Prepared and filed by St Ives Financial

We will contact you next week to organize a teleconference with you to confirm that we have suitably addressed the matters set out in your letter. In the meantime if you have any comments or queries on this matter please do not hesitate to contact me on +44 20 7111 7117.

APPENDIX 1

CRITICAL ACCOUNTING POLICIES - TECHNICAL PROVISIONS

Overview

The provision for unearned premiums in respect of general business represents the proportion of premiums written relating to periods of insurance subsequent to the balance sheet date and is calculated principally on a pro rate basis.

The long term (life) business provision is derived from actuarial valuation. For with profits business, the calculation includes explicit allowance for vested bonuses (including those vesting following valuation at the balance sheet date). Implicit allowance is made for future reversionary bonuses through the use of a net premium valuation method employing a reduced valuation rate of interest. No provision is made for terminal bonuses.

We establish property and casualty loss reserves to account for the anticipated ultimate costs of all losses and related loss adjustment expenses (‘‘LAE’’) on losses that have already occurred. We establish reserves for reported losses and LAE, as well as for incurred but not yet reported (‘‘IBNR’’) losses and LAE. Loss reserve estimates are based on known facts and on interpretation of circumstances including our experience with similar cases and historical claims payment trends. We also consider the development of loss payment trends, levels of unpaid claims, judicial decisions and economic conditions.

We use a variety of statistical techniques and a number of different bases to set reserves, depending on the business unit and line of business in question. Our reserving managers consider claims developments separately for each line of business and subdivide certain lines of business by major claim types or sub-classifications of business. Large claims impacting each relevant account are also generally assessed separately, either being reserved at the face value of the loss adjusters’ estimates or projected separately in order to allow for the future development of large claims.

Reserves are calculated gross of any reinsurance recovery, with a separate asset recorded for the reinsurer’s share having due regard to collectability.

Claims provisions relating to long term permanent disability claims in the United States and Scandinavia are included after reflecting interest expected to be earned. In addition, claims are similarly discounted where there is a particularly long period from incident to claims settlement and where there exists a suitable claims payment pattern from which to calculate the discount. In defining those claims with a long period from incident to claims settlement, those categories of claims where the average period of settlement is six years or more from the balance sheet date has been used as a guide. The total discount applied to these reserves was £721 million in 2004 and £662 million in 2003. The income statement impact of this discounting was a £156 million credit in 2004 and a £133 million credit in 2003. Other than with respect to the above reserves, we do not record reserves on a discounted basis.

We analyze claims progressions according to calendar year accident periods, underwriting years and notification years, with allowances for pure IBNR claims affected in the last instance by way of analyses of claims notification delay tables.

The process of estimating loss reserves is subject to a number of internal and external variables such as changes in claims handling procedures, inflation, judicial trends and legislative changes. Many of these items involve considerable uncertainty. Additionally, there may be significant reporting lags between the occurrence of the insured event and the time it is actually reported to us. We continually refine reserve estimates in a regular ongoing process as experience develops and further losses are reported and settled. We reflect adjustments to reserves in the results of the periods in which we make such adjustments. In establishing reserves, we take into account estimated recoveries for salvage and subrogation.

1 of 8

We derive reserves for reported losses and LAE from estimates of future payments that will be made in respect of reported losses, including settlement costs, having regard to our particular experience with the type of risk involved.

Local regulation in some countries also requires the establishment of catastrophe equalization reserves. Such reserves defer a portion of income with respect to a line of business to future periods in which catastrophe losses might occur (as a result of such factors as hail, nuclear incidents, storms, floods and pollution) in that line of business. Catastrophe equalization reserves are required under U.K. statutory requirements to be presented under the heading ‘‘Technical Provisions’’ in our consolidated balance sheet. Catastrophe equalization reserves are not included in the loss development tables and the reconciliation of reserves for losses and LAE presented within “Item 4 – Information on the Company– Property and Casualty Reserves”.

From time to time we supplement our claims and underwriting processes by utilizing third party adjusters, appraisers, engineers, inspectors, other professionals and information sources to assess and settle catastrophe and non-catastrophe related claims.

Because the establishment of loss reserves is an inherently uncertain process, there can be no assurance that ultimate losses and LAE will not exceed existing loss reserves. Actual losses and LAE may deviate, perhaps substantially, from estimates of reserves reflected in our consolidated financial statements.

We believe based on the information currently available to us that overall, our loss reserves as of December 31, 2005 were appropriate.

The following table shows loss and loss adjustment expense reserves by region as of December 31, 2005, net of reinsurance:

	Personal	Commercial	Total
	£’m	£’m	£’m
U.K.	X	X	X
U.S.	X	X	X
Scandinavia	X	X	X
International	X	X	X

Total	X	X	X

Reserving for non-Asbestos reserves:

How non-Asbestos reserves are set

We analyze reserves by line of business within the various regions. A single line of business may be written in one or more of the regions. Case reserves are established on each individual claim and are adjusted as new information becomes known during the course of handling the claim. Lines of business for which claims data (e.g. paid claims and case reserves) emerge over a long period of time are referred to as long-tail lines of business. Lines of business for which claims data emerge more quickly are referred to as short-tail lines of business. Within the Group’s operations a typical short-tail line of business is property. The longest tail lines of business include personal accident in Scandinavia, workers’ compensation in USA, commercial liability, and professional liability.

Our reserving managers regularly review reserves for both current and prior accident years using the most current claim data. These reserve reviews incorporate a variety of actuarial methods and judgments and involve extensive analysis. For most lines of business, a variety of actuarial methods are reviewed and the reserving managers select methods and specific assumptions appropriate for each line of business based on the current circumstances affecting that line of business. These selections incorporate input from claims personnel, underwriters and operating management on reported claims trends and other factors that could affect the reserve estimates.

2 of 8

For short-tail lines of business, development of paid claims and case reserves is reliable and indicative of ultimate losses. The method used is based on the amount of paid and case estimates for reported claims.

For long-tail lines of business, initial development of paid claims and case reserves is less reliable and, accordingly, may not be fully indicative of ultimate losses. At early stages of an accident year’s development relying solely on the pure development of that year may give erroneous indications and additional information such as the prior accident years’ loss ratios and pricing and inflation trends help give a fuller picture in evaluating the likely ultimate cost of claims arising in that accident year. As the maturity of an accident year develops, the credibility of the specific paid and incurred development of that accident year increases and less reliance needs to be placed on other information such as prior accident years’ loss ratios.

As such there is no explicit segregation between early and late stage development. The process is more one of considering the relative importance of different sources of information and methods on the derivation of the estimated ultimate cost of claims for that accident year.

The reserving process is a “continuous” process, in that the estimates of ultimate claims costs are regularly updated to reflect the latest information on the actual paid and incurred activity and the other factors identified previously. The methodology and approach is not necessarily changed but the parameters in the methods and selection process are updated to derive the most appropriate estimate of the likely ultimate cost. In setting the level of reserves to be booked management will take on board the likely future liabilities disclosed by the estimation process as well as consideration of other factors such as the maturity of the accident year, trends observed over the recent past and the level of volatility within a particular line of business.

The reserve movements shown on pages xx-yy have generally arisen from new or improved information becoming available. Where claims development has differed compared with that previously assumed in estimating reserves, appropriate reserving action has then been taken.

[In our 2005 Form 20-F we will include a section on main developments in 2005]

Current trends contributing to reserve uncertainty

We are a multi-line multi-national property and casualty insurer. We are therefore subject to reserve uncertainty stemming from a number of conditions, including but not limited to those noted above, any of which could be material at any point in time for any line of business. Certain issues may become more or less important over time as conditions change. As various market conditions develop, we must assess whether those conditions constitute a long-term trend that should result in a reserving action (i.e. increasing or decreasing the reserve). Below is a discussion of certain market conditions that Company management has observed during 2005.

[In our 2005 Form 20-F, we will include a discussion on current market conditions]

The level of reserves for US workers compensation, at 31 December 2005, is £xxx million. The most important assumption for workers compensation reserves is the loss development factors, which implicitly allows for future inflation trends. The Company has reviewed the historical variation in paid loss patterns. If the paid loss development patterns change by x%, a change that is within historical variation, the estimated reserves would change by £xxx million, in either direction.

Additionally, for U.S. workers compensation we will include the following information in the 2005 Form 20-F filing:

	•	The number of claims pending at each of the last 3 balance sheet dates;
	•	The number of claims paid during each of the last 3 years;
	•	The number of claims reported during each of the last 3 years;
	•	The aggregate amount paid for each of the last three years.

