This is because incurred development is reasonably reliable for periods over 2 years old, and the BF method does not add much value beyond that point for our lines of insurance. The expected loss ratio can be based upon an average of the other projections, a historical average, an expected loss ratio used in rate filings, and/or judgment. Given our growth and mix shifts, especially the changes in our limits profile (shift to higher policy limits), the expected loss ratios in the more recent periods may vary significantly from historical levels.

The following chart illustrates the calculation for the BF incurred ultimate loss amount for accident semester ending December 2002. This result is included in column (6) of Exhibit A.

		Accident
		Semester
	Bornhuetter-Ferguson using Incurred Loss Development	Ending Dec-02
(a)	Earned Premium		16,161,630
(b)	Expected Loss Ratio		63.0	%
(c)=(a)X(b)	Expected Ultimate Losses		10,181,827
(d)	Cumulative Loss Development Factor (6 mo to Ultimate) Per Incurred Loss Development		1.192
(e)=1-1/(d)	% Unreported		16.1	%
(f)	Actual Incurred Loss @ 12/31/02		8,106,511
(f)+((c)X(e))	Estimated Ultimate Loss		9,748,218

Indicated Ultimate Losses(Column (9)) – After consideration of the six sets of Paid and Incurred projections (in columns (1) through (6)), and all of the issues involved in those selections, we make our indicated ultimate loss selections for each accident semester. For this segment, we determined that the incurred methods are more reliable than the paid methods. Therefore, our selected ultimate losses are the average of the ultimate losses from the three incurred methods.

Sometimes, we may judgmentally select an ultimate loss amount for a few of the periods (usually the most recent periods) that are not based upon the six standard projections. It may be that the projected loss amount from the standard methods does not lead to a reasonable ultimate severity, pure premium, and/or loss ratio.We would expect severity and pure premium to have trends that reasonably reflect internal and external trends in loss costs and inflation.We discuss these trends (as well as the frequency trend) with line management in order to verify the reasonableness of our assumptions. We do not necessarily expect to match the trends selected by Pricing, but management should understand the reasons

Page 18

for the differences. We also expect the loss ratio to be somewhat level, other than changes due to business operations, rate levels, or mix.

Consider the following chart, a section from Exhibit A:

					(21)
	(9)		(20)		=(9) / (20)			(22)			(25)		(26)
Accident	Indicated		Indicated					Semi-Annual
Semesters	Ultimate		Ultimate		Ultimate			Change In			Pure		Loss
Ending	Loss ($000)		Counts		Severity			Severity			Premium		Ratio
PRIOR 3 yrs		35,379		6,035		5,862						189		49.1	%
Jun-1999		11,186		1,888		5,926						212		64.6	%
Dec-1999		13,257		2,144		6,185			4.4	%		211		70.7	%
Jun-2000		14,046		2,173		6,464			4.5	%		224		79.5	%
Dec-2000		10,318		1,826		5,650			-12.6	%		183		65.9	%
Jun-2001		10,125		1,512		6,699			18.6	%		199		68.6	%
Dec-2001		8,175		1,219		6,709			0.2	%		172		56.2	%
Jun-2002		8,805		1,164		7,568			12.8	%		197		61.9	%
Dec-2002		9,695		1,350		7,182			-5.1	%		186		60.0	%
Total		120,987		19,309		6.8	%	4 point exponential trend
						6.5	%	8 point exponential trend
Tot Paid Loss		93,602
Required Reserves		27,385
Held Reserves		28,038
Reserve Adequacy		654

Severity= Ultimate Losses / Ultimate Counts
Pure Premium= Ultimate Losses / Earned Exposures
Loss Ratio= Ultimate Losses / Earned Premium

If we do not believe that the severity is reasonable, we may select a different ultimate loss amount to make the resulting severity more reasonable. A revised selection would also be tested against the other parameters for reasonableness. For this segment, the severity for the last accident semester is 5% lower than the previous period (i.e. six months ago) but it is 7% higher than it was a year ago and the fitted annual trend of about 6.5% is reasonable. Changes in our mix of business may be causing the volatility in severity over the recent periods. The pure premiums and loss ratios that result from the selected losses also appear to be reasonable. As a result, we conclude that the ultimate loss selections are reasonable.

The Required Reserves and Reserve Adequacy are then calculated by using the Identities:

Required Reserves = Total Ultimate Losses – Total Paid Losses = 27,385
Reserve Adequacy = Held Reserves – Required Reserves = 654

Page 19

Therefore, based upon this Accident Period analysis, our total held reserves are$654,000 conservative.

Claim Counts and Frequency– The following chart contains columns (14) through (19) of Exhibit A.

	(14)			(15)			(16)		(17)		(18)		(19)
Accident							Paid		Closed w/ Pay		Incurred		Recorded
Semesters	CWP Rate			Ultimate			Counts		Counts		Counts		Counts
Ending	@ 6 Months			CWP Rate			Method		Method		Method		Method
PRIOR 3 yrs								6,030		6,029		6,032		6,035
Jun-1999		26.3	%		37.9	%		1,885		1,886		1,888		1,887
Dec-1999		29.4	%		40.4	%		2,144		2,145		2,145		2,142
Jun-2000		27.6	%		41.3	%		2,132		2,134		2,175		2,171
Dec-2000		26.3	%		39.8	%		1,821		1,822		1,827		1,825
Jun-2001		30.7	%		41.8	%		1,483		1,486		1,514		1,509
Dec-2001		29.2	%		42.5	%		1,198		1,196		1,222		1,215
Jun-2002		32.4	%		47.2	%		1,080		1,074		1,179		1,148
Dec-2002		28.7	%		43.1	%		1,177		1,104		1,449		1,379
								18,950		18,876		19,431		19,311

TheCWP Rateis the percentage of reported claims that are “Closed Without any Loss Payment.” In column (14), this percentage is calculated at 6 months of development. Column (15) shows our projections of the Ultimate CWP Rates. Changes in CWP rates tend to be due to process changes. For example, the previous process may have been to open claims as soon as they were reported, without sufficiently verifying whether coverage existed. Under the new process, we do not open claims until we have additional information regarding the validity of the claim. This would cause the CWP rate to decrease. Note thatthis change in process should not affect the “Closure Rate,” since the calculation of Closure Rate excludes claims closed without payment.

Claim counts shown in Columns (16) through (19) represent our projections of estimated ultimate counts of claims with loss payment for each accident semester. These estimates are made using 4 different sets of projections. For each method, we sort and analyze the feature counts by accident semester.

• The “Paid Count Method” uses feature counts for claims that have had loss payments, whether the features are currently open or closed.

• The “Closed With Pay Count Method” uses feature counts for claims that have closed with loss payment.

• The “Incurred Count Method” uses feature counts for claims that have closed with loss payment, plus claims that are currently open (whether or not there have been payments on them).

Page 20

• The “Recorded Count Method” uses feature counts for all claims that have been recorded. The projected ultimate recorded counts are multiplied by (100% minus the Ultimate CWP Rates in column (15)) for the same respective accident periods to derive the Ultimate Counts in Column (19). We do this to get the Ultimate Counts for claims WITH Loss Payment.

In this example, the projected counts using the Paid Count Methods are lower than those using the Incurred and Recorded Count Methods for the most recent accident periods. This is because of the closure rate issues discussed previously.Since the closure rate is lower than in the past, the projections of ultimate counts using paid counts will tend to be understated.

The following chart shows the selectedUltimate Counts, considering the projection methods, underlying information, and judgments discussed above. We tend to give more weight to the Incurred and Recorded Count methods in our selections since the changing closure rate is causing distortions in the paid count development. Also shown are the resultingfrequencies, the change in frequency from period to period, and the 4-point and 8-point annual fitted exponential trends. These fitted trends represent the average annual change in frequency, considering the historical selections over the past two years (4 points) and four years (8 points).

					(23)
	(20)		(28)		= (20)/(28)			(24)
Accident	Indicated							Semi-Annual
Semesters	Ultimate		Earned		Ultimate			Change In
Ending	Counts		Exposures		Frequency			Frequency
PRIOR 3 yrs		6,035		187,526		3.22	%
Jun-1999		1,888		52,642		3.59	%
Dec-1999		2,144		62,827		3.41	%		-4.8	%
Jun-2000		2,173		62,734		3.46	%		1.5	%
Dec-2000		1,826		56,287		3.24	%		-6.3	%
Jun-2001		1,512		50,881		2.97	%		-8.4	%
Dec-2001		1,219		47,667		2.56	%		-13.9	%
Jun-2002		1,164		44,804		2.60	%		1.6	%
Dec-2002		1,350		52,158		2.59	%		-0.3	%
Total		19,309		617,528		-7.6	%	4 point exponential trend
						-10.5	%	8 point exponential trend

We would expect frequency to have trends that reasonably reflect the Company’s mix of business and/or the industry results. Recent reductions in the frequency tend to be due to thechange in our mix of business as well as other external causes affecting the industry.We discuss this with management, in order to check the reasonableness of our assumptions. If we do not believe that the frequency is reasonable, we may select a different ultimate count to make the resulting frequency more reasonable. However, changes in the counts may also change the resulting severities.

Once we determine that the selected indicated loss amounts, frequencies, severities, pure premiums, and loss ratios are what we consider to be reasonable, we are done with this phase of the analysis. However, we may re-visit some of these selections after we have done the Record Period and IBNR analyses if they result in significantly different conclusions.

