SEC Filing | White Mountains Insurance Group

View:

David T. Foy
Executive Vice President and
Chief Financial Officer

July 5, 2006

Mr. Jim B. Rosenberg
Senior Assistant Chief Accountant
Securities and Exchange Commission
Division of Corporation Finance
450 Fifth Street, N.W.
Washington, D.C. 20549

Re: White Mountains Insurance Group, Ltd. (the “Company”)
Form 10-K for Fiscal Year Ended December 31, 2005
Filed on March 7, 2006
File No. 001-08993

Dear Mr. Rosenberg,

I refer to your letter dated May 24, 2006 (the “Comment Letter”). In connection with the Company’s response, we acknowledge the following:

· The Company is responsible for the adequacy and accuracy of the disclosure in our filings;

· staff comments or changes to disclosure in response to staff comments do not foreclose the Commission from taking any action with respect to our filings; and

· the Company may not assert staff comments as a defense in any proceeding initiated by the Commission or any person under federal securities laws of the United States.

Set forth below are the Company’s responses to your comments contained in the Comment Letter. The italicized and numbered paragraphs and headings below correspond to those set forth in the Comment Letter.

Item 1, Business, page 2

OneBeacon Loss and LAE table, page 14

White Mountains Re Loss and LAE, page 23

1. We believe that for the reconciliations of net reserve amounts to the gross reserve amounts in the table to be compliant with Industry Guide 6, it is necessary to include a line item that arrives at the cumulative deficiency (redundancy) on a gross basis (i.e., the difference between the gross reserve and the gross re-estimated reserve) for each year presented. Please confirm to us that you will include this line item in your Form 10-K in future filings.

We confirm that we will include a line item that arrives at a gross cumulative deficiency (redundancy) in our future Form 10-K filings.

Item 7. Management’s Discussion and Analysis of Financial Condition and Results of Operations, page 43

Results of Operations for the Years Ended December 31, 2005, 2004 and 2003, page 43

2. We note that you have adjusted comprehensive net income by excluding the change in net unrealized gains from Symetra’s fixed maturity portfolio because the liabilities related to the life insurance and structured settlement products that these assets support are not marked-to-market. Such measure more likely would be permissible if management reasonably believes it is probable that the financial impact of the item will disappear or become immaterial within a near-term finite period. As such, please cease presenting and discussing this measure or demonstrate to us supplementally why this measure is not prohibited by Item 10(e) of Regulation S-K, how your disclosure complies with that regulation, why this measure is useful to investors and how management uses the measure.

Adjusted comprehensive net income is a useful measure to investors and we believe that our use of it complies with Item 10(e) of Regulation S-K. We have addressed each of your questions in separate sections below:

Usefulness to investors

Due to the economics of Symetra’s life insurance and structured settlements businesses, which differ from White Mountains’ property/casualty (re)insurance businesses, “adjusted comprehensive net income” is a more useful measure to our investors of White Mountains’ overall performance than the GAAP measure of “comprehensive net income”.

One of White Mountains’ core operating principles for its property/casualty (re)insurance businesses is to invest for total return. Thus, for those businesses, a dollar of unrealized gains is just as valuable as a dollar of realized gains or net investment income. This is why, prior to the acquisition of our Symetra stake, we had focused on comprehensive net income.

Unlike the total return philosophy of our property/casualty (re)insurance businesses, Symetra attempts to earn a “spread” between what it earns on its investments and what it pays out on its products. In order to try to fix this spread, Symetra invests in a manner that tries to match the duration and cash flows of its investments with the required cash outflows associated with its life insurance and structured settlements products. As a result, Symetra typically earns the same spread on in-force business whether interest rates fall or rise. In the calculation of comprehensive net income under GAAP, fixed maturity investments are marked-to-market while the liabilities to which those assets are matched are not. Thus, we believe that looking to comprehensive net income would overstate or understate the total economic result at White Mountains. Accordingly, following the acquisition of our stake in Symetra, we believe adjusted comprehensive net income is a more useful measure as it excludes the unrealized gains or losses associated with Symetra’s fixed maturity investments.

Further, as we note in our description to investors of why we believe “adjusted comprehensive net income” is a more useful measure, our view is that the intrinsic value of Symetra’s business declines when interest rates decline over the long term — the opposite indication from what an increase in comprehensive net income may tell an investor (and vice

versa when rates rise). At any given time, some of Symetra’s structured settlement obligations may extend 40 or 50 years into the future, which is further out than the longest maturing fixed maturity investments regularly available for purchase in the market (typically 30 years). For these long-dated products, Symetra is unable to fully match the obligation with assets until the remaining expected payout schedule comes within the duration of securities available in the market. If at that time, these fixed maturity investments have yields that are lower than the yields expected when the structured settlement product was originally priced, the spread for the product will shrink and Symetra will ultimately harvest lower returns for its shareholders. GAAP comprehensive net income increases when rates decline, which would suggest an increase in the value of Symetra — the opposite of what is happening to the intrinsic value of the business.

Management’s use of the measure

As discussed above, prior to acquiring our stake in Symetra, we principally looked to comprehensive net income for measuring White Mountains’ overall performance. Following the acquisition, we now principally look to adjusted comprehensive net income in assessing our performance. We report this measure in our quarterly reports to our Board of Directors. We also use this measure to calculate White Mountains’ performance for both our short-term (annual bonus) and long-term incentive plans.

Compliance with Item 10(e) of Regulation S-K

We believe that the use of adjusted comprehensive net income as a measure of performance is not prohibited by Item 10(e) Regulation S-K for the following reasons:

· the presentation of comprehensive net income, the most directly comparable GAAP financial measure, is presented with equal prominence;

· adjusted comprehensive net income is reconciled to comprehensive net income for all periods presented, as the change in unrealized gains (losses) from Symetra’s fixed maturity portfolio is clearly labeled and identified;

· We have disclosed the reasons why we believe that adjusted comprehensive net income is useful information to investors on page 56;

· adjusted comprehensive net income is not a non-GAAP liquidity measure;

· adjusted comprehensive net income does not eliminate or smooth items identified as non-recurring, infrequent or unusual;

· adjusted comprehensive net income is not included on the face of, or in the accompanying notes to, our GAAP financial statements or on the face of any pro forma financial information required to be disclosed by Article 11 of Regulation S-X;

· the titling of the measure as “adjusted comprehensive net income” is a clear indication that it is a variant of the standard GAAP measure, so as not to be confusing.