3 of 8

Reserving for Asbestos reserves:

How Asbestos reserves are set

We have exposure to liabilities for asbestos related losses arising from the sale of commercial liability and multi-peril policies prior to 1987. After 1987, policy wordings contained more prescriptive, and in many cases absolute, exclusions for these types of exposure thereby considerably reducing the potential for loss.

In the United States coverages provided under which these liabilities have emerged were in most cases with smaller commercial customers and involved small policy aggregate limits and limits to coverage. We wrote a limited amount of excess coverage in the United States, most of which were excess policies on top of our own primary covers as opposed to primary policies written by other insurers. As a result, to a large extent, we were able to maintain underwriting and policy wording discipline. Excess policies are insurance policies, which provide coverage in excess of the policy limits of another insurance policy, sometimes referred to as the primary policy. In other words, primary policies provide insurance coverage only to a defined limit of liability. Excess policies provide additional coverage beyond this liability limit.

In the UK asbestos liabilities have emerged most frequently in the Employers Liability line of business. This coverage provides unlimited liability to individuals injured in the workplace. The effect of reinsurance purchased against large individual losses is dampened by the often multi-year nature of the liabilities, which means a deductible for each year of exposure applies before losses can be recovered from the reinsurers.

Reserving for Asbestos claims is subject to a range of uncertainties that are generally greater than those presented by other types of claims. These include long reporting delays, unresolved legal issues on policy coverage and the identity of the insureds. As a consequence, traditional loss reserving techniques cannot wholly be relied on and we have employed specialized techniques to determine reserves in a prudent manner using the extensive knowledge of both internal Asbestos experts and external legal and professional advisors.

U.K. risks written in the U.K.

The majority of business in this area is employers’ liability (“EL”) written through U.K. Commercial with a small amount of public and products liability. The underlying method for the estimation of asbestos requirements for U.K. EL depends critically on establishing a distribution of expected deaths from asbestos related disease, which is then adjusted to allow for the delay between claim and death of claimant. Calibration of the resultant distribution to Group experience of reported claims allows an estimate of future numbers of claims against the Group to be produced.

Average claim cost is monitored from claim notifications over time and, from this data, after adjustment for inflation, a view is taken of current average claims cost taking into account evidence of trends etc. The average costs observed will reflect the proportion of claims cost being borne by the Group as a result of current sharing agreements amongst insurers, which in turn reflects the Group’s proportion of claimant exposure periods.

We have used the U.K. Government Health & Safety Executive projections of probable numbers of future deaths from Mesothelioma, published in December 2003, as the basis to determine our U.K. reserves for future expected asbestos cases and applied this to our average cost, which is adjusted to allow for future inflation. The effect of using this data on our reserves is to increase the anticipated number of asbestos claims, which are spread over an extended reporting period.

US risks written in the UK

U.S. asbestos exposure arises from a variety of sources including: London market ‘direct’ business written through Marine operations many years ago, inward reinsurance exposures also written through the London market and from participation in UK aviation pools.

4 of 8

The methods for reserving this segment are very similar to those described in the ‘US Risks written in the US’ section set out below.

U.S. risks written in the U.S.

Claimants include both primary defendants and peripheral defendants. Primary asbestos defendants manufactured and distributed asbestos products and most are expected to exhaust the majority of available insurance coverage.

The peripheral category is a newer group of defendants brought into litigation due to bankruptcies in the primary defendant group. Some may have manufactured, distributed or installed asbestos containing products, but exposure is more limited. Others owned or operated premises where asbestos products were used, giving rise to premises rather than products claims.

In establishing reserves for US asbestos claims we evaluate each insured’s estimated liability for such claims using a ground up approach. We consider a variety of factors, including the State where underlying claims have been brought, past, current and anticipated future claim activity, disease mix, past settlement amounts for similar claims, dismissal rates, allocated claim adjustment expenses, and potential bankruptcy impact.

Having evaluated the insured’s probability for asbestos claims we then evaluate each insured’s insurance program for such claims. We consider each insured’s total available coverage, including policies issued by us. We also consider relevant judicial interpretations of policy language and applicable defenses or determinations.

Once the gross ultimate exposure for indemnity and allocated claim adjustment expenses is determined for each insured by each policy year, we calculate our ceded reinsurance projection based on any applicable facultative or treaty reinsurances.

The approaches described above were last used in 2003 for US/US risks, 2004 for US/UK risks and 2004 for UK/UK risks. Thereafter paid and incurred emergence has been monitored to track development against prior expectations. If emergence differs from expectations reserves may be adjusted accordingly.

Asbestos reserves in all segments contain a high degree of uncertainty. Factors contributing to this higher degree of uncertainty include:

	•	plaintiffs’ expanding theories of liability, compounded by inconsistent court decisions and judicial interpretations;

	•	a few large claims, accompanied by a very large number of small claims or claims made with no subsequent payment, often driven by intensive advertising by lawyers seeking claimants;

	•	the tendency for speculative, inflated and/or unsupported claims to be made to insurers, with the aim of securing a settlement on advantageous terms;

	•	the long delay in reporting claims and exposures, since the onset of illness and disability arising from exposure to harmful conditions may only become apparent many years later, for example, cases of mesothelioma can have a latent period of up to 40 years;

	•	inadequate loss development patterns;

	•	difficult issues of allocation of responsibility among potentially responsible parties and insurers;

5 of 8

	•	complex technical issues that may give rise to delays in notification arising from unresolved legal issues on policy coverage and the identity of the insureds;

	•	the tendency for social trends and factors to influence jury verdicts; and

	•	developments pertaining to the Group’s ability to recover reinsurance for claims of this nature.

The following table shows the relevant reserve strengthening recorded in the period:

	2003	2004	2005

US written in US	xxx	xxx	xxx
US written in UK	xxx	xxx	xxx
UK written in UK	xxx	xxx	xxx

Total	xxx	xxx	xxx

The technical provisions include £xx million (net of discount) for asbestos in the U.K. and U.S. These provisions can be analyzed by where the risks were written and by survival ratio. Survival ratio is an industry standard measure of a company’s reserves expressing recent year claims payments or notifications as a percentage of liabilities.

The factors that led us to increase our reserves for US exposures in 2003 included:

•	Plaintiffs expanding focus to defendants beyond ‘traditional’ asbestos manufacturers and distributors because of

	•	An increase in the number of insureds seeking bankruptcy protection because of asbestos related litigation

	•	Exhaustion of policy limits

•	Plaintiffs drawing defendants in a wide cross section of defendants who previously only had peripheral or secondary involvement in asbestos litigation such as companies who distributed or incorporated asbestos containing parts in their products or who operated premises where asbestos was present.

•	Attempts to reopen or reclassify previously settled claims

•	Filing of claims under the non-aggregate premises or operation sections of general liability policies

•	Assertions by plaintiffs that insurers had a duty of care to protect the public from the dangers of asbestos.

The main factor that led us to increase asbestos reserves for UK exposures was a new study by the UK Health and Safety Executive indicating that the incidence of population mesothelioma deaths was likely to be greater than previously anticipated.

[We will include similar disclosures for 2004 and 2005 if necessary]

6 of 8

The following tables outline the asbestos provisions as at 31 December 2005, 2004 and 2003 analyzed by risk and survival ratio for U.K. and U.S. risks.

	Total	UK Risks	US Risks	US Risks
		written in UK	written in UK	written in US
	£m	£m	£m	£m
As at 31 December 2005
Provisions
Net of reinsurance	x	x	x	x
Net of discount	x	x	x	x
Claims paid	x	x	x	x
Survival Ratios – On Payment
(Gross of discount)
One Year	x	x	x	x
Three Years	x	x	x	x
Survival Ratios – On Notification
(Gross of discount)
One Year	x	x	x	x
Three Years	x	x	x	x

	Total	UK Risks	US Risks	US Risks
		written in UK	written in UK	written in US
	£m	£m	£m	£m
As at 31 December 2004
Provisions
Net of reinsurance	x	x	x	x
Net of discount	x	x	x	x
Claims paid	x	x	x	x
Survival Ratios – On Payment
(Gross of discount)
One Year	x	x	x	x
Three Years	x	x	x	x
Survival Ratios – On Notification
(Gross of discount)
One Year	x	x	x	x
Three Years	x	x	x	x

	Total	UK Risks	US Risks	US Risks
		written in UK	written in UK	written in US
	£m	£m	£m	£m
As at 31 December 2003
Provisions
Net of reinsurance	x	x	x	x
Net of discount	x	x	x	x
Claims paid	x	x	x	x
Survival Ratios – On Payment
(Gross of discount)
One Year	x	x	x	x
Three Years	x	x	x	x
Survival Ratios – On Notification
(Gross of discount)
One Year	x	x	x	x
Three Years	x	x	x	x

We have not included an analysis of numbers of claims, and therefore the average claim amount, as it is not particularly meaningful for asbestos related business because settlements are

7 of 8

generally made with insureds, where a large number of individual claims may be settled simultaneously.