Page 21

As calculated above in column (9) of Exhibit A, our total held reserves are$654,000 conservative based upon this accident period analysis. We may reduce the reserves by that amount, or we may change the reserves by an amount other than that. We base this judgment upon the credibility of the indications in the review. When credibility is higher, the overall reserve change is closer to the indicated amount. The credibility is higher if our projections are relatively consistent with each other and the indications are consistent with prior reviews. On the other hand, if the projections are not reasonably consistent or if there are recent changes in the trends, we attach less credibility to the current review.

The Record Period and IBNR analyses will determine how the adequacy is distributed by type of reserve, and how we should implement the changes by category.

Page 22

Exhibit B –Accident Period Average Incurred Loss Development

This is one of the standard projection methods we use in our analysis for most segment reviews. It tends to be the method that is weighted most heavily in our selections of Indicated Ultimate Losses.

The top portion of Exhibit B (un-shaded area) contains actual data in a triangular format. The section of exhibit B shown below includes the actual data from the last 8 accident semesters, evaluated at 6-month intervals (semi-annual). The figures in theBlue Shadedcells are “projected” data points, and they will be discussed later. The last column shows Ultimate Severities that result from the analysis that follows. Note that these ultimate severities are also carried over to column (11) of Exhibit A, as discussed previously.

Semiannual
Accident
Periods	AVERAGE INCURRED LOSSES - ACCIDENT PERIOD ANALYSIS																Ultimate
Ending	1		2		3		4		5		6		7		8		Severity
Jun-99		4,315		5,241		5,457		5,704		5,786		5,787		5,822		5,865		5,918
Dec-99		4,830		5,839		5,985		5,975		6,088		6,058		6,068		6,136		6,192
Jun-00		6,277		6,306		6,180		6,140		6,283		6,269		6,307		6,378		6,436
Dec-00		5,440		5,411		5,274		5,440		5,456		5,497		5,530		5,592		5,643
Jun-01		6,155		6,126		6,269		6,366		6,461		6,509		6,548		6,622		6,682
Dec-01		5,657		5,850		6,189		6,356		6,451		6,499		6,538		6,611		6,671
Jun-02		5,513		6,756		6,959		7,147		7,253		7,307		7,351		7,434		7,501
Dec-02		5,289		6,453		6,647		6,826		6,928		6,979		7,021		7,100		7,165

Each data point in theAverage Incurred Lossdevelopment triangle =
[Incurred Loss dollars] divided by [Incurred Count]
where Incurred Count = claim features closed with loss payment + open features

Also recall that incurred losses that we use in our analysis are made up of paid losses plusadjuster-set case reserves.

The ending month of each accident semester is in the left-hand column. The evaluation points (across the top) represent 6-month periods. The first evaluation point is the same date as the end of each respective accident period. Each successive evaluation point is 6 months later. The last (i.e. most recent, or current) evaluation point for each accident semester is the end of December, 2002. This is sometimes referred to as the “Last Diagonal”. This last diagonal is shown inRedon the chart.

For example, for the accident semester ending December, 2001, the loss amount and count data that underlies the average incurred losses (inBold Blue, with the current evaluation being inRed) in the above chart is as follows:

Page 23

		Evaluated at Month-End
	Accident Semester
	Ending Dec-01	Dec, 2001		Jun, 2002		Dec, 2002
(a)	Paid Losses ($000)		646		2,414		4,238
(b)	Adjuster Case Reserves ($000)		6,719		5,295		3,653
(c)=(a)+(b)	Incurred Losses ($000)		7,365		7,709		7,891
(d)	Features closed w/ Payment		322		677		969
(e)	Open features		980		641		307
(f)=(d)+(e)	Incurred Count		1,302		1,318		1,275
(g)=(c)/(f)	Average Incurred Loss ($)		5,657		5,850		6,189

The middle portion of Exhibit B contains theage-to-age development factors, or “link ratios”in a triangular format. Each link ratio represents the “development” from one evaluation point to the next. For example, in the next chart, the link ratios for the accident semester ending December, 2001 are calculated as follows:

The development (of average incurred losses from the triangle at the top portion of Exhibit B) from evaluation point 1 to evaluation point 2 (or from Dec, 2001 to June, 2002) =5,850 / 5,657 = 1.034. This means that, during that 6 month maturity period, the average incurred losses for that accident period increased by 3.4 percent. Similarly, from June, 2002 to Dec, 2002, the ratio was6,189 / 5,850 = 1.058,or a 5.8 percent increase in the average incurred loss during that interval.

These calculations are done for every pair of data points on the triangle. (Notice that the last diagonal in the chart below is again coloredRed. Also, the 2^nd-last diagonal is coloredBlue).

The purpose of this is to seehow the claims have developed historically. This historical information is then used, along with other information and judgment, to estimate how the claims will develop in the future. If the data were “well-behaved,” you would expect the development factors to be consistent down each column. This would indicate that claim reporting, reserving and settlement patterns have been consistent in the past.

You can see in the following table that the development factors are not consistent for State XYZ. We need to consider other parts of our analysis, as well as other information management can provide to try and understand the reasons for this inconsistent pattern. We use that information toselect the development factors for estimated future development.

In order to assist in this process, weaverage the development factors down each columnin several ways. Wealso look at selectionswe made at the same intervalsfrom previous reviews.This information is near the bottom of Exhibit B. Significant portions of this are also included in the chart below, along with the selected factors and the resulting ultimate severities.

Page 24

Semiannual	Average Incurred Losses
Accident	Age-to-Age Development Factors
Periods
Ending	1-2		2-3		3-4		4-5		5-6		6-7		7-8
Jun-99		1.215		1.041		1.045		1.014		1.000		1.006		1.007
Dec-99		1.209		1.025		0.998		1.019		0.995		1.002
Jun-00		1.005		0.980		0.993		1.023		0.998
Dec-00		0.995		0.975		1.031		1.003
Jun-01		0.995		1.023		1.016
Dec-01		1.034		1.058
Jun-02		1.225

	Average Factors and Selected Factors
	1-2		2-3		3-4		4-5		5-6		6-7		7-8
Avg. Last 4		1.062		1.009		1.010		1.015		0.996		1.009		1.005
Prior Select @ 6 Months		1.014		1.001		1.022		1.016		1.002		1.008		1.003
Prior Select @ 3 Months		1.130		1.030		1.007		1.021		1.007		1.011		1.009
Selected Factor (a)		1.220		1.030		1.027		1.015		1.007		1.006		1.011
Cumulative Factor
(b) = (next (b)) X (a)		1.355		1.110		1.078		1.050		1.034		1.027		1.020

	Accident Semester Ending
	Dec-02		Jun-02		Dec-01		Jun-01		Dec-00		Jun-00		Dec-99
Last Diagonal (c)
(from Average Incurred
Loss Triangle)		5,289		6,756		6,189		6,366		5,456		6,269		6,068
Ultimate Severity
(d) = (b) X (c)		7,165		7,501		6,671		6,682		5,643		6,436		6,192

“Avg. Last 4”means the arithmetic average of the last 4 factors from that respective development interval (i.e. from the column directly above). This tells us how the average incurred losses have developed over that interval over the past 4 semesters.
For example, for the 1-2 development interval,
“Avg. Last 4”= (.995+.995+1.034+1.225) / 4 =1.062.

Since we review most segments every 3 months, the“Prior Selections”are shown for the most recent review(@ 3 months), and the review prior to that(@ 6 months). This gives us some perspective on how the actual development has changed relative to our prior reviews, and how our opinions have changed with updated information.

TheSelected Factorsare coloredgreenin the chart above. The most significant amount of judgment goes into the selection of the Development Factor for the first age interval, “1-2”. This is because these claims are the least mature. Therefore, there is less information known about them. The selected factor of1.220is higher than the average of the last 4, as well as both of the prior selections for that interval. The actual development factor from the most recent 6 months

Page 25

(i.e. thelast diagonal) was1.225. This is the highest that it has been in history and the selection shows that we expect this higher development in the future.

Recall the discussion of the“Average Adjuster Case Reserves”from Exhibit A. They decreased (at the 6-month evaluation point) for each of the past 3 semesters. Not surprisingly, the“Average Incurred Losses”have also decreased for each of the past 3 semesters (at the 6-month evaluation point, i.e. the first column). Therefore, we expect the future development on incurred losses to be higher than history.

TheBlue Shadedportion in the chart at the beginning of this section (and at the top of Exhibit B) showshow we expect the average incurred losses to develop over timebased upon our selected factors. For example, for accident semester ending December, 2002, the current evaluation of the average incurred losses (last diagonal) is5,289per claim. When this is multiplied by the selected 1-2 development factor of1.220, the resulting average in the firstBlue Shadedcell of that accident period is6,453. That is what we project the average incurred losses to be for accident semester December 2002, when they are evaluated 6 months later (at June 2003). Similar calculations are done for each development period and each accident period.

When the selected age-to-age development factors are multiplied by each other from the current development point (last diagonal) to the ultimate development (when all claims are expected to be closed), the resulting factor is called the“Cumulative”loss development factor (LDF). The Ultimate Severity for each accident period is then the amount at thelast diagonal, multiplied by the Cumulative factor. For example, for the Accident Semester ending December, 2002:

Ultimate Severity = 5,289 X 1.355 = 7,165

As explained previously (in the discussion of Exhibit A),ultimate severities are multiplied by the indicated ultimate counts, to derive the ultimate lossesfrom this method. Both the ultimate severities and the ultimate losses are carried onto Exhibit A, to consider in the final selections.