Critical Accounting Estimates, page 65

1. Loss and Loss Adjustment Expenses, pages 65-75

3. We believe your disclosure regarding the estimation of the reserve for loss and loss adjustment expenses could be improved to better explain the judgments and uncertainties surrounding this estimate and the potential impact on your financial statements. We believe in order to meet the principal objectives of MD&A this disclosure should enable the investor to understand: 1) management’s process for establishing the estimate 2) the reasons for changes in the historical

estimate 3) whether, and to what extent, management has adjusted their assumptions used in the most recent estimate for trends or other factors identified from past experience and 4) the potential variability in the most recent estimate and the impact this variability may have on reported results, financial condition and liquidity. Please keep this objective in mind in drafting your responses to comments listed below. Please also consider providing any additional information, in disclosure-type format, to achieve this objective.

· Please disclose the frequency of your procedures for determining management’s best estimate of loss reserves on both an annual and interim reporting basis.

· Recorded loss reserves for property and casualty insurers generally include a provision for uncertainty, when management determines that existing actuarial methodologies and assumptions do not adequately provide for ultimate loss development. Such a provision may be explicit (i.e., represented by the excess of recorded reserves over actuarial indications) or implicit in the assumption-setting process. So that investors can better understand the inherent uncertainties in your business and degree of conservatism that you have incorporated in your loss reserve estimates, please address as part of your discussion of the methodologies used how you determine the provision for uncertainty.

· We acknowledge the discussion of the key assumptions that materially affect the estimate of the reserves included on page 65 for OneBeacon and on page 74 for White Mountains Re. In addition to your current discussion, please disclose the following:

i. For each of your key assumptions and for your provision for uncertainty, quantify and explain what caused them to change historically over the periods presented.

ii. Discuss whether and to what extent management has adjusted each of the key assumptions and the provision for uncertainty used in calculating the most recent estimate of the reserve given the historical changes, current trends observed and/or other factors as discussed in i. above. This discussion should reconcile the historical changes, the current trends and/or other factors observed to what management has calculated as its most recent key assumptions.

· We acknowledge your tabular presentation on page 71 for OneBeacon and on page 74 for White Mountains Re of the potential variability in the most recent estimate of your loss reserve. Explain why management believes the scenarios quantified are reasonably likely.

We will comply with your request by including additional disclosures in our future Form 10-K filings. An example of the disclosure we intend to include for OneBeacon can be found in Attachment A and for White Mountains Re in Attachment B.

Please feel free to call me at (203) 453-1681 should you have any questions.

Sincerely,

/s/ David T. Foy

ATTACHMENT A

CRITICAL ACCOUNTING ESTIMATES

Management’s Discussion and Analysis of Financial Condition and Results of Operations discusses the Company’s consolidated financial statements, which have been prepared in accordance with GAAP. The financial statements presented herein include all adjustments considered necessary by management to fairly present the financial position, results of operations and cash flows of White Mountains. The preparation of financial statements in conformity with GAAP requires management to make estimates and assumptions that affect the reported amounts of assets and liabilities as of the date of the financial statements and the reported amounts of revenues and expenses during the reporting period.

In the current year presentation of financial information, certain amounts in the prior period financial statements have been reclassified to conform with the current presentation. White Mountains has completed numerous significant transactions during the periods presented that have affected the comparability of the financial statement information presented herein.

On an ongoing basis, management evaluates its estimates, including those related to loss and LAE reserves, purchase accounting, reinsurance estimates and its pension benefit obligations. Management bases its estimates on historical experience and on various other factors that are believed to be reasonable under the circumstances, the results of which form the basis for making judgements about the carrying values of assets and liabilities that are not readily apparent from other sources.

Management believes that its critical accounting policies affect its more significant estimates used in the preparation of its consolidated financial statements. The descriptions below are summarized and have been simplified for clarity.

1. Loss and Loss Adjustment Expenses

OneBeacon

Non-Asbestos and Environmental Reserves

OneBeacon establishes loss and LAE reserves that are estimates of amounts needed to pay claims and related expenses in the future for insured events that have already occurred. The process of estimating reserves involves a considerable degree of judgment by management and, as of any given date, is inherently uncertain.

Reserves are typically comprised of (1) case reserves for claims reported and (2) reserves for losses that have occurred but for which claims have not yet been reported, referred to as incurred but not reported (“IBNR”) reserves, which include a provision for expected future development on case reserves. Case reserves are estimated based on the experience and knowledge of claims staff regarding the nature and potential cost of each claim and are adjusted as additional information becomes known or payments are made. IBNR reserves are derived by subtracting paid loss and LAE and case reserves from estimates of ultimate loss and LAE. Actuaries estimate ultimate loss and LAE using various generally accepted actuarial methods applied to known losses and other relevant information. Like case reserves, IBNR reserves are adjusted as additional information becomes known or payments are made.

Ultimate loss and LAE are generally determined by extrapolation of claim emergence and settlement patterns observed in the past that can reasonably be expected to persist into the future. In forecasting ultimate loss and LAE with respect to any line of business, past experience with respect to that line of business is the primary resource, but cannot be relied upon in isolation. OneBeacon’s own experience, particularly claims development experience, such as trends in case reserves, payments on and closings of claims, as well as changes in business mix and coverage limits, is the most important information for estimating its reserves. External data, available from organizations such as statistical bureaus, consulting firms and reinsurance companies, is sometimes used to supplement or corroborate OneBeacon’s own experience, and can be especially useful for estimating costs of new business. For some lines of business, such as “long-tail” coverages discussed below, claims data reported in the most recent accident year is often too limited to provide a meaningful basis for analysis due to the typical delay in reporting of claims. For this type of business, OneBeacon uses a selected loss ratio method for the initial accident year or years. This is a standard and accepted actuarial reserve estimation method in these circumstances in which the loss ratio is selected based upon information used in pricing policies for that line of business, as well as any publicly available industry data, such as industry pricing, experience and trends, for that line of business.