The level of reserves for asbestos, at 31 December 2005, is £xxx million. A key assumption in the estimation of the asbestos reserves is the emergence rate of claims. If the emergence rate should increase by x% the estimated reserves need would increase by £xxx million.

8 of 8

APPENDIX 2

ESTIMATING US ASBESTOS LIABILITIES

An insurer's asbestos-related claim liabilities cannot be projected by traditional actuarial methods due to such factors as the long latency period between asbestos exposure and disease manifestation and the resulting potential for involvement of multiple policy periods for individual claims. Therefore, we relied primarily on an exposure-based analysis, involving a detailed review of individual policy terms and exposures for a sample group of insureds, and then extrapolating the estimated liabilities of the sample group to all insureds.

In addition to our exposure-based analysis, we used two other non-traditional reserving approaches to corroborate our estimates of our asbestos liabilities:

	•	Aggregate loss development factor (LDF) projections, prepared for all tiers combined and separately by tier (see description of tiers below).

	•	A paid loss market share technique, based on comparing our paid losses to those of the industry. Several highly judgmental selections are required in the application of the market share technique, and thus we use it primarily as a reasonableness check on the other two methods described above.

All approaches incorporate a provision for asbestos building cases (i.e., alleged Property Damage (PD) cases) as well as asbestos Bodily Injury (BI) claims.

As noted above, there is considerable uncertainty in the estimation of the ultimate asbestos - related liabilities of an insurance company. For this reason, we incorporated various assumptions in each of the approaches used such that several potential outcomes are produced. At various steps in our analysis process we have labeled the assumptions and resulting outcomes as Scenario 1, Scenario 2, and Scenario 3. The Scenario I and Scenario 3 values form a range of ultimate loss and ALAE estimates based on our selected parameters. However, our actual losses may be ultimately less than the designated Scenario 1 outcome or greater than the designated Scenario 3 outcome because of the uncertainty regarding a number of the assumptions relied upon in the approaches used in our review.

The selected losses are based primarily on the exposure-based analysis with some consideration given to the LDF projections. In the exposure-based analysis we collect detailed claim and policy information for each account in the selected sample. We then use this account-specific information to project our ultimate liability, either using a Third Party asbestos BI model or by directly estimating our ultimate liability based on the account-specific information. The exposure-based analysis is the most comprehensive of the three techniques as it reflects more of our specific claims experience and contract exposure information than the other techniques. Therefore, we gave this technique the greatest weight in our overall selections.

The LDF approach reflects our claim experience on an aggregate basis. Using calendar year payment and reporting patterns derived from Third Party asbestos models, we projected ultimate loss and expense both gross and net of reinsurance.

The market share technique estimates our share of the US insurance industry's net asbestos liabilities based on the company's share of industry asbestos net claim payments in recent years. We generally view the market share technique as the least reliable of the three methods, as it is subject to distortions caused by large or unusual settlements, and because several highly judgmental selections are required in its application.

The discussion below describes the exposure based method on which we place greatest reliance.

Background

The insurance industry faces potentially significant liability from asbestos-related BI claims

1 of 6

and from alleged PD claims related to asbestos products installed in buildings. Asbestos BI and PD claims have typically been filed as products and completed operations claims by policyholders involved in the manufacture, distribution or installation of asbestos-containing materials.

For asbestos-related products claims, the general use of aggregate limits serves to limit an insurance company's exposure from claims presented on this basis to such limit plus expenses (if in addition to limit) less previously settled products liability claims. Also, coverage afforded on a combined single limits (CSL) basis provides only one application of limits to the combined BI and PD claims. These policy provisions make the determination of an upper bound of products liability for a given insured possible, although cumbersome due to the volume of data required.

As the resources of traditional asbestos defendants have been depleted, asbestos plaintiffs have named additional defendants including companies who did not manufacture or distribute asbestos products, but rather owned or operated premises containing asbestos. Thus, these companies have begun filing non-products (i.e., premises) asbestos claims. There is considerable uncertainty regarding the definition of an occurrence for premises claims, the applicable aggregate or occurrence limits, and the possibility of aggregating premises asbestos claims to reach excess layers of insurance or reinsurance. Our analysis generally treats these premises claims similarly to the BI products claims discussed above.

In addition, in an effort to seek further insurance coverage beyond that limited by products aggregates, some insureds with installation activities that have substantially eroded their products coverage are also presenting new asbestos-related claims as non-products operations claims or attempting to reclassify old products claims as such. The extent to which insureds will be successful in obtaining coverage on this basis is currently being determined through negotiation and litigation. Several of our insureds are currently pursuing non-products coverage for such claims.

Tier Assignments

To facilitate our analysis, we have established a five-tier system for grouping asbestos defendants. The first two Tiers are for major defendants, Tiers 3 and 4 are for peripheral defendants, and Tier 5 is for railroads. The first step in our analysis was to review our asbestos claim listing and assign each insured to a Tier. The Tiers are described below:

	Tier 1 -	This is the group of primary asbestos defendants who are expected to exhaust nearly all available insurance coverage.


	Tier 2 -	These are asbestos defendants who became involved in asbestos litigation shortly after the initial Tier 1 defendants.


		Some Tier 2 defendants will exhaust all available insurance coverage. However for others their share of total industry liabilities is relatively small such that insurance coverage in layers above $50 million is not expected to be materially eroded. We split Tier 2 into three categories: Tier 2-High, Tier 2-Medium and Tier 2-Low, according to our evaluation of their potential asbestos liability.


	Tier 3 -	This group includes the newer group of defendants that manufactured, distributed, or installed an asbestos-containing product and are being brought into the asbestos litigation arena due to the bankruptcy of first and second tier companies. Tier 3 defendants include manufacturers and distributors of asbestos-containing products whose exposure is more limited than that of Tiers 1 and 2 due to limited product distribution or limited exposure from encapsulated products.


		We further split Tier 3 into two categories: Tier 3-High and Tier 3-Low. The Tier 3-High list reflects those peripheral defendants with the greatest exposure to asbestos losses. Although these companies are often regional or national players (rather than local) or have been named in a significant number of lawsuits in recent years, their exposure is not great enough to warrant including them in Tier 2. However, their losses are expected to be significantly greater than the remaining Tier 3

2 of 6

		companies, which are now labeled Tier 3-Low.


	Tier 4 -	Tier 4 defendants were not involved in the manufacture or distribution of asbestos containing products, but rather owned or operated a facility where asbestos products were used and third parties (and in some cases employees) were exposed to asbestos on the defendant's premises. Tier 4 defendants are filing premises claims.


	Tier 5 -	Tier 5 includes railroads facing liabilities from exposed workers under FELA. It is our understanding that many settlement agreements have been reached between railroads and their insurance carriers that would facilitate the reserving process. Also, due to the nature of FELA exposures and attorney involvement, we think that the process of claim reporting is advanced for this category of exposures relative to other similar occupational exposures.

The importance of the Tier 3 versus 4 distinction relates to the application of non-products occurrence limits rather than products coverage to the Tier 4 insureds' liabilities. There remains significant uncertainty with regard to the definition of an occurrence for asbestos premises claims. Our estimates are based on the assumption that all asbestos premises claims for an individual defendant will be aggregated and treated as a single occurrence. This assumption results in Tier 4 insureds being treated much like Tier 3 insureds and the distinction between the two categories becomes less important. We note that our asbestos BI model does differentiate between Tier 3 and 4 insureds in certain parameter assumptions such as the expense-to-indemnity ratio (discussed in greater detail below).

Exposure-Based Analysis

Our exposure-based analysis involved the following steps:

	•	Select a sample group of defendants for detailed individual analysis. The ultimate liability for each sample insured was either projected using a Third Party asbestos BI model or directly estimated based on a review of the account information.


	•	Extrapolate the projected liability on the sample insureds to reflect exposures not included in the sample group.


	•	Develop estimates of ultimate liabilities for categories of defendants not contemplated in the extrapolated model results (e.g., Tier 5, non-products).