There is another reasonableness test done on Exhibit B. We compare the adequacies that would be derived from several different selections of loss development factors. This chart is from the box in the middle of Exhibit B, about 2/3 of the way across the page. It uses the identities as discussed previously:

Required Reserves = Total Ultimate Losses – Total Paid Losses
Reserve Adequacy = Held Reserves – Required Reserves

Reserve Adequacy based on defaulted and actual selections
of loss development factors using Average Incurred Development

	Adequacy
Loss Development Factors	($000)
Average Last 4		3,835
Second Diagonal		1,951
Last Diagonal		3,154
Selected Avg Inc Indication		932
Selected Ultimate Indication		654

According to the final selections of indicated ultimate losses, the loss reserve adequacy is654. This calculation is summarized on Exhibit A. The chart shows that, according to our selections from the average incurred development method, the adequacy would be932. We relied upon

Page 26

this method, as well as the incurred loss development and the incurred Borhuetter-Ferguson method, for our final selections.

One should note that, had we used default selections for the loss development factors from the average incurred development, our adequacy would have been higher. These “default” selections, as shown in the chart, are based on factors from the“Average of the Last 4", the“Second Diagonal”and the“Last Diagonal.”For example, the factors on the “Last Diagonal” are shown inRed above (on the triangle of Age-to-Age Development Factors). If the current losses would develop at the rate indicated by this set of factors, adequacy would be3,154. Similarly, if the current losses would develop according to the factors along theSecond-Last Diagonal, as shown inblueabove, adequacy would be1,951.

On average, our selected factors are higher than the default factors. This is because we expect the average incurred losses to develop at a higher rate in the future than they have in the recent past.Higher selected development factors lead to higher ultimate losses, which lead to higher required reserves, which lead to lower reserve adequacy. Therefore, even though our selected adequacy is outside of the range of the “default” selections, it is reasonable, based upon other information we have gained through the analysis.

Page 27

Exhibit C – Summary of Record Period Analysis

This exhibit summarizes our Record Period Analysis for this segment. Since it is a Record Period analysis, theclaims are sorted and analyzed by record date. We utilize 6-month record periods (i.e. record semesters) for our analysis. Therefore, each record semester represents all claims that have been recorded during the 6-month period ending at the end of the designated month (in the left-hand column of the exhibit).

A Record Period analysis measures theadequacy of our case reserves. In other words, the estimated ultimate losses for each record period include losses for claims that have already been recorded on the Company’s books. They do NOT include losses for unrecorded claims. Therefore, theyexclude IBNR.

The information summarized on this exhibit is similar to the information summarized on Exhibit A. Theissues involved in the analysis of record period losses are similar to the issues for accident period losses.The calculations of the components of the analyses are also very similar. Therefore, the focus of this discussion will be to compare and contrast the results of Exhibit C (Record Period, or RP) with Exhibit A (Accident Period, or AP).

In both the RP and AP analyses, the adequacies for the Paid methods are higher than those of the Incurred methods. The reasons (changes in closure rate) were discussed with Exhibit A, and the same reasoning is relevant for record period development. However, almost every one of the defaulted AND selected adequacies is LOWER for the RP analysis than for the same respective methods in the AP analysis. Thus, the case reserve adequacy is LOWER than the total reserve adequacy. This is summarized in the following chart, which pulls information from both Exhibits A and C:

	(1)			(2)			(3)			(4)			(5)			(6)			(9)
													Avg.			BF
	Paid			Avg. Paid			BF Paid			Incurred			Incurred			Incurred
Reserve	Method			Method			Method			Method			Method			Method			Indicated
Adequacy	($000)			($000)			($000)			($000)			($000)			($000)			($000)
	Accident Period Analysis (Exhibit A)
Selected		1,148			1,347			1,545			629			932			400			654
Avg Last 4		3,132			(2,025	)					3,261			3,835
2nd Diag		2,865			(3,318	)					624			1,951
Last Diag		(7,001	)		(6,264	)					3,470			3,154
	Record Period Analysis (Exhibit C)
Selected		(882	)		(116	)		(250	)		(1,079	)		(1,108	)		(1,061	)		(1,029	)
Avg Last 4		1,021			(2,991	)					559			1,378
2nd Diag		1,284			(3,280	)					(1,436	)		242
Last Diag		(8,089	)		(7,008	)					1,646			1,614

Page 28

This is a consistent result. Since the actual record period development results in lower adequacy, the selected adequacy would be expected to be lower for case reserves than for total reserves.

Thetiming difference between when accidents occur, and when they are recorded will help explain how severities differ between the analyses. Accidents are reported and recorded AFTER they occur, and severity is normally expected to increase over time, due to inflation. Therefore, for a given period-ending date,the Record Period severity (for accidents from earlier periods) is expected to be LOWER than the Accident Period severityfor the same respective period-ending date. The following chart illustrates this reasonableness test for this segment:

	Ultimate Severity
	Exh A (21)		Exh C (12)
Semesters	Accident		Record
Ending	Periods		Periods
Jun-1999		5,926		5,405
Dec-1999		6,185		6,271
Jun-2000		6,464		6,657
Dec-2000		5,650		5,524
Jun-2001		6,699		6,781
Dec-2001		6,709		6,640
Jun-2002		7,568		7,434
Dec-2002		7,182		7,172

In this example, the relationship is reasonable for most of the periods. For the periods in which it is not, there may be some large claims for which the timing difference is not consistent. Also, the severity trend is not consistent in this segment; there was a significant decrease in severity in the latter half of 2000. The underlying data should be investigated to determine whether this inconsistency is reasonable.

Theindicated ultimate counts should also be similar, in aggregate, between the two analyses.If frequency is relatively flat, and we are growing in volume, the aggregate claim counts should be higher for the accident period analysis than for the record year analysis due to the expected time lag between the occurrence and the recording of accidents. In this segment, we were actually declining in premium and exposure volume over most of the past two years, until a recent growth spurt. This may be partially driven by recent rate increases by our competitors due to industry-wide severity increases. In addition, frequency has been decreasing over most of the period, but it has flattened out over the past year. The aggregate accident period counts are slightly higher than the aggregate record period counts. This is a reasonable result.

We have determined the following:

Adequacy of Total Reserves, per Accident Period analysis

Adequacy of Case Reserves, per Record Period analysis

Page 29

Since IBNR Reserves = Total Reserves – Case Reserves, then:
    Adequacy of IBNR Reserves
= Adequacy of Total Reserves – Adequacy of Case Reserves
= 654 – (1,029) = 1,683

We do a separate analysis of “Late Reported Claims” by lag period, in order to independently determine IBNR reserve adequacy. We compare the results of that analysis to the results above to test for reasonableness.

Page 30

Exhibit D – Summary of Estimated IBNR

This exhibit summarizes our IBNR analysis for this segment. The section on “Types of Reserves” briefly explained what IBNR reserves are, and some of the theory behind the calculations. The focus of this section will be to relate the theory to the actual application of the results and to discuss some of the issues involved.

Exhibit E (discussed later) illustrates some of the details behind the calculations and selections of future pure premiums used in this exhibit (column (4)).In order to estimate the required IBNR reserves, we group the claims by lag period(i.e. the length of time from claim occurrence date to record date). Then,the claims for each lag period are sorted and analyzed by accident date.

IBNR Reserves represent estimates of losses for claims that have already occurred, but have not yet been recorded by the Company. These are sometimes referred to as“Late Reported Claims.”The later the claim is recorded (after the accident occurrence), the longer the lag period.

Exhibit D summarizes 4 1/2 years ofrequired IBNR, by accident quarter. The relevant accident periods are shown in column (3). The most recent period would have the largest proportion of required IBNR, since it would have the largest proportion of unreported claims. Therefore, we will focus on the most recent accident quarter. The following chart shows columns (1) through (9) from the December, 2002 row of Exhibit D:

(Column)	Description	Amount
(1)	Prior Review Future Pure Premium		$35.65
(2)	Calculated Pure Premium using 6-mo. Emerged		$34.05
(3)	Quarterly Record w/in Accident Period Ending	Dec-2002
(4)	Total Future Pure Premium		$33.52
(5)	Earned Exposures		26,942
(6)	Earned Premium		$8,437,252
(7)	Required IBNR =(4) X (5)		$ 903,099
(8)	Required IBNR Factor =(7) / (6)		10.7%
(9)	Current IBNR Factor		11.6%

     Pure Premium is defined as Losses per Exposure(or per Earned Car Year).

At the time of the prior review, we projected that the required IBNR reserves were$35.65per exposure (Column (1)) for the most recent accident quarter. However, we now have information on claims that have been reported (or “emerged”) since that evaluation date, on accidents that occurred prior to that date. Based upon the emergence over the past 6 months, we now (retrospectively) project that the required IBNR reserves should have been$34.05per exposure (Column (2)) for the most recent accident quarter. Therefore, theactual emergence has been slightly lower than expected for this period.

Page 31

The following chart shows the first 4 columns of Exhibit D for the 8 most recent accident quarters:

(1)	(2)		(3)	(4)
	Calculated		Quarterly	Selected
Prior Review	Pure Premium		Record within	Total
Future	Using 6 month		Accident Periods	Future
Pure Premium	Emerged		Ending	Pure Prem
5.14		3.62	Mar-2001		3.44
5.69		4.11	Jun-2001		4.00
6.81		5.14	Sep-2001		4.78
7.58		5.64	Dec-2001		5.47
8.95		6.33	Mar-2002		6.59
11.31		9.00	Jun-2002		8.92
15.82		13.83	Sep-2002		11.74
35.65		34.05	Dec-2002		33.52

If you compare all of column (2) to column (1) on Exhibit D, you can see that we have experienced “favorable”IBNR emergencein all accident periods. (Note that this result contrasts with our actual results for most segments in the early part of 2003, in which we experienced unfavorable IBNR emergence). The result in this case studymay be due to:

• Fewer claims than expected were reported (i.e. lower frequency than expected)

• The severity of the “late reported claims” has been lower than expected

• There may have been a process change that impacts the timing of claim reporting and/or the severity of late reported claims

• There may be external forces that impact timing of claim reporting and/or the severity of the late reported claims

Our selected pure premiums are based upon the actual emergence and development of “late reported claims” (by reporting lag period within each accident period), as well as judgment. They also include an expected level of inflation, since our current IBNR reserves need to be at the cost level that is relevant to each respective accident and record period. The selected“Future Pure Premiums”are shown in column (4). We selected$33.52per exposure for the most recent accident period. The details of the calculations that make up these Future Pure Premiums are included in Exhibit E, and explained later in this section.