Uncertainties in estimating ultimate loss and LAE are magnified by the time lag between when a claim actually occurs and when it is reported and settled. This time lag is sometimes referred to as the “claim-tail”. The claim-tail for most property coverages is typically short (usually a few days up to a few months). The claim-tail for liability/casualty coverages, such as automobile liability, general liability, products liability, multiple peril coverage, and workers compensation, can be especially long as claims are often reported and ultimately paid or settled years, even decades, after the related loss events occur. During the long claims reporting and settlement period, additional facts regarding coverages written in prior accident years, as well as about actual claims and trends may become known and, as a result, OneBeacon may adjust its reserves. If management determines that an adjustment is appropriate, the adjustment is booked in the accounting period in which such determination is made in accordance with GAAP. Accordingly, should reserves need to be increased or decreased in the future from amounts currently established, future results of operations would be negatively or positively impacted, respectively.

In determining ultimate loss and LAE, the cost to indemnify claimants, provide needed legal defense and other services for insureds and administer the investigation and adjustment of claims are considered. These claim costs are influenced by many factors that change over time, such as expanded coverage definitions as a result of new court decisions, inflation in costs to repair or replace damaged property, inflation in the cost of medical services and legislated changes in statutory benefits, as well as by the particular, unique facts that pertain to each claim. As a result, the rate at which claims arose in the past and the costs to settle them may not always be representative of what will occur in the future. The factors influencing changes in claim costs are often difficult to isolate or quantify and developments in paid and incurred losses from historical trends are frequently subject to multiple and conflicting interpretations. Changes in coverage terms or claims handling practices may also cause future experience and/or development patterns to vary from the past. A key objective of actuaries in developing estimates of ultimate loss and LAE, and resulting IBNR reserves, is to identify aberrations and systemic changes occurring within historical experience and accurately adjust for them so that the future can be projected reliably. Because of the factors previously discussed, this process requires the use of informed judgment and is inherently uncertain.

OneBeacon’s actuaries use several generally accepted actuarial methods to evaluate its loss reserves, each of which has its own strengths and weaknesses. OneBeacon places more or less reliance on a particular method based on the facts and circumstances at the time the reserve estimates are made. These methods generally fall into one of the following categories or are hybrids of one or more of the following categories:

· Historical paid loss development methods: These methods use historical loss payments over discrete periods of time to estimate future losses. Historical paid loss development methods assume that the ratio of losses paid in one period to losses paid in an earlier period will remain constant. These methods necessarily assume that factors that have affected paid losses in the past, such as inflation or the effects of litigation, will remain constant in the future. Because historical paid loss development methods do not use case reserves to estimate ultimate losses, they can be more reliable than the other methods discussed below that look to case reserves (such as actuarial methods that use incurred losses) in situations where there are significant changes in how case reserves are established by a company’s claims adjusters. However, historical paid loss development methods are more leveraged (meaning that small changes in payments have a larger impact on estimates of ultimate losses) than actuarial methods that use incurred losses because cumulative loss payments take much longer to equal the expected ultimate losses than cumulative incurred amounts. In addition, and for similar reasons, historical paid loss development methods are often slow to react to situations when new or different factors arise than those that have affected paid losses in the past.

· Historical incurred loss development methods: These methods, like historical paid loss development methods, assume that the ratio of losses in one period to losses in an earlier period will remain constant in the future. However, instead of using paid losses, these methods use incurred losses (i.e., the sum of cumulative historical loss payments plus outstanding case reserves) over discrete periods of time to estimate future losses. Historical incurred loss development methods can be preferable to historical paid loss development methods because they explicitly take into account open cases and the claims adjusters’ evaluations of the cost to settle all known claims. However, historical incurred loss development methods necessarily assume that case reserving practices are consistently applied over time. Therefore, when there have been significant changes in how case reserves are established, using incurred loss data to project ultimate losses can be less reliable than other methods.

· Expected loss ratio methods: These methods are based on the assumption that ultimate losses vary proportionately with premiums. Expected loss ratios are typically developed based upon the information used in pricing, and are multiplied by the total amount of premiums written to calculate ultimate losses. Expected loss ratio methods are useful for estimating ultimate losses in the early years of long-tailed lines of business, when little or no paid or incurred loss information is available.

· Adjusted historical paid and incurred loss development methods: These methods take traditional historical paid and incurred loss development methods and adjust them for the estimated impact of changes from the past in factors such as inflation, the speed of claim payments or the adequacy of case reserves. Adjusted historical paid and incurred loss development methods are often more reliable methods of predicting ultimate losses in periods of significant change, provided the actuaries can develop methods to reasonably quantify the impact of changes.

OneBeacon performs an actuarial review of its recorded reserves each quarter. OneBeacon’s actuaries compare the previous quarter’s estimates of paid loss and LAE, case reserves and IBNR to amounts indicated by actual experience. Differences between previous estimates and actual experience are evaluated to determine whether a given actuarial method for estimating losses and LAE should be relied upon to a greater or lesser extent than it had been in the past. While some variance is expected each quarter due to the inherent uncertainty in loss and LAE, persistent or large variances would indicate that prior assumptions and/or reliance on certain reserving methods may need to be revised going forward.

In its selection of recorded reserves, management historically gave greater weight to adjusted paid loss development methods, which are not dependent on the consistency of case reserving practices, over methods that rely on incurred losses. In recent years, the amount of weight given to methods based on incurred losses has increased with management’s confidence that OneBeacon’s case reserving practices have been more consistently applied. However, at this time, management continues to rely more heavily on paid loss development methods over incurred loss development methods when recording reserves.