Given the potential magnitude of Tier 1 and 2 defendants' asbestos liabilities, the relative maturity of their loss experience, and the greater body of information typically available on these major defendants, we included all open Tier 1 and 2 accounts in our sample.

A number of the sample group insureds were analyzed with a Third Party asbestos BI model. The model analysis involves estimating each insured's aggregate asbestos costs from the ground up, allocating the costs across years of coverage, and comparing the losses allocated to each year to our available coverage. To estimate the ground-up costs for each insured, we used claims information, such as the insured's number of filed and settled claims, the insured's indemnity and expense payments and the insured's coverage block dates. While complete defendant-level information was not available for all insureds, a significant amount of information was available. This information was supplemented as necessary with Third Party benchmark distributions, which vary by tier and were created based on industry data.

In addition, we used coverage information for each insured evaluated using the model, such as the effective dates of coverage, policy limits and attachments, aggregate limits, expense treatment, and any coverage exclusions.

Some cases within the sample group were not analyzed using the model. The reasons for this include:

	•	Account has been settled
	•	Limits have been exhausted or are near exhaustion

3 of 6

	•	Coverage was denied because insured could not provide evidence of coverage
	•	Coverage was disclaimed based on absolute asbestos exclusion
	•	Claims are not expected to reach our layer due to high attachment point and low level of claim activity, and defendant-level claim information is not available for modeling
	•	Insured has no BI exposure, only PD
	•	Our liability hinges on outcome of unresolved coverage issues that cannot be readily modeled

Model Analysis

For the cases run through the asbestos BI model, the first step in the calculations is to project the ultimate ground up indemnity and expense for each account. Building on the claims experience to date (either actual or estimated using the model's benchmark parameters by tier), the asbestos model projects future claim filings for each calendar year through 2050. The model then applies trended average indemnity amounts to the future filings, and loads the projected indemnity for expenses (i.e., defense of the insured, not DJ expenses). This produces a ground up ultimate indemnity and expense indication for each insured that is then spread across policy years.

The amounts by policy year are compared to our policies written in each year to give an estimate of our potential liability across all policies relating to an insured account. This process is repeated for 101 trials of the asbestos BI model, where the parameters are varied stochastically in each run.

Future Claim Counts

Future claim counts were estimated by applying a claim reporting pattern to the estimated number of claims filed to date. A simple loss development approach was used for the Tier 3-Low and 4 insureds, while a Bornhuetter-Ferguson type methodology was used for Tier 2 and 3-High insureds.

The claim reporting pattern was determined from a review of a major asbestos manufacturer by researchers in connection with fairness hearings on the asbestos trust fund established by the bankruptcy court. The manufacturer model estimates the size of the population of workers exposed to asbestos from 1940 to 1979 that are still alive to generate claims in the future. The population is moved forward in time using mortality and morbidity assumptions based on epidemiological studies. Claim frequency rates are based on the manufacturer's claim experience.

We adjusted the filing pattern based on malignant claims (i.e., mesothelioma and cancer) to include a substantial provision for non-malignant claims (e.g., asbestosis and pleural injuries). We also adjusted the resulting pattern to reflect the potential impact of the Center for Claims Resolution (CCR) class action filing (referred to as "Georgine") and its denial, and the recent surge in case filings (the majority assumed to be for non-malignancies) that we have observed in the industry. Additionally, we adjusted the pattern applied to Tier 3 and 4 companies to reflect their later entry into the asbestos litigation process relative to Tier 1 and 2 companies.

Indemnity Severities

The model also requires an estimate of the current indemnity severity for each insured. For most of the insureds we were able to provide information regarding ground-up indemnity payments and claims filed, closed, and pending. Based on this information we calculated various indemnity severity indications, as follows:

	•	Cumulative paid indemnity divided by total claims closed

	•	Paid indemnity for a recent period divided by the number of claims closed during the same time period

We generally used the indemnity severity based on the most recent time period, unless we believed that the volume of claims closed in the recent period was not credible.

4 of 6

The model was run with various scenarios regarding trends in future claim severity. Different severity trends were assumed for mesothelioma, lung-cancer, and non-malignant claims; an overall weighted average was then calculated with the weights changing over time to reflect projected changes in the mix of claims by disease over time. The overall trend was then varied in each run of the model to reflect variations in the trends for the individual disease types or variations in the disease mix.

Expense-to-Indemnity Ratios

We estimated expense-to-indemnity ratios to include a provision for expense. We used ground-up information regarding the expense payments of our insured accounts. Based on this data, we calculated various expense-to-indemnity ratios as follows:

	•	Paid expense divided by paid indemnity.

	•	Average expense divided by average indemnity, where the average indemnity was described above and the average expense was calculated as the paid expense divided by the sum of closed and pending claims. The latter calculation recognizes that expenses paid during a time period are not directly related to the claims closed during that time period.

The two calculations of expense-to-indemnity ratios described above were calculated two ways: reflecting cumulative data and based on recent period experience. The expense-to-indemnity ratio used in the model simulations was selected based on the recent period experience, unless we believed that the claim volume in the recent period was not credible.

Additionally, for Tier 3-Low and Tier 4 defendants, the initial expense-to-indemnity ratios are assumed to decline from their current level.

Benchmark Assumptions.

In the few instances where defendant-level claim information for a particular insured was not available, we used benchmark assumptions compiled from industry data. These benchmark distributions vary by tier. Individual insured-specific claim experience assumptions were used for most of the modeled insureds.

Comparison to our Coverage

Where our share of the ground-up loss has already been negotiated for certain insureds and an actual or estimated cost share percentage was provided, we utilized that directly in our calculations to provide an estimate of our share of the ground-up loss and ALAE.

Where a cost share percentage was not available, our liabilities were estimated by allocating losses to years within the insured's coverage block and comparing the allocated losses to our policy limits and attachment points. Within the indicated coverage block, losses are allocated to policy year based on benchmark distributions that vary by tier. These benchmark distributions reflect a continuous trigger coverage scenario. Where we did not have data on the actual coverage block dates for its insureds, we stochastically generated coverage block starting and ending dates based on industry averages by tier.

Expenses on primary and umbrella policies are payable in addition to limits. For excess policies expenses are generally included within the aggregate limits.

Net indications are derived by applying the outward reinsurance program to the gross ultimate losses on a policy-by-policy basis.

Extrapolation of Sample

Since the exposure-based analysis was applied only to a sample of our insured accounts, it was necessary to extrapolate the results to reflect our insured accounts with asbestos claims that were

5 of 6

not included in the sample.

We estimated the percentage of ultimate losses expected for the sample relative to our total exposure on available coverage for known insureds. The percentages are selected considering the number of insured accounts and the paid, case outstanding, and reported losses for the insured accounts.

The extrapolated ultimate losses reflect both the number of insured accounts and the reported losses of the sample relative to all of our identified open accounts. Further, the extrapolation was performed on a more refined basis than summarized in the table above. It was performed separately for various size groups of insured accounts:

	•	Tier 3-High
	•	Tiers 3-Low and 4 with reported losses greater than $5 million
	•	Tiers 3-Low and 4 with reported losses between $1 million and $5 million
	•	Tiers 3-Low and 4 with reported losses between $100,000 and $1 million
	•	Tiers 3-Low and 4 with reported losses less than $100,000

Since all open Tier 1 and 2 accounts were included in our sample group, no extrapolation is needed for these Tiers (or equivalently, the extrapolation factor is 1.00) .

After extrapolating the indications for the sample group to all open Tier 1-4 accounts, we then add the amount of losses paid on closed accounts to the estimates for the open accounts.

Pure IBNR

We also added a loading for IBNR for the emergence of currently unidentified insured accounts.

We include a pure IBNR provision for only Tier 3 and 4 accounts. We did not include a provision for pure IBNR on the Tier 1 and 2 accounts, because all possible exposure for this group has been identified.

To calculate the pure IBNR provision on Tier 3 and 4 accounts, we first sorted all insured accounts to reflect the date an asbestos claim was first reported to us for the insured. Next we calculated the average number of new accounts reported per year in recent years, and judgmentally selected the expected future rate of account reporting per year. We also judgmentally selected the number of years that insured accounts will continue to emerge at the selected rate. For Tier 3-High accounts, we assume that new accounts with asbestos claims will emerge for another 10 to 20 years. For Tier 3 Low and 4 accounts, we assume new accounts with asbestos claims will emerge for another 15 to 25 years. Multiplying the annual rate of future reporting by the number of years produces the expected number of future accounts.