Page 32

The following chart shows columns (3) through (9) of Exhibit D for the 8 most recent accident quarters:

(3)	(4)		(5)		(6)		(7)		(8)			(9)
Quarterly							=(4)X(5)		=(7)/(6)
Rec w/n Acc	Total								Required			Current
Periods	Future		Earned		Earned		Required		IBNR			IBNR
Ending	Pure Prem		Exposures		Premium		IBNR		Factors			Factors
Mar-2001		3.44		26,502		7,425,622		91,225		1.2	%		3.1	%
Jun-2001		4.00		24,379		7,323,851		97,579		1.3	%		3.0	%
Sep-2001		4.78		24,375		7,367,666		116,551		1.6	%		4.1	%
Dec-2001		5.47		23,292		7,189,243		127,483		1.8	%		4.5	%
Mar-2002		6.59		22,123		7,035,903		145,689		2.1	%		4.9	%
Jun-2002		8.92		22,681		7,197,385		202,219		2.8	%		5.7	%
Sep-2002		11.74		25,217		7,724,378		295,929		3.8	%		6.9	%
Dec-2002		33.52		26,942		8,437,252		903,099		10.7	%		11.6	%

The required IBNR in column (7) represents the amount of IBNR reserves that we expect to need for payment of claims that have occurred but have not yet been recorded for each respective accident period. Since pure premium is “Losses per Exposure”, you need to multiply it by the number of exposures during that period to get the expected amount of losses. For the accident quarter ending December 2002, shown above, this calculation is as follows:

Required IBNR	=	[Future Pure Premium]	X	[Earned Exposures]
		=      [33.52]	X	[26,942]
		=      903,099

In order to carry the appropriate level of IBNR reserves on the Company’s books, we assign “IBNR Factors” to each trailing 3-month period of earned premium. Therefore, our IBNR reserves will change as our premium volume changes. This allows our IBNR reserves to keep up with inflation, as well as our increased exposure to claims.

The required IBNR factors in column (8) are then calculated as follows in the example above:

IBNR Factor	=	[Required IBNR]	/	[Earned Premium]
	= =	[903,099]      10.7%	/	[8,437,252]

The required factors in column (8) are less than the current factors in column (9). This is not surprising, since we experienced favorable emergence. There is a test that we do to check the reasonableness of our required factors in column (8). We compare these to the factors that result from the actual emergence from the past 6 months, as well as the average 6-month emergence from the past 18 months. This information is shown in the following excerpt from Exhibit D.

Page 33

				(11)			(13)
(3)	(8)			6-mo				avg 6-mo
Quarterly	Selected			Emerged #1				Emerged #2
Record within	Required			Required				Required
Accident Periods	IBNR			IBNR				IBNR
Ending	Factors			Factors				Factors
Sep-2000					1.3	%			1.3	%
Dec-2000					1.5	%			1.5	%
Mar-2001		1.2	%		1.8	%			1.7	%
Jun-2001		1.3	%		1.9	%			1.8	%
Sep-2001		1.6	%		2.1	%			2.1	%
Dec-2001		1.8	%		2.9	%			2.9	%
Mar-2002		2.1	%		4.3	%			3.8	%
Jun-2002		2.8	%		10.7	%			10.5	%
Sep-2002		3.8	%
Dec-2002		10.7	%

Each current required factor in column (8) would be compared to the “emerged” required factors in columns (11) and (13) from 2 quarters prior (2 rows up). This shows that the selected required factors are reasonable, based upon the recent emergence patterns.

The bottom portion of Exhibit D summarizes the IBNR Reserve Adequacy, by comparing the required IBNR Reserves to the Carried IBNR Reserves. This is summarized below:

($000's)
2,437	Indicated IBNR (Sum column 7)
4,452	Carried IBNR
2,015	Adequacy = Carried — Indicated

This shows that our IBNR Reserves are$2.015 million conservative. The Indicated IBNR at the bottom of column (7) is the sum of the Required IBNR for all accident periods, based upon the calculations as illustrated above.

As mentioned previously,IBNR Reserves = Total Reserves – Case Reserves.

Expected IBNR reserve adequacy per the Accident Period and Record Period analyses, was:

= [Adequacy of Total Reserves (Exhibit A)] – [Adequacy of Case Reserves (Exhibit C)]

= 654 – (1,029) = 1,683

Difference in IBNR adequacy per separate analyses

= [Adequacy per IBNR analysis] – [Adequacy per accident and record period analyses]

= 2,015 — 1,683 = 332

Page 34

Since ourtotal carried loss reservesfor this segment are28,038(as shown on Exhibit A), thisdifference in IBNR adequacy of 332 is about one percent. This is a reasonable difference.

We revise our IBNR factors in the indicated direction, in order to move our carried IBNR reserves toward the indicated amount. By decreasing IBNR Reserves and increasing Case Reserves, we will obtain a reserve level that is consistent with our indications. Our Case, IBNR and Total Loss Reserves will all be at a level that is adequate for expected future payments on claims for which we are liable.

Page 35

Exhibit E – IBNR Analysis

In order to estimate the required level of IBNR reserves, we need to estimate the expected future pure premiums by accident quarter. These selected pure premiums were shown in column (4) of Exhibit D. They are determined by summing the estimated (inflated) future pure premiums for each report lag period within each accident quarter. We select these lag pure premiums by grouping the claim data by Lag Period, and then sort and analyze it by accident quarter for each Lag Period. Exhibit E summarizes the steps involved in this process.

Step 1:Selection of Ultimate Losses by accident period for each report Lag group. We do this for 8 quarterly Lag groups (from Lag 0 quarters to Lag 7 quarters); and for 5 annual Lag groups (from Lag 2 years to Lag 6 years).

The Lag 0 quarter includes all claims that are recorded in the same quarter in which the accidents occurred. Therefore, there are no late reported claims as of the end of the accident quarter. TheLag 1quarter includes all claims that are recorded in the quarter following the quarter in which the accidents occurred. The following chart is an excerpt from Page 1 of Exhibit E, showing the development of incurred losses for Lag 1 by accident quarter, as well as the selected development factors and ultimate loss amounts.

Quarterly
Record w/in Accident	INCURRED LOSSES QUARTERLY LAG 1 - IBNR ANALYSIS
Period End	0		1		2		3		4		Ultimate
Jun-01		473,607		400,046		393,187		374,912		353,414		320,368
Sep-01		523,663		444,461		382,911		458,488		390,422		372,461
Dec-01		772,190		700,711		604,762		577,760				524,747
Mar-02		553,743		496,540		466,983						414,268
Jun-02		559,263		621,759								508,232
Sep-02		673,829										527,186

Loss Development Factors (LDF)	0-1	1-2	2-3	3-4	4-5
Jun-01	0.845	0.983	0.954	0.943	0.923
Sep-01	0.849	0.862	1.197	0.852
Dec-01	0.907	0.863	0.955
Mar-02	0.897	0.940
Jun-02	1.112
Average Last 4	0.941	0.912	1.015	0.932	0.963
Weighted Avg. Last 4	0.940	0.905	0.996	0.942	0.970
Select	0.957	0.921	0.977	0.952	0.972
Cumulative	0.782	0.817	0.887	0.908	0.954
Ultimate Loss = [Incurred Loss (last diag)] X [Cumulative LDF]	527,186	508,232	414,268	524,747	372,461

Page 36

The “0” development column represents the incurred losses evaluated as of the end of the quarter that the claims were recorded. For example, theRedamount of673,829in the above chart represents the incurred loss amount for claims that occurred in the quarter ending September 2002, and were recorded in the quarter ending December 2002 (i.e. one Lag quarter), evaluated as of the end of December 2002. Note that the accident quarter ending December 2002 has not yet experienced any lag 1 claims (since those would be recorded in the future). Thus, the most recent accident period in the lag 1 triangle is September 2002.

In order to select loss development factors for the IBNR analysis, we go through a similar process that we do for the accident period and record period analyses. We use averages of the link ratios, as well as judgment in the selection process. We go through this selection process for each of the Report Lag groups.

Step 2:Summarize the ultimate losses for all Lag groupsinto an exhibit like that shown on Page 2 of Exhibit E. A section from this exhibit is shown below. Note that the Selected Ultimate Losses from the Lag 1 analysis (above) are transferred to this chart in the Lag 1 column.

Quarterly
Rec w/n Acc
Periods	ULTIMATE LOSSES QUARTERLY LAG 1-6 EMERGENCE
Ending	1		2		3		4		5		6
Jun-01		320,368		77,327		52,106		13,905		23,755		16,108
Sep-01		372,461		34,944		53,028		29,579		8,235
Dec-01		524,747		40,514		48,554		33,479
Mar-02		414,268		106,068		54,175
Jun-02		508,232		61,886
Sep-02		527,186

Step 3: Ultimate losses are “Inflated” to the cost level of the last diagonal using the selected pure premium trend. We have selected an annual pure premium trend of +4.0 percent. This is based upon judgment, considering the historical severity and frequency trends for this segment. This is done because our objective is to estimate the required IBNR Reserves as of the current date. Therefore, the losses must be at cost levels that are consistent with the present value of money and are relevant for each respective record period. The following chart is from the bottom of page 2 of Exhibit E, and illustrates this point.