Upon completion of each quarterly review, OneBeacon’s actuaries select indicated reserve levels based on the results of the actuarial methods described previously, which are the primary consideration in determining management’s best estimate of required reserves. At December 31, 2005 and 2004, the differences between OneBeacon’s total held reserves, which represent management’s best estimate of required reserves, and the actuarially indicated reserve level were insignificant. However, in making its best estimate, management also considers other qualitative factors that may lead to a difference between held reserves and actuarially recommended levels in the future. Typically, these factors exist when management and the company’s actuaries conclude that there is insufficient historical incurred and paid loss information or that trends included in the historical incurred and paid loss information are unlikely to repeat in the future. Such factors include, among others, recent entry into new markets or new products, improvements in the claims department that are expected to lessen future ultimate loss costs and legal and regulatory developments.

Construction Defect Claims

OneBeacon’s general liability and multiple peril lines of business have been significantly impacted by a large number of construction defect claims. Construction defect is a liability allegation relating to defective work performed in the construction of structures such as apartments, condominiums, single family dwellings or other housing, as well as the sale of defective building materials. Such claims seek recovery due to damage caused by alleged deficient construction techniques or workmanship. Much of the recent claims activity has been generated by plaintiffs’ lawyers who approach new homeowners, and in many cases homeowner associations with large numbers of homeowners in multi-residential complexes, about defects or other flaws in their homes. Claims for construction defects began with claims relating to exposures in California. Then, as plaintiffs’ lawyers organized suits in other states with high levels of multi-residential construction, construction defect claims were reported in nearby western states, such as Colorado and Nevada, and eventually throughout the country. The reporting of such claims can be quite delayed as the statute of limitations can be up to ten years. Court decisions have expanded insurers’ exposure to construction defect claims as well. For example, in 1995 California courts adopted a “continuous trigger” theory in which all companies that had ever insured a property that was alleged to have been damaged by defective construction must respond to the claimant, even if evidence of the alleged damage did not appear until after the insurance period had expired. As a result, claims may be reported more than ten years after a project has been completed as litigation can proceed for several years before an insurance company is identified as a potential contributor. Claims have also emerged from parties claiming additional insured status on policies issued to other parties (e.g., such as contractors seeking coverage on a sub-contractor’s policy). Further, in reserving for these claims, there is additional uncertainty due to the potential for further unfavorable judicial rulings and regulatory actions.

A large number of construction defect claims have been identified relating to coverages that OneBeacon had written in the past through Commercial Union and General Accident and their subsidiaries in California, Colorado, Nevada, Washington and Oregon. Management has sought to mitigate future construction defect risks in all states by no longer providing insurance to certain residential general contractors and sub-contractors involved in multi-habitational projects. Mitigating actions also included initiating the withdrawal from problematic sub-segments within OneBeacon’s construction book of business, such as street and road construction, water, sewer and pipeline construction. As a result of these actions, management believes that the number of reported construction defect claims relating to coverages written in the past peaked in 2004 and will continue to decline.

Asbestos and Environmental (“A&E”) Reserves

OneBeacon’s reserves include provisions made for claims that assert damages from A&E related exposures. Asbestos claims relate primarily to injuries asserted by those who came in contact with asbestos or products containing asbestos. Environmental claims relate primarily to pollution and related clean-up cost obligations, particularly as mandated by federal and state environmental protection agencies. In addition to the factors described above under “Non-Asbestos and Environmental Reserves “ regarding the reserving process, OneBeacon estimates its A&E reserves based upon, among other factors, facts surrounding reported cases and exposures to claims, such as policy limits and deductibles, current law, past and projected claim activity and past settlement values for similar claims, as well as analysis of industry studies and events, such as recent settlements and asbestos-related bankruptcies. The cost of administering A&E claims, which is an important factor in estimating loss reserves, tends to be higher than in the case of non-A&E claims due to the higher legal costs typically associated with A&E claims.

A large portion of OneBeacon’s A&E losses resulted from the operations of the Employers Group, an entity acquired by one of the legacy companies in 1971. These operations, including business of Employers Surplus Lines Insurance Company and Employers Liability Assurance Corporation, provided primary and excess liability insurance for commercial insureds, including Fortune 500-sized accounts, some of whom subsequently experienced claims for A&E losses. OneBeacon stopped writing such coverage in 1984.

OneBeacon’s liabilities for A&E losses from business underwritten in the recent past are substantially limited by the application of exclusionary clauses in the policy language that eliminated coverage for such claims. After 1987 for pollution and 1992 for asbestos, most liability policies contained industry-standard absolute exclusions of such claims. In earlier years, various exclusions were also applied, but the wording of those exclusions was less strict and subsequent court rulings have reduced their effectiveness.

OneBeacon also incurred A&E losses via its participation in industry pools and associations. The most significant of these pools was Excess Casualty Reinsurance Association (“ECRA”), which provided excess liability reinsurance to U.S. insurers from 1950 until the early 1980s. ECRA incurred significant liabilities for A&E, of which OneBeacon bears approximately a 4.7% share, or $65 million at both December 31, 2005 and 2004, which is fully reflected in OneBeacon’s loss and LAE reserves.

More recently, since the 1990s, OneBeacon has experienced an influx of claims from commercial insureds, including many non-Fortune 500-sized accounts written during the 1970s and 1980s, who are named as defendants in asbestos lawsuits. As a number of large well-known manufacturers of asbestos and asbestos-containing products have gone into bankruptcy, plaintiffs have sought recoveries from peripheral defendants, such as installers, transporters or sellers of such products, or from owners of premises on which the plaintiffs’ exposure to asbestos allegedly occurred. At December 31, 2005, 592 policyholders had asbestos-related claims against OneBeacon. In 2005, 128 new insureds with such peripheral involvement presented asbestos claims under prior OneBeacon policies.

Historically, most asbestos claims have been asserted as product liability claims. Recently, insureds who have exhausted the available products liability limits of their insurance policies have sought payment for asbestos claims under the premises and operations coverage of their liability policies. It is more difficult for plaintiffs to establish losses as stemming from premises and operations exposures, which requires proof of the defendant’s negligence, rather than products liability under which strict legal liability applies. Hence, there are fewer of such claims and there is a great deal of variation in damages awarded for the actual injuries. Additionally, several accounts that seek such coverage find that previously paid losses exhausted the aggregate limits under their policies. In these situations there is no coverage for these claims. There are currently 243 active claims against OneBeacon without product liability coverage asserting operations or premises coverage.