We then multiplied the number of future accounts by the average account severity. The severities are based on the extrapolated model results, adjusted downward to reflect our judgment that future accounts will generally settle for less than accounts that have already been identified. We believe that it will be more difficult to prove the potential culpability (i.e. product identification and exposure to that product as the cause of disease) of asbestos defendants that have not already been drawn into the litigation. In other words, we believe that the plaintiffs bar has already identified those companies with the strongest connection to asbestos for whom they believe they will have the greatest potential success in achieving awards or settlements for their clients.

6 of 6

APPENDIX 3

PROPERTY AND CASUALTY RESERVES

General

We establish property and casualty loss reserves to account for the anticipated ultimate costs of all losses and related loss adjustment expenses (“LAE”) on losses that have already occurred. We establish reserves for reported losses and LAE, as well as for incurred but not yet reported (“IBNR”) losses and LAE. Loss reserve estimates are based on known facts and on interpretation of circumstances including our experience with similar cases and historical claims payment trends. We also consider the development of loss payment trends, levels of unpaid claims, judicial decisions and economic conditions.

We use a variety of statistical techniques and a number of different bases to set reserves, depending on the business unit and line of business in question. Our reserving managers consider claims developments separately for each line of business and subdivide certain lines of business by major claim types or subclassifications of business. Large claims impacting each relevant account are also generally assessed separately, either being reserved at the face value of the loss adjusters’ estimates or projected separately in order to allow for the future development of large claims.

Reserves are calculated gross of any reinsurance recovery, with a separate asset recorded for the reinsurer’s share having due regard to collectability.

We analyze claims progressions according to calendar year accident periods, underwriting years and notification years, with allowances for pure IBNR claims effected in the last instance by way of analyses of claims notification delay tables.

Local regulation in some countries also requires the establishment of catastrophe equalization reserves. Such reserves defer a portion of income with respect to a line of business to future periods in which catastrophe losses might occur (as a result of such factors as hail, nuclear incidents, storms, floods and pollution) in that line of business. Catastrophe equalization reserves are required under U.K. statutory requirements to be presented under the heading “Technical Provisions” in our consolidated balance sheet. Catastrophe equalization reserves are not included in the loss development tables and the reconciliation of reserves for losses and LAE presented below.

1 of 12

In September 2003, we published the results of a review of our net loss reserves carried out by Tillinghast-Towers Perrin (“the Tillinghast Review”). This review found that our loss reserves at June 30, 2003 were within a reasonable range. However, as the loss reserves were below the Tillinghast best estimate and given the continued uncertainty in relation to asbestos and environmental claims and workers’ compensation claims, we increased our net loss reserves by £563 million.

During the fourth quarter 2004 we undertook a detailed review of reserves held by our U.S. business and, in line with others in the industry, identified the need to strengthen our U.S. loss reserves. We consequently strengthened these loss reserves by £160 million. Of the strengthening £95 million related to workers’ compensation business with the balance across a number of lines including specialty and core segments.

We believe based on the information currently available to us that overall, our loss reserves as of December 31, 2004 were appropriate.

Loss Reserve Development

The tables below present changes in the historical property and casualty reserves that we established in 1995 and subsequent accounting years. The top line of the tables shows the estimated reserves for unpaid losses and LAE set up as of each balance sheet date. Each amount in the top line represents the estimated amount of future payments to be made for losses and LAE for losses occurring in that year and in prior years. The upper (paid) portion of the tables presents the cumulative amounts paid through each subsequent year on those losses for which reserves were carried as of each balance sheet date. The lower (reserve re-estimated) portion of the tables shows the re-estimate of the initially recorded reserves as of each succeeding year end, ignoring claims paid, at the latest period end rate of exchange. The estimate changes, as more information becomes known about the actual losses for which the initial reserves were set up and as the rate of exchange changes. The cumulative redundancy/(deficiency) line reflects the cumulative changes in estimate since the initial reserve was established. The cumulative redundancy/(deficiency) is equal to the initial reserve as restated for exchange less the liability re-estimated as of December 31, 2004.

Reserves for losses and LAE are an accumulation of the estimated amounts necessary to settle outstanding claims as of the date for which the reserve is stated. The following data is cumulative and therefore ending balances should not be added since the amount at the end of each calendar year includes activity for both the current and prior years.

As described in “Item 3—Key Information” we changed our accounting policy in 2001 and we now discount provisions for outstanding claims and the related reinsurance recoveries for those categories of claims where there is a particularly long period from occurrence to claims settlement and where there exists a suitable claims payment pattern. The impact of this change in accounting policy was that the claims provisions for asbestos and environmental claims were discounted for the first time in 2001. In 2001, the additional discount related to asbestos and environmental claims amounted to £296 million, with £176 million accounted for as a prior year adjustment. In 2002, the total discount related to asbestos and environmental claims amounted to £368 million. In 2003, the total discount related to asbestos and environmental claims amounted to £377 million. In 2004, the total discount related to asbestos and environmental claims amounted to £487 million. In the loss development tables presented in this document we have omitted the discount of the asbestos and environmental claims provisions, and hence the reserves shown in the tables differ from the reserves in the financial statements.

2 of 12

CONSOLIDATED LOSS DEVELOPMENT—GROSS OF REINSURANCE
(U.K. GAAP BASIS (1))

	As of December 31,

	1996		1997		1998		1999		2000		2001		2002		2003		2004	2005

	(£in millions)

Initial net reserves for unpaid losses and LAE	9,305		9,388		9,391		10,834		10,992		11,229		11,066		10,831		10,731	x

Initial retroceded reserves	2,241		2,240		2,152		3,183		3,476		4,488		4,511		4,275		4,008	x


Initial gross reserves	11,546		11,628		11,543		14,017		14,468		15,717		15,577		15,106		14,739	x

Exchange adjustment (2)	(542	)	(356	)	(285	)	(412	)	(773	)	(933	)	(594	)	(380	)	—	x

As restated for exchange	11,004		11,272		11,258		13,605		13,695		14,784		14,983		14,726		14,739	x

Fair value adjustment to acquired loss reserves							130
Paid (cumulative) as of:
One year later	3,320		3,469		3,564		4,796		5,032		5,554		4,887		3,992			x
Two years later	5,125		5,303		5,562		7,653		8,328		8,726		7,225
Three years later	6,313		6,655		7,119		9,794		10,659		10,669
Four years later	7,251		7,702		8,197		11,379		12,062
Five years later	7,950		8,403		9,106		12,281
Six years later	8,462		9,085		9,594
Seven years later	8,920		9,456
Eight years later	9,198
Nine years later
Reserve re-estimated as of:
One year later	11,147		11,397		11,387		14,430		15,660		16,734		16,089		15,911	(3)		x
Two years later	10,871		11,080		11,617		15,354		16,477		17,400		17,134
Three years later	10,639		11,214		12,064		16,109		17,204		18,240
Four years later	10,739		11,513		12,365		16,562		17,898
Five years later	11,094		11,882		12,602		16,963
Six years later	11,423		11,971		12,864
Seven years later	11,494		12,278
Eight years later	11,805
Nine years later
Cumulative redundancy (deficiency)	(801	)	(1,006	)	(1,606	)	(3,228	)	(4,203	)	(3,456	)	(2,151	)	(1,185	)		x

(1)      As described in “Item 3—Key Information” we changed our accounting policy in 2001 and we now discount provisions for outstanding claims and the related reinsurance recoveries for those categories of claims where there is a particularly long period from occurrence to claims settlement and where there exists a suitable claims payment pattern. The impact of this change in accounting policy was that the claims provisions for asbestos and environmental claims were discounted for the first time in 2001. In the loss development tables presented in this document we have omitted the discount of the asbestos and environmental claims provisions, and hence the reserves shown in the tables differ from the reserves in the financial statements.
(2)      The exchange adjustment is the difference between the initial reserve at the original rate of exchange and the initial reserve at the rate of exchange ruling on December 31, 2004 as per note 6 to our consolidated financial statements.
(3)      Includes an amount of £704 million in relation to reserves held by our U.S. business, which were included in previous years on a net basis.