Quarterly
Rec w/n Acc	INFLATED ULTIMATE LOSSES QUARTERLY LAG 1-6 EMERGENCE
Periods			(using a +4.0% Pure Premium Trend)
Ending	1		2		3		4		5		6
Jun-01		336,466		80,420		53,661		14,180		23,989		16,108
Sep-01		387,359		35,987		54,078		29,870		8,235
Dec-01		540,412		41,316		49,032		33,479
Mar-02		422,472		107,113		54,175
Jun-02		513,240		61,886
Sep-02		527,186

Page 37

Note that the June 2002 lag 1 inflated value of 513,240 is equal to the projected ultimate loss amount of 508,232 in the previous chart, multiplied by one-quarter of the 4.0 percent trend to bring its monetary value forward to the current cost level:

(508,232) X [(1.04) ^ (1/4)] = 513,240

Step 4: Divide the inflated ultimate loss amounts by the earned exposures (for each respective quarter), to convert them to pure premium amounts.The results are shown on the top of Page 3 of Exhibit E, and summarized on the chart below. This allows us to normalize these values.

Step 5: Select projected pure premiums for each lag period, based upon this summary, as well as judgment. Those results are also shown at the bottom of the following chart.

	Quarterly
	Rec w/n Acc	INFLATED PURE PREMIUMS – QTRLY LAG 1-6
Earned	Periods	(Inflated Ultimate Losses / Earned Exposures)
Exposures	Ending	1		2		3		4		5		6
24,379	Jun-01		13.801		3.299		2.201		0.582		0.984		0.661
24,375	Sep-01		15.892		1.476		2.219		1.225		0.338
23,292	Dec-01		23.202		1.774		2.105		1.437
22,123	Mar-02		19.096		4.842		2.449
22,681	Jun-02		22.629		2.729
25,217	Sep-02		20.906
	Avg Last 8		22.272		3.685		2.708		1.314		0.767		1.614
	Avg Last 4		21.458		2.705		2.243		1.048		0.638		0.732
	Select		21.458		2.705		2.243		1.048		0.638		0.732

Step 6: Inflate the Selected Pure Premiums by the Pure Premium trend(of +4.0 percent) to the record within accident periods for which the Late Reported Claims are expected to be reported.

For example, the selected Pure Premium for Lag 2 is2.705. The accident quarters that have future claims recorded 2 quarters after their occurrence are the accident quarters ending September 2002 and December 2002. All accident periods older than that no longer need IBNR reserves from Lag 2 for the current analysis. This is because those accidents have already been recorded as of the end of December 2002. However, the pure premium of2.705 is at the cost level of December 2002 recorded values. Therefore, this pure premium needs to be inflated to the monetary level that is relevant for each future record period.

The chart at the bottom of Page 3 of Exhibit E shows the results of these calculations. The chart below is a section from that exhibit to illustrate the calculations.

Page 38

               Selected Pure Premiums from chart above:

	21.458		2.705		2.243		1.048		0.638		0.732		0.521
Quarterly
Rec w/n Acc																	Total
Periods	FUTURE PURE PREMIUMS BY QUARTERLY LAG, INFLATED																Future
Ending	1		2		3		4		5		6		7		8 - 27		Pure Prem
Jun-01														0.526		3.476		4.00
Sep-01												0.740		0.532		3.510		4.78
Dec-01										0.645		0.747		0.537		3.545		5.47
Mar-02								1.058		0.651		0.754		0.542		3.580		6.59
Jun-02						2.266		1.069		0.657		0.762		0.548		3.615		8.92
Sep-02				2.732		2.288		1.079		0.664		0.769		0.553		3.651		11.74
Dec-02		21.670		2.759		2.310		1.090		0.670		0.777		0.558		3.687		33.52

The Lag 2 Selected Pure Premium of2.705is inflated by one-quarter of the 4.0 percent trend for accidents that occur in the quarter ending September 2002 (since they will be recorded in the quarter ending March 2003), and by two-quarters (i.e. 1/2 of a year) of the trend for accidents that occur in the quarter ending December 2002 (since they will be recorded in the quarter ending June 2003):

(2.705) X [(1.04) ^ (1/2)] = 2.759

Step 7: For each accident quarter, calculate the Total Future Pure Premium by summing all Lag periods’ future pure premiums.For example, the Total Future Pure Premium for accident quarter ending December 2002 is33.52.This is the sum of the future pure premiums for accidents that occurred during this quarter, but are recorded in future quarters:

Lag 1 = claims recorded in the first quarter of 2003 = future pure premium of 21.670

Lag 2 = claims recorded in the second quarter of 2003 = future pure premium of2.759

and so on, across the Dec-02 row of the chart above.

These Total Future Pure Premiums are then transferred to column (4) of Exhibit D (Summary of Estimated IBNR), in order to calculate the Total Required IBNR Reserves.

Page 39

Section VII: A Case Study — Loss Adjustment Expense Reserve Review

When a claim occurs, the ultimate amount of the loss is not known until final settlement (payment) of that claim. Through the life of the claim, we need to make sure that our loss reserves are adequate for all future payments on that claim. However, we also incur expenses to adjust claims. Costs incurred in this “loss adjustment” process are called “loss adjustment expenses” (LAE). Like loss reserves, we also need to make sure that our loss adjustment expense reserves are adequate for the future payment of these expenses.

There are two major categories of loss adjustment expenses:

• Defense and Cost Containment (DCC) Expenses. These are sometimes referred to as ALAE (Allocated Loss Adjustment Expenses), or more accurately as Legal ALAE. Since 1998, this category includes:

• Defense and litigation-related expenses, whether internal or external

• Medical cost containment

• Other related expenses incurred in defense of claims

• Adjusting Expenses. These are sometimes referred to as ULAE (Unallocated Loss Adjustment Expenses). Since 1998, this category includes:

• Claim adjusting expenses, whether internal or external

• Fees and salaries of those involved in a claim adjusting function

• Other related expenses incurred in determination of coverage

We hold both Case and IBNR Reserves for each expense category. We may revise any or all of the following parameters in order to achieve the desired changes to Case and/or IBNR LAE Reserves for a given segment:

• ReviseCase LAE Reservesby changing:

• Average Reserves, applicable to open claims below theThreshold. (Note that the Threshold for DCC expense reserves is usually $15,000 per claim; however very few case reserve amounts exceed that Threshold. There is no Threshold for Adjusting expense reserves).

• Inflation Factor, applicable to Averages in following months.

• ReviseIBNR LAE Reservesby changing:

• IBNR Factors, applicable to IBNR Loss Reserves

We evaluate the adequacy of our LAE reserve segments at least two times per year. DCC expense reserves are analyzed separately from Adjusting expense reserves. The segments for which we evaluate LAE reserves are similar to those for which we evaluate Loss reserves. However, we have fewer LAE segments than Loss segments. As discussed in a previous section, our segmentation changes over time. For example, for Bodily Injury, we evaluate about 32-35 states separately for losses (with the remaining states grouped into 2 or 3 segments). However, for LAE we evaluate Bodily Injury for about 10-15 states separately (with the remaining states grouped into 2 or 3 segments). We group the remaining states because they are not credible for independent analysis. This is due to lower volume, and resultant volatility of the data.

The segment being reviewed in this Case Study is for a sample state and coverage for personal auto. This review was done during October 2002, using claim expense data as of September 30, 2002.Note that the data in this example is not from any specific segment and any similarity to a specific segment would be coincidental. Also, the selections made are not necessarily the same as those that would be made in an actual review. The results of this case study are also not intended to represent the actual results of the company.However,

Page 40

our intent is to illustrate and discuss many of the issues that we consider during our analysis, in order to make reasonable selections. The calculations involved in the process will also be explained.

The identities for loss reserves are also relevant for LAE reserves, as follows:

[LAE Reserve Adequacy] =
[Carried (or Held) LAE Reserves] –
[Required (or Indicated) LAE Reserves]

[Required (or Indicated) LAE Reserves] =
[Total Indicated (or Selected or Estimated) Ultimate LAE] –
[Total Paid LAE]

Ultimate LAEis derived differently for each of the two major LAE categories. In general, we attempt to determine how these expenses will develop in the future based upon how they have developed in the past. In order to make reasonable selections, we look at several parameters and also consider the business issues that underlie the data.

There are several exhibits that we include in our reviews to summarize our analysis. We also use these exhibits to discuss our reviews with the relevant business units. In this section, we will present and describe the summary exhibits. One summarizes the DCC Expense analysis (Exhibit DCC); the other summarizes the Adjusting Expense analysis (Exhibit ADJ). These are presented on the next two pages. Following that, we provide an overview of each exhibit, as well as sections from the exhibits. The sections are separated to discuss the issues involved in the underlying data, the judgments that need to be made in the selection process, and the calculations. We will also discuss how different components of the analysis relate to each other.

Note that the LAE reserve review for a segment is not done in the same month as the loss reserve review for that same segment. Therefore, when loss projections are used in the LAE review, they are comparable but not equal to the projections from the loss review. Also note that “rounding” in the exhibits, as well as the “order of calculation” may make some of the figures appear slightly out of balance.