Immediately prior to White Mountains’ acquisition of OneBeacon, OneBeacon purchased a reinsurance contract with NICO under which OneBeacon is entitled to recover from NICO up to $2.5 billion in the future for asbestos claims arising from business written by OneBeacon in 1992 and prior, environmental claims arising from business written by OneBeacon in 1987 and prior, and certain other exposures. Under the terms of the NICO Cover, NICO receives the economic benefit of reinsurance recoverables from certain of OneBeacon’s third party reinsurers in existence at the time the NICO Cover was executed (“Third Party Recoverables”). As a result, the Third Party Recoverables serve to protect the $2.5 billion limit of NICO coverage for the benefit of OneBeacon. Any amounts uncollectible from third party reinsurers due to dispute or the reinsurers’ financial inability to pay are covered by

NICO under its agreement with OneBeacon. Third Party Recoverables are typically for the amount of loss in excess of a stated level each year. Of claim payments in the past 11 years, approximately 51% of asbestos and environmental losses have been recovered under the historical third party reinsurance.

In June 2005, OneBeacon completed an internal study of its A&E exposures. This study considered, among other items, (1) facts, such as policy limits, deductibles and available third party reinsurance, related to reported claims; (2) current law; (3) past and projected claim activity and past settlement values for similar claims; (4) industry studies and events, such as recent settlements and asbestos-related bankruptcies; and (5) collectibility of third-party reinsurance. Based on the study, OneBeacon increased its best estimate of its incurred losses ceded to NICO, net of underlying reinsurance, by $353 million ($841 million gross) to $2.1 billion, which is within the $2.5 billion coverage provided by the NICO Cover. OneBeacon estimates that the range of reasonable outcomes around its best estimate is $1.7 billion to $2.4 billion, versus a range of $1.5 billion to $2.4 billion from its previous study that was conducted in 2003. Due to the NICO Cover, there was no impact to income or equity from the change in estimate.

The increase in the estimate of incurred A&E losses was principally driven by raised projections for claims related to asbestos (particularly from assumed reinsurance business), and for mass torts other than asbestos and environmental, particularly lead and sexual molestation. The increase was partially offset by reduced projections of ultimate hazardous waste losses.

As noted above, OneBeacon estimates that on an incurred basis it has used approximately $2.1 billion of the coverage provided by NICO at December 31, 2005. Since entering into the NICO Cover, $26 million of the $2.1 billion of utilized coverage relates to uncollected amounts from third party reinsurers through December 31, 2005. Net losses paid totaled approximately $701 million as of December 31, 2005, with $94 million paid in 2005. Asbestos payments during 2005 reflect payments resulting from intensified efforts by claimants to resolve asbestos claims prior to the potential enactment of Federal asbestos legislation. To the extent that OneBeacon’s estimate of ultimate A&E losses as well as the estimate and collectibility of Third Party Recoverables differs from actual experience, the remaining protection under the NICO Cover may be more or less than the approximate $404 million that OneBeacon estimates remained at December 31, 2005.

OneBeacon’s reserves for A&E losses, net of Third Party Recoverables but prior to NICO recoveries, are $1.3 billion at December 31, 2005. An industry benchmark of reserve adequacy is the “survival ratio”, computed as a company’s reserves divided by its historical average yearly loss payments. This ratio indicates approximately how many more years of payments the reserves can support, assuming future yearly payments are equal to historical levels. OneBeacon’s survival ratio was approximately 18.6 at December 31, 2005, which was computed as the ratio of A&E reserves, net of Third Party Recoverables, of $1.3 billion plus the remaining unused portion of the NICO Cover of $404 million, to the average loss payments, net of Third Party Recoverables, in the past three years. The average loss payments used to calculate OneBeacon’s survival ratio were net of a large commutation ($64 million) in 2003 with a third party reinsurer. White Mountains believes that as a result of the NICO Cover and its historical third party reinsurance programs, OneBeacon should not experience material financial loss from old A&E exposures under current coverage interpretations and that its survival ratio compares favorably to industry survival ratios.

OneBeacon’s reserves for A&E losses at December 31, 2005 represent management’s best estimate of its ultimate liability based on information currently available. OneBeacon believes the NICO Cover will be adequate to cover all of its A&E obligations. However, as case law expands, medical and clean-up costs increase and industry settlement practices change, OneBeacon may be subject to asbestos and environmental losses beyond currently estimated amounts. Therefore, OneBeacon cannot guarantee that its A&E loss reserves, plus the remaining coverage under the NICO Cover, will be sufficient to cover additional liability arising from any such unfavorable developments. See Note 3 to the financial statements for more information regarding White Mountains’ A&E reserves.

OneBeacon A&E Claims Activity

OneBeacon’s A&E claim activity for the last two years is illustrated in the table below.

	Year Ended December 31,
A&E Claims Activity	2005		2004
Asbestos
Accounts with asbestos claims at the beginning of the year	664		642
Accounts reporting asbestos claims during the year	128		112
Accounts on which asbestos claims were closed during the year	(200	)	(90	)
Accounts with asbestos claims at the end of the year	592		664

Environmental
Accounts with environmental claims at the beginning of the year	644		674
Accounts reporting environmental claims during the year	180		110
Accounts on which environmental claims were closed during the year	(329	)	(140	)
Accounts with environmental claims at the end of the year	495		644

Total
Total accounts with A&E claims at the beginning of the year	1,308		1,316
Accounts reporting A&E claims during the year	308		222
Accounts on which A&E claims were closed during the year	(529	)	(230	)
Total accounts with A&E claims at the end of the year	1,087		1,308

OneBeacon’s Loss and LAE Reserves by Line of Business

OneBeacon’s net loss and LAE reserves by line of business at December 31, 2005 and 2004 were as follows:

Net loss and LAE reserves by class of business	December 31, 2005						December 31, 2004
($ in millions)	Case		IBNR		Total		Case		IBNR		Total
Workers compensation	$	195.2	$	132.6	$	327.8	$	362.1	$	135.5	$	497.6
Personal automobile liability	445.5		174.7		620.2		530.7		244.1		774.8
Multiple peril	310.4		236.0		546.4		359.3		264.3		623.6
Commercial automobile liability	140.2		65.6		205.8		203.8		82.8		286.6
General liability	106.1		227.2		333.3		121.6		151.1		272.7
Homeowners/Farmowners	81.1		41.4		122.5		82.2		41.8		124.0
Other	115.5		59.9		175.4		97.5		84.0		181.5
Total	$	1,394.0	$	937.4	$	2,331.4	$	1,757.2	$	1,003.6	$	2,760.8

For OneBeacon, the range of reserve estimates at December 31, 2005 was evaluated to consider the strengths and weaknesses of the actuarial methods applied against OneBeacon’s historical claims experience data. The following table shows the recorded reserves and the high and low ends of OneBeacon’s range of reasonable loss reserve estimates at December 31, 2005. The high and low ends of OneBeacon’s range of reserve estimates in the table below are based on the results of various actuarial methods described above. The recorded reserve for each line is the result of the actuarial method that management believes to be most appropriate based on known facts and trends.

OneBeacon net loss and LAE reserves by line of business
Range and recorded reserves

	December 31, 2005
($ in millions)	Low		Recorded		High
Workers compensation	$	291	$	327.8	$	385
Personal automobile liability	560		620.2		659
Multiple peril	499		546.4		658
Commercial automobile liability	191		205.8		231
General liability	282		333.3		372
Homeowners/Farmowners	110		122.5		125
Other	159		175.4		180
Total	$	2,092	$	2,331.4	$	2,610

The recorded reserves represent management’s best estimate of unpaid loss and LAE by line of business. OneBeacon uses the results of several different actuarial methods to develop its estimate of ultimate reserves. While OneBeacon has not determined the statistical probability of actual ultimate paid losses falling within the range, management believes that it is reasonably likely that actual ultimate paid losses will fall within the ranges noted above because the ranges were developed by using several different generally accepted actuarial methods.

The probability that ultimate losses will fall outside of the ranges of estimates by line of business is higher for each line of business individually than it is for the sum of the estimates for all lines taken together due to the effects of diversification. The diversification effects result from the fact that losses across OneBeacon’s different lines of business are not completely correlated. Although management believes OneBeacon’s reserves are reasonably stated, ultimate losses may deviate, perhaps materially, from the recorded reserve amounts and could be above the high end of the range of actuarial projections. This is because ranges are developed based on known events as of the valuation date, whereas the ultimate disposition of losses is subject to the outcome of events and circumstances that may be unknown as of the valuation date.

The following table expresses where OneBeacon’s net recorded loss and LAE reserves for each line of business fall within the range at December 31, 2005 and 2004:

OneBeacon net loss and LAE reserves by line of business
expressed as a percentage of the range

	December 31, 2005		December 31, 2004
Workers compensation	39	%	29	%
Personal automobile liability	61	%	95	%
Multiple peril	30	%	14	%
Commercial automobile liability	37	%	47	%
General liability	57	%	60	%
Homeowners/Farmowners	83	%	95	%
Other	78	%	90	%
Total	46	%	45	%

For some types of claims, such as worker’s compensation, management used forecasting models that consider the unique loss development characteristics of these types of claims. As a result of the trends suggested by these models, management chose a point estimate that was at a higher point in the range at December 31, 2005 as compared to the prior year. Management also selected a point estimate higher in the range for newer and/or growing segments of business, in part based on their view that actuarial methods that rely on historical loss and LAE patterns may have a higher degree of uncertainty for these businesses. As these segments accumulate more historical data, management’s selections place greater reliance on the emerging experience. For personal automobile liability, this resulted in OneBeacon recording reserves at the higher end of the range in 2004, reflecting a more conservative view of emerging favorable loss experience. Management selected a point estimate lower in the range for personal automobile liability in 2005, although still above the mid-point, based on a decision to place greater reliance on the favorable loss trends that had begun to emerge in the prior year. For multiple peril, OneBeacon’s recorded reserves at December 31, 2005 were near the lower end of the range but at a higher point in the range than at the previous year end. The movement within the range for multiple peril was the result of management’s decision to place greater reliance on actuarial estimates that relied on incurred losses than in historical periods, based on the assessment that OneBeacon’s case reserving processes have been more consistently applied in recent periods. For commercial automobile liability, management recorded reserves at December 31, 2005 at a point somewhat lower in the range than previous years as a result of placing greater reliance on actuarial estimates that rely on incurred losses. Similar to the situation noted above for multiple peril, management determined it was appropriate to place greater reliance on an estimate produced by incurred loss methods, based on the assessment that OneBeacon’s case reserving processes have been more consistently applied in recent periods as compared to prior years. For homeowners and “other” (principally shorter tailed lines of business such as ocean and inland marine insurance) recorded reserves remain at the high end of their respective ranges, as management’s selections reflect a conservative approach to recognition of recent favorable incurred loss development patterns.

ATTACHMENT B

White Mountains Re

White Mountains Re A&E Reserves

White Mountains Re’s A&E exposure is primarily from reinsurance contracts written between 1974 through 1985 by Folksamerica predecessor companies (MONY Reinsurance and Christiania General). The exposures are mostly higher layer excess of loss treaty and facultative coverages with relatively low limits exposed for each claim. Folksamerica has a specialized unit that handles claims relating to A&E exposures. The issues presented by these types of claims require specialization, expertise and an awareness of the various trends and jurisdictional developments. Net incurred loss activity for asbestos and environmental in the last two years was as follows:

Net incurred loss and LAE activity	December 31,
($ in millions)	2005			2004
Asbestos	$	62.0		$	2.6
Environmental	(3.4		)	.1
Total	$	58.6		$	2.7

In 2004, White Mountains Re experienced an increase in the number and amount of reported asbestos claims, primarily due to increased claim filings from uninjured claimants who may have been exposed to asbestos. This resulted in a change in our assumption regarding the level of projected future asbestos claims to be paid. As a result of this change, during 2005, White Mountains Re completed a detailed, ground up asbestos study on all reported Folksamerica insureds that had over $250,000 of asbestos claims as well as a significant sample of all other insureds with reported asbestos claims of less than $250,000. Comparing estimates generated by the study to Folksamerica exposed limits by underwriting year led to an increase of approximately $50 million in IBNR during the third quarter of 2005.