3 of 12

CONSOLIDATED LOSS DEVELOPMENT—NET OF REINSURANCE
(U.K. GAAP BASIS (1))

	As of December 31,

	1996		1997		1998		1999		2000		2001		2002		2003		2004	2005

	(£ in millions)

Initial net reserves for unpaid losses and LAE	9,305		9,388		9,391		10,834		10,992		11,229		11,066		10,831		10,731	x
Exchange adjustment (2)	(420	)	(270	)	(201	)	(258	)	(485	)	(430	)	(232	)	(229	)	—	x
																		x
As restated for exchange	8,885		9,118		9,190		10,576		10,507		10,799		10,834		10,602		10,731	x

Fair value adjustment to acquired loss reserves							130
Paid (cumulative) as of:
One year later	2,499		2,706		2,863		3,772		4,176		3,875		3,801		2,633			x
Two years later	3,905		4,133		4,588		5,934		6,474		6,263		5,380
Three years later	4,942		5,277		5,735		7,370		8,137		7,466
Four years later	5,652		6,108		6,562		8,417		9,006
Five years later	6,153		6,656		7,157		9,060
Six years later	6,536		7,073		7,501
Seven years later	6,842		7,355
Eight years later	7,055
Nine years later
Reserve re-estimated as of:
One year later	8,897		9,085		9,112		11,117		11,553		11,692		11,758		10,940			x
Two years later	8,709		8,780		9,193		11,605		11,830		12,408		11,929
Three years later	8,501		8,819		9,533		11,856		12,474		12,626
Four years later	8,527		9,121		9,642		12,294		12,752
Five years later	8,800		9,315		9,801		12,528
Six years later	8,974		9,308		10,077
Seven years later	8,977		9,608
Eight years later	9,279
Nine years later
Cumulative redundancy (deficiency)	(394	)	(490	)	(887	)	(1,822	)	(2,245	)	(1,827	)	(1,095	)	(338	)		x

(1) As described in “Item 3—Key Information” we changed our accounting policy in 2001 and we now discount provisions for outstanding claims and the related reinsurance recoveries for those categories of claims where there is a particularly long period from occurrence to claims settlement and where there exists a suitable claims payment pattern. The impact of this change in accounting policy was that the claims provisions for asbestos and environmental claims were discounted for the first time in 2001. In the loss development tables presented in this document we have omitted the discount of the asbestos and environmental claims provisions, and hence the reserves shown in the tables differ from the reserves in the financial statements.
(2) The exchange adjustment is the difference between the initial reserve at the original rate of exchange and the initial reserve at the rate of exchange ruling on December 31, 2004 as per note 6 to our consolidated financial statements.

4 of 12

2004

The adverse development of £338million for 2004 was mainly due to reserve strengthening in the United Kingdom of £174million and in the United States of £156million.

The reserve strengthening of £174million in the U.K. all arose in accident years 1995 and prior and primarily related to U.K. asbestos reserves. In preparing our U.K. asbestos reserve estimates we use U.K. Health and Safety Executive data tables. During 2004 we adopted the latest updated data tables which indicated a higher incidence of asbestos claims but which were spread over a longer period.

The reserve strengthening of £156million in the U.S. primarily related to accident years 2003, 2000 and 1995 and prior. Of the £156million, £95million related to strengthening of reserves for workers compensation. Following the adoption of a more consistent approach to claims handling practices across the group, resulting in more consistent data for analysis, and higher than expected medical inflation, workers compensation claims emergence in 2004 was higher than that previously expected.

2003

The adverse development of £731million in 2003 primarily arose from reserve strengthening in the U.S. of £617million and the U.K. of £75million.

Of the £617million reserve strengthening in the U.S., £491 million arose in accident years 1999 to 2001. The reserve strengthening in the U.S. was mainly due to increases in asbestos, workers compensation and general liability following the completion of a review of loss reserves by Tillinghast - Towers Perrin (“Tillinghast”). The review by Tillinghast established that actual claims emergence across the asbestos, workers compensation and general liability lines of business was running higher than that assumed within our previous loss reserve estimates.

The reserve strengthening of £75million in the U.K. primarily arose in accident years 2000, 1998, and 1995 and prior, which was offset by reserve decreases in 2001 and 2002. The reserve increases primarily related to increases in asbestos reserves following the completion of the review by Tillinghast. The review by Tillinghast established that actual claims emergence on asbestos reserves was running higher than that assumed within our previous loss reserve estimates.

2002

The adverse development of £734million in 2002 primarily arose from reserve strengthening in the U.S. of £406million, in the U.K. of £147million and in International of £121million.

The reserve strengthening of £406million in the U.S. primarily arose in accident years 2001,1999, and 1995 and prior. The reserve strengthening in the U.S. was mainly due to increases in asbestos, workers compensation and auto liability. An asbestos reserve review was undertaken by the Group in 2002 where the latest information projected a higher number of claims than had previously been assumed. Asbestos reserves in the U.S. were increased by £106million (before discounting). The remaining reserve increases of £300million primarily related to workers compensation and auto liability and were largely caused by higher than expected claims inflation.

The reserve strengthening of £147million in the U.K. primarily arose in accident years 1998 to 2001 and 1995 and prior. The reserve increases primarily related to increases in asbestos and auto liability reserves. The increase in asbestos reserves of £101million (before discounting) followed the findings of the internal asbestos review noted above. The increases in auto liability reserves arose after the emergence of higher than expected losses due to an increase in the number of younger drivers being insured.

The reserve strengthening of £121million in International primarily arose in accident years 2001, 1999, and 1995 and prior. The reserve increases primarily related to auto liability reserves and arose due to higher than expected claims inflation.

5 of 12

In 2000, goodwill was increased by an amount that included £130 million as a fair value adjustment to the Orion acquired claims provisions for accident years 1997 to 1999. As this item is not a profit or loss account item for U.K. GAAP, it is shown as an adjustment to the deficiency for the year. For U.S. GAAP purposes, this item is reflected in the income statement. Loss reserve estimates are based on known facts and on interpretation of circumstances including our experience with similar cases and historical claims payment trends. We continually refine reserve estimates in a regular ongoing process as experience develops and further losses are reported and settled. The adjustment was based on an actuarial evaluation undertaken as part of this ongoing process in the fourth quarter of 2000. Claims handling and estimating procedures are inevitably affected by integration activity following an acquisition and this increases uncertainty concerning the adequacy of loss reserves for a period.

A substantial proportion of our property and casualty business, principally personal automobile and household, is short tail and therefore losses are reported and settled relatively quickly. By three years later over 60% of the total claims net of reinsurance initially reserved will generally expected to have been paid and by five years later the expected payment percentage is approximately 75%.

Other than for the reinsurance arrangements with The Chubb Corporation described below, there have been no portfolio transfers significant enough to distort the reserve development tables. The tables presented include the run-off of acquired operations only for those periods subsequent to acquisition. Run-off on the claims reserves of Trygg-Hansa and Orion, both acquired in 1999, were reflected in the loss development tables for 2000 for the first time. A number of non-material acquisitions have been made in the periods covered by the development tables, none of which has significantly affected run-off.

The merger of Royal Insurance and Sun Alliance in 1996 did not bring about any material change in reserving methodology or require additional reserving other than as described below. Some refinement to process took place in the respective actuarial units but on the whole the methodologies used by the two companies produced reserve levels of a consistent strength. It was general practice in both Royal Insurance and Sun Alliance to run a variety of statistical techniques on each reserving population and to select the technique that best approximated the ultimate costs of settling the claim. The selection of method was driven by the nature of the product and the claims experience data available. Methods used included chain ladder, Bornhuetter-Ferguson and average cost per claim. We continue to run a variety of statistical methods on reserving populations. The selection of the actuarial technique adopted has been affected by the nature of the products selected by the Group post-merger, the availability of data following changes to reserving systems and the introduction of consistent case handling and reserving philosophies. We continue to refine both reserving population and method selection with the aim of improving reserving quality. The most significant methodology change after the merger was the use of consulting actuaries in 1999 to model pollution and asbestos risks on our inbound reinsurance book of business covering old U.S. risks.

A reinsurance arrangement between Sun Alliance in the United Kingdom and The Chubb Corporation in the United States was terminated with effect from January 1, 1997. The effect of the change to the reinsurance arrangement was to transfer a portfolio of £175 million of existing net assumed reserves to Chubb. This transfer had no effect on the reserve development redundancies/deficiencies.

Other than for the issues discussed above, we have had no material reserve strengthening or releases, nor have we noted any exceptional trends in our claims development in the nine years preceding 2004.

The loss development tables presented are on a U.K. GAAP basis excluding the impact of discounting on provisions for asbestos and environmental claims and the related reinsurance recoveries. The trends reflected would be materially unchanged if these tables were presented on a U.S. GAAP basis, with the exception of the Orion goodwill adjustment of £130 million discussed above.

The table below reconciles, as of the dates indicated, the gross loss reserve information presented above to the reserves presented in our consolidated financial statements.