Page 41

EXHIBIT DCC

State ABC Auto DCC (ALAE) as of September 30, 2002

ESTIMATED ULTIMATE LEGAL ALAE — ACCIDENT PERIOD ANALYSIS

	(1)			(2)			(3)			(4)		(5)			(6)		(7)		(8)		(9)
Semiannual				Avg.			Paid ALAE			Paid		Indicated			Indicated		Indicated
Accident	Paid ALAE			Paid ALAE			to Paid Loss			Legal ALAE		Ultimate			Ultimate		Ultimate		Earned
Periods	Method			Method			Method			to date		Legal ALAE			Loss		Loss		Premium		Earned
Ending	Ult ($000)			Ult ($000)			Ult ($000)			($000)		($000)			($000)		Counts		($000)		Exposures
Mar-1996		1,101			1,100			1,086			1,123		1,096			35,491		14,653		54,916		569,550
Sep-1996		1,289			1,288			1,273			1,313		1,283			41,356		16,489		61,783		655,755
Mar-1997		1,854			1,853			1,838			1,880		1,848			52,466		19,047		72,796		771,143
Sep-1997		1,946			1,945			2,156			1,950		2,016			63,887		20,316		82,704		878,255
Mar-1998		1,670			1,670			1,672			1,617		1,671			63,949		21,982		87,680		976,034
Sep-1998		1,597			1,597			1,614			1,483		1,603			67,391		23,252		90,664		1,084,940
Mar-1998		1,920			1.921			1,877			1,706		1,906			81,384		27,133		101,491		1.290.850
Sep-1999		2,238			2,290			2,243			1,918		2,273			86,263		28,436		109,072		1.391.208
Mar-2000		2,735			2,783			2,742			2,159		2,753			94,879		28,155		113,434		1.388.546
Sep-2000		3,184			3,239			3,197			2,257		3,207			90,612		26,717		118,967		1.388.550
Mar-2001		3,401			3,460			3,423			2,077		3,428			90,485		26,376		129,931		1,451,354
Sep-2001		3,918			4,011			3,986			1,977		3,964			98,939		28,092		145,179		1,560,979
Mar-2002		4,415			4,456			4,588			1,351		4,486			126,712		31,500		167,889		1,731,315
Sep-2002		4,693			4,812			4,733			254		4,746			130,955		35,803		193,690		1,902,253
Total		36,012			36,422			36,427			23,066		36,280			1,124,770		347,951		1,530,195		17,040,732
Paid ALAE		23,066			23,066			23,066					23,066
		12,946			13,356			13,362		Required Reserves			13,214
		12,136			12,136			12,136		Held Reserve			12,136
		(810	)		(1,220	)		(1,226	)	Reserve Adequacy			(1,078	)

[Additional columns below]

[Continued from above table, first column(s) repeated]

	(10)		(11)			(12)		(13)		(14)			(15)
Semiannual
Accident	Ultimate		Ultimate					Selected		Selected			Ult. Utillz.
Periods	Legal ALAE		Projected			Projected		Legal ALAE		Ultimate			Prior Rev
Ending	Counts		Utilization			Counts		Counts		Utilization			(03/31/02)
Mar-1996		4,772		32.6	%		4,772		4,772		32.6	%		32.6	%
Sep-1996		5,330		32.3	%		5,330		5,330		32.3	%		32.3	%
Mar-1997		6,597		34.6	%		6,598		6,597		34.6	%		34.6	%
Sep-1997		6,622		32.6	%		6,622		6,622		32.6	%		32.6	%
Mar-1998		6,713		30.5	%		6,714		6,714		30.5	%		30.5	%
Sep-1998		6,689		28.8	%		6,689		6,689		28.8	%		28.8	%
Mar-1998		7,373		27.2	%		7,374		7,374		27.2	%		27.5	%
Sep-1999		7,380		26.0	%		7,381		7,381		26.0	%		25.8	%
Mar-2000		6,982		24.8	%		6,983		6,983		24.8	%		24.5	%
Sep-2000		6,723		25.2	%		6,723		6,723		25.2	%		25.8	%
Mar-2001		6,605		25.0	%		6,607		6,606		25.0	%		25.7	%
Sep-2001		6,954		24.8	%		6,954		6,954		24.8	%		25.7	%
Mar-2002		7,831		24.9	%		7,838		7,834		24.9	%		26.2	%
Sep-2002		8,816		24.6	%		8,819		8,818		24.6	%		N/A
Total		95,387					95,406		95,396		27.4	%
Paid ALAE

	(16)			(17)			(18)			(19)			(20)			(21)			(22)			(23)		(24)		(25)
																										Selected Ult
Semiannual										Selected			Selected									Projected		Selected		Legal ALAE
Accident				Average			Paid			Ultimate			Ultimate			Selected			Selected			Average		Ultimate		Severity
Periods	Paid Ult			Paid Ult			to Paid Ult			Legal			Legal			Ultimate			Ultimate			Paid ALAE		Legal ALAE		(Prior Preview)
Ending	ALAE/Loss			ALAE/Loss			ALAE/Loss			ALAE/Loss			ALAE/EP			ULAE/EP			Loss/EP			Severity		Severity		(03/31/02)
Mar-1996		2.36	%		2.36	%		2.31	%		2.34	%		1.88	%					80.19	%		261		258		247
Sep-1996		2.65	%		2.65	%		2.60	%		2.64	%		2.08	%					79.09	%		310		308		312
Mar-1997		2.88	%		2.93	%		2.89	%		2.90	%		2.43	%		10.10	%		83.64	%		398		394		408
Sep-1997		3.51	%		3.58	%		3.53	%		3.54	%		2.70	%		11.28	%		76.17	%		482		477		445
Mar-1998		3.76	%		3.82	%		3.78	%		3.79	%		2.64	%		11.23	%		69.64	%		524		519		462
Sep-1998		3.96	%		4.05	%		4.03	%		4.01	%		2.73	%		9.56	%		68.15	%		577		570		549
Mar-1998		3.48	%		3.52	%		3.62	%		3.54	%		2.67	%		9.41	%		75.47	%		569		573		536
Sep-1999		3.58	%		3.67	%		3.61	%		3.62	%		2.45	%		9.53	%		67.61	%		546		538		N/A

[Additional columns below]

[Continued from above table]

RETROSPECTIVE COMPARISION OF INDIACTED ULTIMATES
(26)	(27)		(28)	(29)		(30)		(31)
Semiannual	Ult. Legal		Semiannual	Ult.Legal		Ult. Legal
Accident	ALAE at		Accident	ALAE at		ALAE at
Periods	03/31/2002		Periods	03/31/2002		09/30/2002		Change
Ending	($000)		Ending	($000)		($000)		($000)
			Prior		11,406		11422		16
Prior		12,265	Sep-1999		2,228		2,273		45
Sep-1999		2,228	Mar-2000		2,611		2,753		143
Mar-2000		2,611	Sep-2000		3,074		3,207		133
Sep-2000		3,074	Mar-2001		3,191		3,428		236
Mar-2001		3,191	Sep-2001		4,094		3,964		(130	)
Sep-2001		4,094	Mar-2002		4,072		4,486		415
Mar-2002		4,072	Sep-2002	NA			4,746	NA
Total		31,534			30,676		31,534		858

Page 42

Exhibit ADJ

State ABC Auto Adjusting (ULAE) as of September 30, 2002

	(1)		(2)		(3)		(4)		(5)		(6)			(7)
									ULAE
									Charged		Ratio of ULAE			Ratio of ULAE
Quarterly	Adjusted						Total		Per ULAE		to Capped			to Uncapped
Calender	ULAE		Paid Loss		Paid Loss		ULAE		Count		Paid Loss			Paid Loss
End Dates	Counts*		Capped		Uncapped		Charged		(4)/(1)		(4)/(2)			(4)/(3)
Dec-99		15,369		36,018,982		36,762,734		5,683,949		370		15.8	%		15.5	%
Mar-00		15,173		35,491,404		36,776,627		5,770,164		380		16.3	%		15.7	%
Jun-00		15,349		40,171,940		41,911,731		6,194,349		404		15.4	%		14.8	%
Sep-00		14,871		39,792,432		41,226,623		7,221,504		486		18.1	%		17.5	%
Dec-00		14,338		36,986,811		38,603,592		7,371,079		514		19.9	%		19.1	%
Mar-01		14,472		36,979,568		38,745,014		7,217,315		499		19.5	%		18.6	%
Jun-01		15,620		39,466,634		41,028,112		6,638,354		425		16.8	%		16.2	%
Sep-01		15,458		41,634,517		45,535,309		7,247,104		469		17.4	%		15.9	%
Dec-01		16,415		41,529,374		43,867,096		7,622,710		464		18.4	%		17.4	%
Mar-02		17,160		44,576,836		47,394,206		8,180,617		477		18.4	%		17.3	%
Jun-02		18,907		47,316,043		50,129,020		8,654,719		458		18.3	%		17.3	%
Sep-02		19,298		53,160,429		56,278,102		9,809,062		508		18.5	%		17.4	%
			Wtd Avg for Oct-99 thru Sep-00							409		16.4	%		15.9	%
			Wtd Avg for Oct-00 thru Sep-01							475		18.4	%		17.4	%
			Wtd Avg for Oct-01 thru Sep-02							477		18.4	%		17.3	%
			Selected @ Sep-02									18.4	%		17.3	%
			Selected @ Mar-02									18.0	%		16.8	%
			Carried ULAE Paid to Paid Ratio												17.2	%

^* Note: Adjusted ULAE Counts is a weighted average of Number Recorded and Number Paid.