Folksamerica sets up claim files for each reported claim by each cedent for each individual insured. In many instances, a single claim notification from a cedent could involve several years and layers of coverage resulting in a file being set up for each involvement. Precautionary claim notices are submitted by the ceding companies in order to preserve their right to pursue coverage under the reinsurance contract. Such notices do not contain an incurred loss amount to Folksamerica, accordingly, an open claim file is not established. As of December 31, 2005, White Mountains Re had approximately 1,339 open claim files for asbestos and 750 open claim files for environmental exposures.

The costs associated with administering the underlying A&E claims by White Mountains Re’s clients tend to be higher than non A&E claims due to generally higher legal costs incurred by ceding companies in connection with A&E claims ceded to White Mountains Re under the reinsurance contracts.

White Mountains Re A&E Claims Activity

White Mountains Re’s A&E claim activity for the last two years is illustrated in the table below.

	Year ended December 31,
A&E Claims Activity	2005		2004
Asbestos
Total asbestos claims at the beginning of the year	1,401		1,185
Incoming asbestos claims due to Sirius Acquisition	—		232
Asbestos claims reported during the year	223		292
Asbestos claims closed during the year	(285	)	(308	)
Total asbestos claims at the end of the year	1,339		1,401

Environmental
Total environmental claims at the beginning of the year	900		743
Incoming environmental claims due to Sirius Acquisition	—		220
Environmental claims reported during the year	65		138
Environmental claims closed during the year	(215	)	(201	)
Total environmental claims at the end of the year	750		900

Total
Total A&E claims at the beginning of the year	2,301		1,928
Incoming A&E claims due to Sirius Acquisition	—		452
A&E claims reported during the year	288		430
A&E claims closed during the year	(500	)	(509	)
Total A&E claims at the end of the year	2,089		2,301

Loss and LAE Reserves by Class of Business

White Mountains Re establishes loss and LAE reserves that are estimates of amounts needed to pay claims and related expenses in the future for reinsured events that have already occurred. The estimation of net reinsurance loss and LAE reserves is subject to the same risk as the estimation of insurance loss and LAE reserves. In addition to those risk factors which give rise to inherent uncertainties in establishing insurance loss and LAE reserves, the inherent uncertainties of estimating such reserves are even greater for the reinsurer, due primarily to: (1) the claim-tail for reinsurers being further extended because claims are first reported to the ceding company and then through one or more intermediary insurers or reinsurers, (2) the diversity of loss development patterns among different types of reinsurance treaties or facultative contracts, (3) the necessary reliance on the ceding companies for information regarding reported claims and (4) the differing reserving practices among ceding companies.

As with insurance reserves, the process of estimating reinsurance reserves involves a considerable degree of judgment by management and, as of any given date, is inherently uncertain. Based on the above, such uncertainty may be larger relative to the reserve for a reinsurer compared to an insurance company, and may take a longer time to emerge.

In order to reduce the potential uncertainty of loss reserve estimation, White Mountains Re obtains information from numerous sources to assist in the process. White Mountains Re’s pricing actuaries devote considerable effort to understanding and analyzing a ceding company’s operations and loss history during the underwriting of the business, using a combination of ceding company and industry statistics. Such statistics normally include historical premium and loss data by class of business, individual claim information for larger claims, distributions of insurance limits provided, loss reporting and payment patterns, and rate change history. This analysis is used to project expected loss ratios for each treaty during the upcoming contract period. These expected ultimate loss ratios are aggregated across all treaties and are input directly into the loss reserving process to generate the expected loss ratios that are used to estimate IBNR.

Upon notification of a loss from a ceding company, White Mountains Re establishes case reserves, including LAE reserves, based upon White Mountains Re’s share of the amount of reserves established by the ceding company and our independent evaluation of the loss. In cases where available information indicates that reserves established by the ceding company are inadequate, White Mountains Re establishes case reserves or IBNR in excess of its share of the reserves established by the ceding company. In addition, specific claim information reported by ceding companies or obtained through claim audits can alert us to emerging trends such as changing legal interpretations of coverage and liability, claims from unexpected sources or classes of business, and significant changes in the frequency or severity of individual claims. Such information is often used to supplement estimates of IBNR.

As mentioned above, there can be a considerable time lag from the time a claim is reported to a ceding company to the time it is reported to the reinsurer. The lag can be several years in some cases. This lag can be due to a number of reasons, including the time it takes to investigate a claim, delays associated with the litigation process, the deterioration in a claimant’s physical condition many years after an accident occurs, etc. In its loss reserving process, White Mountains Re assumes that such lags are predictable, on average, over time and therefore the lags are contemplated in the loss reporting patterns used in its actuarial methods. This means that, as a reinsurer, White Mountains Re must rely on such actuarial estimates for a longer period of time after reserves are first estimated than does a primary insurance company.

Backlogs in the recording of assumed reinsurance can also complicate the accuracy of loss reserve estimation. As of December 31, 2005, there were no significant backlogs related to the processing of assumed reinsurance information at White Mountains Re.

White Mountains Re relies heavily on information reported by ceding companies, as discussed above. In order to determine the accuracy and completeness of such information, White Mountains Re’s U.S. underwriters, actuaries, and claims personnel perform regular audits of Folksamerica’s ceding companies. While regular ceding company audits are not customary outside the United States, Sirius International’s staff regularly reviews information from ceding companies for unusual or unexpected results. Any material findings are discussed with the ceding companies. White Mountains Re sometimes encounters situations where we determine that a claim presentation from a ceding company is not in accordance with contract terms. In these situations, White Mountains Re attempts to resolve the dispute directly with the ceding company. Most situations are resolved amicably and without the need

for litigation or arbitration. However, in the infrequent situations where a resolution is not possible, White Mountains Re will vigorously defend our position in such disputes.