6 of 12

CONSOLIDATED RECONCILIATION OF RESERVES FOR LOSSES AND LAE
(U.K. GAAP BASIS)

			As of December 31,

			2003		2004		2005

					(£ in millions)
Net reserves for unpaid losses and LAE at beginning of year			11,066		10,831		x
Reinsurance recoveries/receivables for unpaid losses and LAE at beginning of year			4,511		4,275		x

Gross reserves for losses and LAE at beginning of year			15,577		15,106		x

Effect of changes in foreign exchange rates			(224	)	(380	)	x
Effect of claims portfolio transfer and acquisitions			(915	)	(122	)	x
Incurred related to:

	Current year		6,773		4,867		x
	Prior years		736		1,185		x

		Total incurred losses and LAE	7,509		6,052		x

Paid related to:
	Current year		1,954		1,925		x
	Prior years		4,887		3,992		x

		Total paid losses and LAE	6,841		5,917		x

Gross reserves for losses and LAE at end of year			15,106		14,739		x
Reinsurance recoverable			4,275		4,008		x

Net reserves for losses and LAE at end of year			10,831		10,731		x
Discount on net asbestos and environmental reserves			(377	)	(487	)	x

Net reserves for losses and LAE at end of year as presented in our consolidated financial statements			10,454		10,244		x

Total incurred as a percentage of gross reserves			48.2	%	40.1	%	x
Current year incurred as a percentage of total incurred			90.2	%	80.4	%	x
Prior years incurred as a percentage of total incurred			9.8	%	19.6	%	x
Total paid as a percentage of net reserves			61.8	%	54.6	%	x
Current year paid as a percentage of total paid			28.6	%	32.5	%	x
Prior years paid as a percentage of total paid			71.4	%	67.5	%	x

Included in the “effect of claims portfolio transfer and acquisitions” for 2004 are the adjustments arising from the disposal of our disability business in Denmark and our property and casualty operation in Peru.

Included in the “effect of claims portfolio transfer and acquisitions” for 2003 are the adjustments arising from the disposals of our property and casualty operations in Australia, New Zealand and Puerto Rico.

Included in the “effect of claims portfolio transfer and acquisitions” for 2002 are the adjustments arising from the disposals of our property and casualty operations in Germany, Benelux and Italy.

7 of 12

Reserves for Asbestos and Environmental Losses

The tables below present the changes in the historical asbestos and environmental reserves established by us for 2002 and subsequent accounting years.

ASBESTOS LOSS DEVELOPMENT TABLE—GROSS OF REINSURANCE
(U.K. GAAP BASIS (1))

	As of December 31,

	2003		2004	2005

	(£ in millions)
Initial net reserves for unpaid losses and LAE	975.7		1,021.1	x
Initial retroceded reserves	253.4		326.6	x

Initial gross reserves	1,229.1		1,347.7	x
Exchange adjustment (2)	(48.8	)	—	x

As restated for exchange	1,180.3		1,347.7	x
Paid (cumulative) as of:
One year later	73.8			x
Two years later
Reserve re-estimated as of:
One year later	1,421.4			x
Two years later
Cumulative redundancy/(deficiency)	(241.1	)		x

ASBESTOS LOSS DEVELOPMENT TABLE—NET OF REINSURANCE
(U.K. GAAP BASIS (1))

	As of December 31,

	2003		2004	2005

			(£ in millions)
Initial net reserves for unpaid losses and LAE	975.7		1,021.1	x
Exchange adjustment (2)	(33.2	)	—	x

As restated for exchange	942.5		1,021.1	x
Paid (cumulative) as of:
One year later	72.2			x
Two years later
Reserve re-estimated as of:
One year later	1,093.3			x
Two years later
Cumulative redundancy/(deficiency)	(150.8	)		x


(1) As described in “Item 3—Key Information” we changed our accounting policy in 2001 and we now discount provisions for outstanding claims and the related reinsurance recoveries for those categories of claims where there is a particularly long period from occurrence to claims settlement and where there exists a suitable claims payment pattern. The impact of this change in accounting policy was that the claims provisions for asbestos and environmental claims were discounted for the first time in 2001. In the loss development tables presented in this document we have omitted the discount of the asbestos and environmental claims provisions, and hence the reserves shown in the tables differ from the reserves in the financial statements.
(2) The exchange adjustment is the difference between the initial reserve at the original rate of exchange and the initial reserve at the rate of exchange ruling on December 31, 2004.

8 of 12

ENVIRONMENTAL LOSS DEVELOPMENT TABLE—GROSS OF REINSURANCE
(U.K. GAAP BASIS (1))

	As of December 31,

	2003		2004	2005

			(£ in millions)
Initial net reserves for unpaid losses and LAE	171.6		159.5	x
Initial retroceded reserves	52.1		54.2	x

Initial gross reserves	223.7		213.7	x
Exchange adjustment (2)	(11.7	)	—	x

As restated for exchange	212.0		213.7	x
Paid (cumulative) as of:
One year later	12.9			x
Two years later
Reserve re-estimated as of:
One year later	223.4			x
Two years later
Cumulative redundancy/(deficiency)	(11.4	)		x

ENVIRONMENTAL LOSS DEVELOPMENT TABLE—NET OF REINSURANCE
(U.K. GAAP BASIS (1))

	As of December 31,

	2003		2004	2005

			(£ in millions)
Initial net reserves for unpaid losses and LAE	171.6		159.5	x
Exchange adjustment (2)	(5.6	)	—	x

As restated for exchange	166.0		159.5	x
Paid (cumulative) as of:
One year later	11.6			x
Two years later
Reserve re-estimated as of:
One year later	168.3			x
Two years later
Cumulative redundancy/(deficiency)	(2.3	)		x

(2) The exchange adjustment is the difference between the initial reserve at the original rate of exchange and the initial reserve at the rate of exchange ruling on December 31, 2004.

We have exposure to liabilities for asbestos related and environmental pollution (“A&E”) losses arising from the sale of commercial liability and multi-peril policies prior to 1987. After 1987, policy wordings contained more prescriptive, and in many cases absolute, exclusions for these types of exposure thereby considerably reducing the potential for loss.

Coverages provided under which these liabilities have emerged were in most cases with smaller commercial customers and involved small policy aggregate limits and limits to coverage. We wrote a limited amount of excess coverage in the United States, most of which were excess policies on top of our own primary covers as opposed to primary policies written by other insurers. As a result, to a

9 of 12

large extent, we were able to maintain underwriting and policy wording discipline. Excess policies are insurance policies, which provide coverage in excess of the policy limits of another insurance policy, sometimes referred to as the primary policy. In other words, primary policies provide insurance coverage only to a defined limit of liability. Excess policies provide additional coverage beyond this liability limit.

Reserving for A&E claims is subject to a range of uncertainties that are generally greater than those presented by other types of claims. These include long reporting delays, unresolved legal issues on policy coverage and the identity of the insureds. As a consequence, traditional loss reserving techniques cannot wholly be relied on and we have employed specialized techniques to determine reserves in a prudent manner using the extensive knowledge of both internal A&E experts and external legal and professional advisors.

	Factors contributing to this higher degree of uncertainty include:

	•	plaintiffs’ expanding theories of liability, compounded by inconsistent court decisions and judicial interpretations;

	•	a few large claims, accompanied by a very large number of small claims or claims made with no subsequent payment, often driven by intensive advertising by lawyers seeking claimants;

	•	the tendency for speculative, inflated and/or unsupported claims to be made to insurers, with the aim of securing a settlement on advantageous terms;

	•	the long delay in reporting claims and exposures, since the onset of illness and disability arising from exposure to harmful conditions may only become apparent many years later, for example, cases of mesothelioma can have a latent period of up to 40 years;

	•	inadequate development patterns;

	•	difficult issues of allocation of responsibility among potentially responsible parties and insurers;

	•	complex technical issues that may give rise to delays in notification arising from unresolved legal issues on policy coverage and the identity of the insureds;

	•	the tendency for social trends and factors to influence jury verdicts; and

	•	developments pertaining to the Group’s ability to recover reinsurance for claims of this nature.

The position in the United States is particularly problematic, as plaintiffs have expanded their focus to defendants beyond the ‘traditional’ asbestos manufacturers and distributors. This has arisen as a consequence of the increase in the number of insureds seeking bankruptcy protection because of asbestos-related litigation and the exhaustion of their policy limits. Plaintiffs, supported by lawyers remunerated on a contingent fee basis, are now seeking to draw in a wide cross section of defendants who previously only had peripheral or secondary involvement in asbestos litigation. This may include companies, which have distributed or incorporated asbestos containing parts in their products or operated premises where asbestos was present. There are also increasing signs of attempts to reopen and reclassify into other insurance coverages previously settled claims, and the filing of claims under the non-aggregate premises or operation section of general liability policies. There are also indications that plaintiffs may seek damages by asserting that insurers had a duty to protect the public from the dangers of asbestos.