	(8)		(9)		(10)			(11)		(12)		(13)			(14)
								Indicated		Carried		Reserve			#ULAE
	Loss Reserves		Incurred ULAE		% Unpaid			ULAE Reserve		ULAE Reserve		Adequacy			Reserves
Capped Case		96,413,019		17,706,870		38	%		6,728,611		6,680,752		(47,859	)		28,058
IBNR Reserve		22,578,552		3,914,139		90	%		3,522,725		3,402,974		(119,751	)
Total		118,991,572		21,621,009					10,251,336		10,083,726		(167,610	)

Page 43

Exhibit DCC – Defense and Cost Containment Expense Analysis

This exhibit summarizes our Analysis of the adequacy of the DCC expense reserves for this segment. It is an Accident Period analysis. Therefore, the claims are sorted and analyzed by accident date. We utilize 6-month accident periods (i.e. accident semesters) for our analysis. Each accident semester represents all claims that have occurred (or are expected to occur) during the 6-month period ending at the end of the designated month (in the left-hand column of the exhibit).

Since this is an Accident Period analysis, it measures the adequacy of our Total DCC Expense Reserves (Case + IBNR). In other words, the estimated ultimate amounts for each accident period include DCC expenses for claims that have already been reported PLUS DCC expenses for claims that have not yet been reported.

The table below, a section from Exhibit DCC summarizes our selection of the Estimated Ultimate DCC expenses by accident semester.

	(1)			(2)			(3)			(4)		(5)
				= (13) X (23)			= (6) X (18)
Semiannual				Avg			Paid ALAE			Paid		Indicated
Accident	Paid ALAE			Paid ALAE			to Paid Loss			Legal ALAE		Ultimate
Periods	Method			Method			Method			to date		Legal ALAE
Ending	Ult ($000)			Ult ($000)			Ult ($000)			($000)		($000)
Mar-1999		1,920			1,921			1,877			1,706		1,906
Sep-1999		2,288			2,290			2,243			1,918		2,273
Mar-2000		2,735			2,783			2,742			2,159		2,753
Sep-2000		3,184			3,239			3,197			2,257		3,207
Mar-2001		3,401			3,460			3,423			2,077		3,428
Sep-2001		3,918			4,011			3,986			1,977		3,964
Mar-2002		4,415			4,456			4,588			1,351		4,486
Sep-2002		4,693			4,812			4,733			254		4,746
Total		36,012			36,422			36,427			23,066		36,280
Paid ALAE		23,066			23,066			23,066					23,066
		12,946			13,356			13,362		Required Reserves			13,214
		12,136			12,136			12,136		Held Reserves			12,136
		(810	)		(1,220	)		(1,226	)	Reserve Adequacy			(1,078	)

Columns 1 through 3 contain three projection methods that we typically use for estimation of the ultimate amount of DCC expenses by accident semester. The Paid and Average Paid Methods (columns (1) and (2)), are similar to the loss projection methods. For thePaid Method(column (1)), we project the Paid DCC expenses to ultimate amount by organizing the historical paid amounts in a triangular format (by accident period and by evaluation period).

For the Average Paid method (column (2)):

Each data point in theAverage Paid ALAE (DCC)development triangle
= [Paid DCC expense dollars] divided by [Paid Count],
where Paid Count = Claim features (closed or open) with DCC payment.

Page 44

We project the Average Paid DCC to ultimate by accident period. This ultimate DCC severity (shown in column (23)) is then multiplied by the Ultimate number of claims that are expected to incur DCC costs (from column (13)). The result (in column (2)) is the Estimated Ultimate DCC expense amount for each accident period. The estimation of ultimate severity and counts is explained later. However, the following chart illustrates this calculation.

					(2)
	(13)		(23)		= (13)*(23)
Semiannual			Projected		Avg
Accident	Selected		Average		Paid ALAE
Periods	Legal ALAE		Paid ALAE		Method
Ending	Counts		Severity		Ult ($000)
Mar-1999		7,374		261		1,921
Sep-1999		7,381		310		2,290
Mar-2000		6,983		398		2,783
Sep-2000		6,723		482		3,239
Mar-2001		6,606		524		3,460
Sep-2001		6,954		577		4,011
Mar-2002		7,834		569		4,456
Sep-2002		8,818		546		4,812

Column (3) is the “Paid ALAE to Paid Loss” or “Paid-to-Paid” method. Similar to other projection methods, this method also organizes the data in a triangular format. However, each data point in the triangle is the ratio of Paid ALAE (DCC) to Paid Loss. We project this ratio to Ultimate for each accident period. This ultimate “Paid-to-Paid” ratio (by accident period) is shown in column (18). This ultimate ratio is then multiplied by the Ultimate Projected Losses (as derived from the loss analysis, and shown here in column (6)) for each respective accident period. The result (in column (3)) is the Estimated Ultimate DCC expense amount for each accident period. There is further discussion of the Paid-to-Paid method below.

In this example, we have given equal weight to each of the three methods in our ultimate selections in column (5). We calculate the required reserves and the reserve adequacy for each of the methods, as well as the selected (indicated) amounts. The results are shown at the bottom of columns (1) through (3), and (5). For this segment, we determined that our DCC expense reserves are inadequate by $1.1 million. As a result of this analysis, we will increase our reserves by changing the case averages and the IBNR factors for the DCC expense category.

When making selections on many of the DCC segments we tend to give a greater weight to the Paid-to-Paid method. The logic of this is that the legal costs for claims tend to be related to their loss costs. Although the losses may develop at a different rate than the expenses, the relationship tends to be somewhat consistent over time.

However, there can be changes in the claim adjustment process that would potentially cause this relationship to change over time. This may be due to changes in the legal/regulatory environment. It may also be due to changes in the company’s loss adjustment process. We discuss these issues with management to better understand the underlying data. We use

Page 45

additional approaches in our projections for segments in which we observe process changes. This is because the historical development may not be relevant for the future.

The following table shows the ALAE/Loss ratios for this segment over the past 8 accident semesters.

(6)		(16)			(17)			(18)			(19)
Indicated	Semiannual	=(1) / (6)			=(2) / (6)			=(3) / (6)			=(5) / (6)
Ultimate	Accident				Average			Paid			Selected
Loss	Periods	Paid Ult			Paid Ult			to Paid Ult			Ultimate
($000)	Ending	ALAE/Loss			ALAE/Loss			ALAE/Loss			ALAE/Loss
81,384	Mar-1999		2.36	%		2.36	%		2.31	%		2.34	%
86,263	Sep-1999		2.65	%		2.65	%		2.60	%		2.64	%
94,879	Mar-2000		2.88	%		2.93	%		2.89	%		2.90	%
90,612	Sep-2000		3.51	%		3.58	%		3.53	%		3.54	%
90,485	Mar-2001		3.76	%		3.82	%		3.78	%		3.79	%
98,939	Sep-2001		3.96	%		4.05	%		4.03	%		4.01	%
126,712	Mar-2002		3.48	%		3.52	%		3.62	%		3.54	%
130,955	Sep-2002		3.58	%		3.67	%		3.61	%		3.62	%

Each of the ALAE/Loss ratios is equal to the
[Ultimate ALAE (DCC) dollars (from each of the methods)] divided by the
[Ultimate Loss dollars (comparable to the loss reserve analysis of this same segment)]

For this segment, the ALAE/Loss ratios have been increasing over the past 4 accident years, but there are not any significant process changes that caused us to adjust the data. In this example, we began spending more on legal expenses to keep our total loss costs lower. This may be due to higher amounts spent on each claim, and/or a higher proportion of claims may be utilizing Legal expenses. These two parameters are discussed below. Note that each additional dollar spent in defense of claims should save the company significantly more than a dollar in reduced loss costs. This in turn tends to keep our rates lower, which benefits our policyholders.

One of the important parameters used in our analysis of DCC expenses is the percentage of claims for which we incur DCC expenses. We refer to this as the “Utilization Ratio.” In other words, Utilization Ratio = [Ultimate DCC Counts] / [Ultimate Claim Counts].

We project our ultimate DCC counts in two different ways. We then make selections of the DCC counts, which give us selected Utilization Ratios. This is illustrated in the following excerpt from Exhibit DCC.

Page 46

(7)		(10)		(11)			(12)		(13)		(14)
Indicated	Semiannual						=(11) X (7)				= (13) / (7)
Ultimate	Accident	Ultimate		Ultimate					Selected		Selected
Loss	Periods	Legal ALAE		Projected			Projected		Legal ALAE		Ultimate
Counts	Ending	Counts		Utilization			Counts		Counts		Utilization
27,133	Mar-1999		7,373		27.2	%		7,374		7,374		27.2	%
28,436	Sep-1999		7,380		26.0	%		7,381		7,381		26.0	%
28,155	Mar-2000		6,982		24.8	%		6,983		6,983		24.8	%
26,717	Sep-2000		6,723		25.2	%		6,723		6,723		25.2	%
26,376	Mar-2001		6,605		25.0	%		6,607		6,606		25.0	%
28,092	Sep-2001		6,954		24.8	%		6,954		6,954		24.8	%
31,500	Mar-2002		7,831		24.9	%		7,838		7,834		24.9	%
35,803	Sep-2002		8,816		24.6	%		8,819		8,818		24.6	%
347,951	Total		95,387					95,406		95,396		27.4	%

Column (7) contains the estimated ultimate Loss Counts. They are comparable to the ultimate counts from our most recent analysis of losses for this segment.

Column (10) contains the estimated ultimate DCC expense counts by accident semester from this analysis. We project the Paid DCC counts to ultimate by organizing the historical paid DCC counts in a triangular format (by accident period and by evaluation period).

Column (11) contains the estimated ultimate Utilization Ratios from this analysis, based upon an analysis of the triangle of Utilization Ratios.

Each data point in the Utilization Ratio development triangle = [Paid DCC expense counts] divided by [Paid Loss Counts].

In Column (12), we calculate the Projected DCC Counts by multiplying the estimated Utilization Ratios in column (11) by the Indicated ultimate Loss Counts in column (7).