White Mountains Re also obtains reinsurance whereby another reinsurer contractually agrees to indemnify White Mountains Re for all or a portion of the reinsurance risks underwritten by White Mountains Re. Such arrangements, where one reinsurer provides reinsurance to another reinsurer, are usually referred to as “retrocessional reinsurance” arrangements. White Mountains Re establishes estimates of amounts recoverable from retrocessional reinsurance in a manner consistent with the loss and LAE liability associated with reinsurance contracts offered to its customers (the “ceding companies”), net of an allowance for uncollectible amounts, if any. Net reinsurance loss reserves represent loss and LAE reserves reduced by retrocessional reinsurance recoverable on unpaid losses.

Although loss and LAE reserves are initially determined based on underwriting and pricing analyses, White Mountains Re regularly reviews the adequacy of its recorded reserves by using a variety of generally accepted actuarial methods, including historical incurred and paid loss development methods. If actual loss activity differs substantially from expectations, an adjustment to recorded reserves may be warranted. As time passes, loss reserve estimates for a given accident year will rely more on actual loss activity and historical patterns than on initial assumptions based on pricing indications.

White Mountains Re’s expected annual loss reporting assumptions are updated once a year, at year end. These assumptions are applied to year-end IBNR to generate expected reported losses for the subsequent year. Interpolation methods are applied to estimate quarterly reported losses, which are then compared to actual reported losses each quarter. Significant differences may result in a change in estimates or a revision in the loss reporting pattern. Expected loss ratios underlying the current accident year are updated quarterly, to reflect new business that is underwritten by the company.

In 2005, White Mountains Re changed its assumptions relating to asbestos reserves, as discussed above in the A&E section. In 2004, White Mountains Re reduced its loss ratio assumptions for property and other short-tailed business written at Sirius in 2002 and 2003, resulting in a decrease of $35 million in loss reserves related to prior years. This was due to improved terms and conditions in reinsurance contracts written after the terrorist attacks of September 11, 2001, which had a greater impact on lowering loss ratios than had been expected in the original loss ratio assumptions. Also in 2004, White Mountains Re increased its estimates of prior year loss reserves by $55 million on long-tailed U.S. casualty contracts written between 1997 and 2001 at Folksamerica and Scandinavian Re. This was due to a higher than expected level of loss reporting, which caused White Mountains Re to raise its expected loss ratio assumptions and to lengthen the expected reporting tail for those contracts.

White Mountains Re’s net loss and LAE reserves by class of business at December 31, 2005 and 2004 were as follows:

Net loss and LAE reserves by class of business	December 31, 2005						December 31, 2004
($ in millions)	Case		IBNR		Total		Case		IBNR		Total
Liability (excluding A&E)	$	829.3	$	782.9	$	1,612.2	$	920.4	$	900.0	$	1,820.4
Property	357.7		342.2		699.9		143.3		318.0		461.3
Accident & Health and Other	233.9		138.4		372.3		279.5		187.5		467.0
A&E	42.0		75.4		117.4		50.2		24.8		75.0
Total	$	1,462.9	$	1,338.9	$	2,801.8	$	1,393.4	$	1,430.3	$	2,823.7

The actuarial methods described above are used to calculate a point estimate of loss and LAE reserves for each company within White Mountains Re. These point estimates are then aggregated to produce an actuarial point estimate for the entire segment. Once a point estimate is established, White Mountains Re’s actuaries estimate loss reserve ranges to measure the sensitivity of the actuarial assumptions used to set the point estimates. These ranges are calculated using similar methods to the point estimate calculation, but with different expected loss ratio and loss reporting pattern assumptions. For the low estimate, more optimistic loss ratios and faster reporting patterns are assumed, while the high estimate uses more conservative loss ratios and slower reporting patterns. These variable assumptions are derived from historical variations in loss ratios and reporting patterns by class and type of business. Due to the inherent difficulties in estimating ultimate A&E exposures, White Mountains Re does not estimate ranges for these reserves.

The actuarial point estimate is management’s primary consideration in determining its best estimate of loss and LAE reserves. In making its best estimate, management also considers other qualitative factors that may lead to a difference between its best estimate of loss and LAE reserves and the actuarial point estimate. Typically, these factors exist when management and the company’s actuaries conclude that there is insufficient historical incurred and paid loss information or that trends included in the historical incurred and paid loss information are unlikely to repeat in the future. These factors may include, among others, changes in the techniques used to assess underwriting risk, more accurate and detailed levels of data submitted with reinsurance applications, the uncertainty of the current reinsurance pricing environment, the level of inflation in loss costs, changes in ceding company reserving practices, and legal and regulatory developments. At December 31, 2005 and 2004, total carried reserves were 0.7% and 1.2% above the actuarial point estimate, respectively.

The following table illustrates White Mountains Re’s recorded net loss and LAE reserves and high and low estimates for those classes of business for which a range is calculated, at December 31, 2005.

Net loss and LAE reserves by class of business	December 31, 2005
($ in millions)	Low		Recorded		High
Liability (excluding A&E)	$	1,490	$	1,612.2	$	1,760
Property	670		699.9		730
Accident & Health and Other	320		372.3		440
Sub-total	$	2,480	$	2,684.4	$	2,930
A&E			117.4
Total			$	2,801.8

The probability that ultimate losses will fall outside of the range of estimates by class of business is higher for each class of business individually than it is for the sum of the estimates for all classes taken together due to the effects of diversification. While White Mountains Re has not determined the statistical probability of actual ultimate losses falling within the range, management believes that it is reasonably likely that actual ultimate losses will fall within the ranges noted above because the ranges were developed by using generally accepted actuarial methods. Although management believes reserves for White Mountains Re are reasonably stated, ultimate losses may deviate, perhaps materially, from the recorded reserve amounts and could be above the high end of the range of actuarial projections.