10 of 12

Against this background and in common with the industry generally, the Group in the United States receives notifications and approaches from, and on behalf of, insureds who previously had peripheral or secondary involvement in asbestos litigation indicating that they may be seeking coverage under Group policies. Given the uncertainties outlined above as to the potential of loss suffered, the availability of coverage and the often long delay in reporting these issues it is difficult to predict the outcome of these notifications and approaches. The greatest difficulty is with estimating whether the Group has any liability as many of these are discharged at no cost to the Group or have been settled below the quantum sought, although there can be no certainty that this will always be the case. It is clear that there is unlikely to be any firm direction in case law or legislation which would allow for these issues to be resolved satisfactorily in the near term and no likelihood of the plaintiffs’ bar in the U.S. easing its aggressive stance with litigation. We, therefore, expect that these notifications and approaches will continue to be received for some time to come. One such approach received during 2004 from General Motors Corporation is now the subject of litigation.

Total net outstanding A&E claims reserves at the end of 2004 amounted to £1,181 million. Net A&E reserves have increased from £1,147 million at the end of 2003, primarily as a result of the increase in asbestos reserves of £151 million before the effect of discounting following reports received from independent actuarial consulting firms offset by settlement of claims and exchange movements. The A&E reserves are mainly in the United Kingdom and the United States (approximately £1,137 million), and to a lesser extent in Canada (approximately £44 million). Amounts recoverable from reinsurers as of December 31, 2004 amounted to £381 million. Unrecoverable amounts, which are not included in the above figure, are not significant and have been fully provided for.

Reserves for environmental liabilities include provision for IBNR claims. The Group has provided for IBNR claims based on modeling performed by both internal experts and external consulting actuaries.

As with other claims reserves, A&E reserves are subject to regular internal review and updating. It is our practice to periodically subject reserves to independent actuarial review and in 1996 following a comprehensive review by an independent actuarial consulting firm, A&E reserves in the United States were strengthened by £117 million and in the United Kingdom by £25 million. Our A&E reserves in the United States were reviewed by independent actuaries in 1999 and our overall level of claims reserves were in the range indicated by the actuaries. Asbestos reserves on inbound reinsurance business in the United Kingdom were reviewed by consulting actuaries in 1999. Resulting from this and other reviews, we released gross reserves of £40 million. The net impact of this release after reinsurance was £6 million. The asbestos reserves were reviewed by independent actuaries in 2001 and, before the impact of discounting the reserves, increased by £200 million in the United Kingdom and by £171 million in the United States. Reviews carried out on asbestos reserves on behalf of the Group during 2002 confirmed that the existing provisions were within the range of reasonable actuarial estimates. They were, however, towards the lower end of that range. In light of this, reserves were strengthened by £101 million before the effect of discounting in the United Kingdom and by £106 million before the effect of discounting in the United States. Asbestos reserves were also strengthened in Australia. U.S. asbestos reserves were further strengthened by £80 million before the effect of discounting following publication of the Tillinghast Report in 2003. United Kingdom asbestos reserves were increased by £179 million before the effect of discounting to move the reserves further up the range of possible outcomes. In 2004 asbestos reserves in the United Kingdom were strengthened by £148 million before the effect of discounting, following publication of revised estimates for future mesothelioma deaths.

In May 2004 we reached agreement in principle with the Administrators of Turner & Newall (“T&N”), an asbestos products manufacturer in administration. This agreement makes available a sum of money towards compensating workers exposed to asbestos by T&N during the years covered by the policies. This should conclude our involvement in any asbestos liabilities of the T&N companies.

Over the past few years, our U.S. operation has been aggressively pursuing A&E claims settlements. This process has involved, where possible, legally enforceable settlement agreements to limit our liabilities. In 2000 the environmental claim payments include the settlement of one large claim for £20 million and in 2001 include one large claim payment of £14 million.

Both the U.S. Senate and the U.S. House of Representatives are considering a bill called the Fairness in Asbestos Injury Resolution Act of 2005. This Bill would establish a privately financed trust

11 of 12

fund to pay people with asbestos-related diseases and an Asbestos Reserves Commission would determine the trust fund payment obligations of insurers. Judicial review would be available in the federal courts, which would remove asbestos claims from the state tort litigation system for the duration of the fund.

As the draft now stands, the trust would be funded with more than $100 billion to provide payments to individuals with asbestos related illnesses. It would be financed through contributions from primary insurers, reinsurers and industrial enterprises. Proposals are still circulating about whether funding would be through lump sum payments, installments or a combination of both. Our share of funding the Trust Fund, should it be established, is necessarily uncertain at this stage.

The draft Bill proposes that the U.S. Department of Labor would administer a no fault trust fund. They will seek to compensate only those individuals who fall sick. Those who may have been exposed to asbestos but show no signs of illness would not benefit from the trust unless they become sick.

Passage of the Bill, in its current or amended form, is far from certain. There are concerns that once the Fund is exhausted claimants would have the right to return to the tort system. The return would be confined to the Federal Court using the same medical conditions as the fund, which are different to the categories applied in the traditional tort system.

12 of 12


Re:	Royal & Sun Alliance Insurance Group plc
	Form 20-F for the fiscal year Ended December 31, 2004
	File Number: 001-15146

1.	We note in your response to prior comment two that you base reserve calculations for long-tail, non-asbestos exposures in the early stage of claim development on prior year accident year loss ratios, changes to pricing mix of business and other factors. You base reserve calculations for such exposures in a later stage of claim development on methods, which incorporate a development pattern assumption. Please expand your proposed disclosure as follows.

	a.	Describe and quantify the difference between early and later stage claims development.
	b.	Describe and quantify the methods and key assumptions used for early stage claims by line of business.
	c.	Describe and quantify the methods and key assumptions used for later stage claims, including the different development pattern assumptions by line of business.
	d.	So that investors can better assess the volatility inherent in your reserve estimates, link your discussion of these expected development patterns to your explanation of changes in prior year reserve estimates as described in your responses to comment four in your October 20, 2005 letter.

2.	Please refer to your response to prior comment two. Your proposed disclosure does not appear to adequately address our comment. We note that you determine asbestos reserves for US risks written in the UK and US risks written in the US on a “ground up” basis. However, you do not discuss the methods and key assumptions used in this “ground up” approach and appear to only provide a general list of factors contributing to uncertainty in this reserve determination process. Please expand your description of the reserve methodology for asbestos reserves for US risks written in the UK and US risks written in the US to include the following information.

	a.	A more detailed description of this “ground up” approach covering the methodology and key assumptions. Clarify whether this approach was used at each reporting date or only at the initial reporting date. Describe any changes in this methodology during the periods presented.
	b.	Claim data for asbestos risks similar to that described for your US workers compensation business in your response to comment three in your October 20, 2005 letter. Explain significant changes over the three year period in amounts per claim.
	c.	A detailed description of the factors considered in reestimating the ultimate liability for cases that have significantly affected the changes in prior year reserve estimates for each period presented.
	d.	The impact of these factors on case emergence rates expected in the future.
	e.	Link this disclosure to your explanation of changes in prior year reserve estimates as described in your response to comment four in your October 20, 2005 letter.

3.	On page 11 of your response, we note that you propose certain new disclosure but have not provided related examples. Without examples it makes it difficult for us to evaluate the adequacy of your proposed disclosure. Please provide the following information in disclosure type format for both asbestos and non- asbestos reserves.

	a.	Explain and quantify each of the significant factors that you considered in your adjustment of the initial actuarial point estimates to produce the final recorded reserves at December 31, 2003 and 2004.
	b.	Explain and quantify those conditions deemed to be indicative of a long term trend that resulted in a reserving action at December 31, 2003 and 2004.
	c.	Link your discussion of these adjustments to your explanation of changes in prior year reserve estimates as described in your response to comment four in your October 20, 2005 letter.

	a.	Royal & Sun Alliance is responsible for the adequacy and accuracy of the disclosure in the Form 20-F;
	b.	Staff comments or changes to disclosure in response to staff comments do not foreclose the Securities and Exchange Commission from taking any actions with respect to the filing; and
	c.	Royal & Sun Alliance may not assert staff comments as a defense in any proceeding initiated by the Securities and Exchange Commission or any person under the federal securities laws of the United States.