The Selected DCC Counts in column (13) are then based upon the projections from the two different methods in columns (10) and (12). For this segment, the two methods produce similar results, so the selection is relatively straightforward.

The Selected Utilization Ratios in column (14) are the Selected Ultimate DCC Counts (column (13)) divided by the Indicated Ultimate Loss Counts (column (7)). We would expect these ratios to be relatively consistent over time. However, internal process changes or changes to the legal or regulatory environment in the state would cause the utilization ratios to change over time. If we observe changes, we need to investigate the reasons for those changes. For this segment, the ratios are somewhat lower than they were in the past. In other words, we have legal expenses for a lower proportion of claims than we had in the past. Our investigation has shown that this is reasonable due to changes in some laws in the state.

We also compare our selected Ultimate Utilization Ratios in column (14) with the selections made in the prior review of this segment (column (15)). For this segment, our recent ratios are somewhat lower than what we selected six months ago. At that time, we did not believe that the

Page 47

law changes would have as significant an impact on the utilization ratio as they have proven to over the past six months.

Even though the utilization ratio has decreased, the average amount that we spend on legal expenses has increased for each claim that has incurred legal expenses. The following excerpt illustrates this.

	(23)		(24)		(25)
Semiannual	Projected		Selected		Select Ult
Accident	Average		Ultimate		Legal ALAE
Periods	Paid ALAE		Legal ALAE		Severity
Ending	Severity		Severity		(Prior Review)
Mar-1999		261		258		247
Sep-1999		310		308		312
Mar-2000		398		394		408
Sep-2000		482		477		445
Mar-2001		524		519		462
Sep-2001		577		570		549
Mar-2002		569		573		536
Sep-2002		546		538		N/A

Column (24) shows that the selected ultimate severities (i.e. average amount of DCC expenses for each claim for which there is DCC expense incurred) have been increasing each period until the most recent one. Also, for several of the more recent periods, the estimate is higher than it was at the prior review (six months ago). Investigation has shown that we purposely spent more to defend against frivolous and/or fraudulent claims, and have been successful in reducing the amount of fraud. The legal changes in the state have been supportive of this cause. We have therefore been able to reduce rates in this state due to lower loss costs.

We also look at the ratios of losses and expenses, in relation to earned premium, to test the reasonableness of our selections, and to determine if there are any trends of which we should be aware. These ratios have been as follows:

	(20)			(21)			(22)
Semiannual	Selected
Accident	Ultimate			Selected			Selected
Periods	Legal			Ultimate			Ultimate
Ending	ALAE/EP			ULAE/EP			Loss/EP
Mar-1999		1.88	%					80.19	%
Sep-1999		2.08	%					79.09	%
Mar-2000		2.43	%		10.10	%		83.64	%
Sep-2000		2.70	%		11.28	%		76.17	%
Mar-2001		2.64	%		11.23	%		69.64	%
Sep-2001		2.73	%		9.56	%		68.15	%
Mar-2002		2.67	%		9.41	%		75.47	%
Sep-2002		2.45	%		9.53	%		67.61	%

Page 48

We generally had improving loss ratios and ULAE (Adjusting expense) ratios, while the Legal ALAE (DCC expense) ratios have been fluctuating for this segment. We determined that these ratios are reasonable. However, more discussion of adjusting expenses will follow.

In columns (26) through (31) we show a retrospective comparison of the selected (indicated) ultimate DCC expense amounts, versus the selections from the prior review. Consistent with the other parameters, as discussed above, the selected amounts have increased since our prior review.

Page 49

Exhibit ADJ –Adjusting Expense Analysis

This exhibit summarizes our Analysis of the adequacy of the Adjusting Expense (ULAE) reserves for this Segment. It is a Calendar Period analysis. We calculate the ratio of Paid Adjusting expenses to Paid losses for each calendar quarter over the past three years. We then estimate what we believe to be the expected ratio going forward. In order to determine the indicated needed Adjusting Reserves, we apply the estimated UNPAID portion to the current amount of Loss Reserves.

The calculation of the Paid-to-Paid ratios is shown in the following excerpt from Exhibit ADJ.

	(2)		(3)		(4)		(6)			(7)
							Ratio of ULAE			Ratio of ULAE
Quarterly	Paid Loss				Total		to Capped			to Uncapped
Calendar	Capped		Paid Loss		ULAE		Paid Loss			Paid Loss
End Dates	@ 50,000		Uncapped		Charged		(4) / (2)			(4) / (3)
Dec-00		36,986,811		38,603,592		7,371,079		19.9	%		19.1	%
Mar-01		36,979,568		38,745,014		7,217,315		19.5	%		18.6	%
Jun-01		39,466,634		41,028,112		6,638,354		16.8	%		16.2	%
Sep-01		41,634,517		45,535,309		7,247,104		17.4	%		15.9	%
Dec-01		41,529,374		43,867,096		7,622,710		18.4	%		17.4	%
Mar-02		44,576,836		47,394,206		8,180,617		18.4	%		17.3	%
Jun-02		47,316,043		50,129,020		8,654,719		18.3	%		17.3	%
Sep-02		53,160,429		56,278,102		9,809,062		18.5	%		17.4	%
	Wtd Avg for Oct-99 thru Sep-00							16.4	%		15.9	%
	Wtd Avg for Oct-00 thru Sep-01							18.4	%		17.4	%
	Wtd Avg for Oct-01 thru Sep-02							18.4	%		17.3	%
	Select @ Sep-02 (current)							18.4	%		17.3	%
	Select @ Mar-02 (prior rev)							18.0	%		16.8	%
	Carried ULAE Paid to Paid Ratio										17.2	%

The Paid Adjusting Expenses in column (4) are used to calculate the ratios to Paid Losses capped at $50,000 per feature, as well as the ratio to Uncapped Paid Losses. We use $50,000 as the basis for the capping of losses, because it is reasonable to assume that staff involvement does not increase proportionally as the size of the claim increases for larger claims.

The capped paid-to-paid ratios in column (6) are used to determine the needed adjusting expense reserve for settling claims that are currently open. The uncapped paid-to-paid ratios in column (7) are used to determine the needed adjusting expense reserve for claims that are not yet reported.

Using the historical ratios, as well as the 12-month-ending averages and the prior selected ratios, we select our estimated paid-to-paid ratios at the bottom of columns (6) and (7). Also note that the paid-to-paid ratio implied by our current carried Adjusting Expense reserves is shown at the very bottom of column (7).

Page 50

For this segment, the ratios have been relatively level over the past two years, although they were lower during the year prior to that. As with other expense categories, we would look at reasons for changes in the ratios over time. The expenses allocated to this category are those related to Adjustment of claims, including fees and salaries of adjusters. Therefore, we would look at changes in claims staffing, claims inventory, claims processing, and loss volume in order to determine reasons for changes in the ratios over time.

We also look at the average amount of Adjusting expense per claim to see if there is a trend in the expense severity. The following excerpt illustrates this.

	(1)		(4)		(5)
					ULAE
					Charged
Quarterly	Adjusted		Total		Per ULAE
Calendar	ULAE		ULAE		Count
End Dates	Counts		Charged		(4) / (1)
Dec-00		14,338		7,371,079		514
Mar-01		14,472		7,217,315		499
Jun-01		15,620		6,638,354		425
Sep-01		15,458		7,247,104		469
Dec-01		16,415		7,622,710		464
Mar-02		17,160		8,180,617		477
Jun-02		18,907		8,654,719		458
Sep-02		19,298		9,809,062		508
Wtd Avg Oct-99 thru Sep-00						409
Wtd Avg Oct-00 thru Sep-01						475
Wtd Avg Oct-01 thru Sep-02						477

The “Adjusted ULAE Counts” used in this calculation are the weighted average of the number recorded and the number paid. The adjusting expense severity for this segment has been fluctuating somewhat, but the average has been relatively level over the past two years.

Columns (8) through (13) illustrate the calculation of our adjusting expense (ULAE) reserve adequacy. Based upon the indications, we would revise our reserves by changing the Case Averages and the IBNR factors for the Adjusting expense category.

	Selected			(8)		(9)		(10)			(11)		(12)		(13)
	Paid/Paid					= (8) X Select					=(9)X(10)				= (12) - (11)
	Ratio					Pd/Pd ratio					Indicated		Carried
	Per (6)							%			ULAE				Reserve
	And (7)			Loss Reserves		Incurred ULAE		Unpaid			Reserve		ULAE Reserve		Adequacy
Capped Case		18.4	%		96,413,019		17,706,870		38	%		6,728,611		6,680,752		(47,859	)
IBNR Reserve		17.3	%		22,578,552		3,914,139		90	%		3,522,725		3,402,974		(119,751	)
Total					118,991,572		21,621,009					10,251,336		10,083,726		(167,610	)

The “Incurred ULAE” in column (9) is the total incurred adjusting expense for adjusting all claims that are currently reserved.

Page 51

For claims that are currently open, this amount equals the
[Selected Capped Paid-to-Paid ratio of 18.4% (column(6)] times the
[current case loss reserves capped at $50,000].

For claims that are not yet reported, it equals the
[Selected Uncapped Paid-to-Paid ratio of 17.3% (column (7)] times the
[current IBNR loss reserves].

The “% Unpaid” in column (10) have been determined from our time tracking studies. The underlying concept is that much of the adjusting expenses are incurred when the claim is opened (or before it is opened), and the remainder is incurred regularly during the life of the claim until it is settled.

The “Indicated ULAE Reserve” in column (11) represents the unpaid portion of the Incurred ULAE. This is compared to the Carried reserve in column (12) in order to determine the Adequacy of the Adjusting expense reserves in column (13).

Page 52