Actuarial Best Estimates and Proxy Methods

General Insurance Reserving Actuarial Best Estimates and Proxy Methods
A commentary from the Actuarial Professions General Insurance Reserving Oversight Committee (GI ROC) to the report from the Board for Actuarial Standards GI Proxy Methods Experts Group (published September 2007).
January 2008
GI Board: Reserving Oversight Committee GI Reserving Actuarial Best Estimates and Proxy Methods January 2008
Comments from the UK Actuarial Professions General Insurance Boards Reserving Oversight Committee (ROC) on Proxy Methods, and on the Board for Actuarial Standards GI Proxy Methods Experts Groups report (the BAS Report) to the FSA on GI Proxy Methods (published September 2007), in the context of proposed Solvency II requirements regarding reserving. The BAS Report is appended to this response (page 8 onwards). ROC Membership Lis Gibson (Chairman) Caroline Barlow Neil Bruce Steven Fisher Neil Hilary Ian Hilder Richard Winter
Introduction The GI Board and CEIOPS have asked the Reserving Oversight Committee to provide their comments on behalf of the Actuarial Profession on the subject of Proxy Methods, and the BAS Report. There has been some concern expressed that the UK Actuarial Profession may appear to be endorsing the use of Proxy Methods as described in the BAS Report and it is in the particular context of that concern that we make these brief comments. In formulating our comments, we have recognised that BAS is responsible for technical actuarial guidance, and we have not sought to provide comments on the sections of the BAS Report which deal with the technical aspects of Proxy Methods and benchmark data. Rather, we have reviewed and commented on the principles behind the potential use of Proxy Methods and the practical and professional issues which we believe fall within the remit of the Actuarial Profession. The BAS Report explains its purpose and context as follows:
One of the cornerstones of Pillar I is an Actuarial Best Estimate of technical provisions. In discussions between CEIOPS (the Committee of European Insurance and Occupational Pensions Supervisors) and the Groupe Consultatif (which represents actuarial associations in the European Union) it was observed that some companies in Europe do not have Actuarial Best Estimates of their technical provisions, either because of lack of knowledge or lack of
data. It is envisaged that supervisors and actuarial associations will, where appropriate, encourage insurers to improve expertise and collect any necessary industry-wide data to allow Actuarial Best Estimates of technical provisions to be produced. In the interim, however, there is a need to find practical solutions (Proxy Methods), especially for smaller firms, where Actuarial Best Estimates are not available. Various member states have been asked to form Expert Groups to consider what Proxy Methods may be possible or appropriate in the absence of Actuarial Best Estimates. Against this background, the Financial Services Authority (FSA) asked the Board for Actuarial Standards (BAS) to set up an Expert Group to consider proxy methods for General Insurance (GI) technical provisions from a UK perspective. This report sets out the thinking of that Expert Group.
Our comments will be in the same context.
Summary of Our Comments and Recommendations Solvency II requires claims provisions to be (Actuarial) Best Estimates, defined as being the (discounted) mean of the distribution of possible outcomes. Proxy Methods are relatively simple methods which can be applied mechanically without a high degree of judgement, when either insufficient data is available to allow the use of more sophisticated actuarial methods, or where companies do not have access to the necessary actuarial expertise. Proxy Methods should make use of such data as is available, but there may be a need to supplement them with the use of suitable benchmark data. If insufficient benchmark data is available, there may be a need to arrange for the collection of such data to support the use of Proxy Methods. Where Proxy Methods are utilised, the intention should still be to produce a Best Estimate of the provisions required. The uncertainty associated with a paucity of data or the absence of appropriate judgement should be reflected in the risk margin rather than in the claims provisions themselves. Consideration of risk margins is beyond the scope of this note. Best Estimates should be produced using the judgement of a suitably qualified and experienced person. By virtue of their training and professional standards, suitably experienced qualified actuaries are ideally suited to exercise this judgement. A regulatory framework which caters for the setting of reserves in the absence of appropriate judgement is exposed to failure and potential abuse, as the mechanical application of methods does not take into account all aspects of the business. The real issue is not so much proxy methods as proxy judgement. Appropriate judgement can sometimes be exercised by experienced insurance professionals who have not qualified as actuaries. It would be practical and reasonable for a regulatory framework to provide transitional arrangements or exceptions for smaller companies, allowing suitable nonactuaries to produce reserves using their experience and judgement.
We recommend that a European benchmark database be created in time for the implementation of Solvency II, to make development data, benchmark development factors and benchmark prior loss ratios available to the market for as many business groups as possible. We set out a recommended approach for companies exempt from the actuarial best estimate basis to follow, which includes the nomination of an alternative expert, training and the use of central benchmarks as a safety net.
Overview of our Understanding of the BAS Report The BAS report has been written on the premise that Proxy Methods are methods for calculating claims provisions which are used by companies who do not use actuaries to assist in setting their technical provisions. It is our understanding of the BAS report that Proxy Methods are essentially defined by the authors to be methods used to calculate reserves utilising some or all of Various historical data Planning and pricing information Industry benchmarks Judgement of insurance professionals, such as claims managers and underwriters
as a substitute for actuarial judgement. The BAS Report sets out some examples of methods which might be used by actuaries as part of their overall methodology, subject to the use of actuarial judgement, to produce best estimates, or could alternatively be used by non-actuaries as a proxy for actuarial judgement. The BAS Report sets out some advantages and disadvantages of Proxy Methods as used by non-actuaries as a proxy for actuarial judgement. They explain that on some occasions the judgement of other professionals may be adequate. They also explain that the use of these methods by actuaries applying judgement is fine. The BAS Report makes an assumption that when Proxy Methods are used, a degree of prudence should be built in. They believe that this is the intention of CEIOPS, although they do also acknowledge that this is inconsistent with the Solvency II requirement for Best Estimates. They do not explain how such caution is to be introduced or measured, or how this may subsequently be allowed for in the calculation of risk margins. The BAS Report appears to make no reference to the discounting of technical provisions. However, the Solvency II requirement for a discounted Best Estimate means that further consideration may need to be given to this aspect.
The BAS Report also sets out some disadvantages of the use of Proxy Methods, including the dangers of nave interpretations of past data or benchmarks and the potential need to allow for future trends beyond the data. They conclude
In short, methods based on the application of benchmark ratios / factors, be they market derived or from other sources, may produce very unstable results. Such methods are best used to validate more sophisticated approaches and to inform judgements, rather than being used in isolation as a basis for setting provisions.
The BAS Report concludes that Proxy Methods cannot be relied upon to provide a meaningful estimate of future claim costs without appropriate judgement and understanding, and that such methods are best used to validate more sophisticated approaches and to inform judgements, rather than being used in isolation as a basis for setting provisions. They also appear to conclude that reliance on the judgement of other insurance professionals may sometimes be a valid proxy for reliance on the judgement of actuaries. Our Thoughts on Proxy Methods The request for consideration of Proxy Methods was prompted by the observation that some companies in Europe do not have Actuarial Best Estimates of their technical provisions, either because of lack of knowledge or lack of data. Solvency II requires claims provisions to be (Actuarial) Best Estimates, defined as being the (discounted) mean of the distribution of possible outcomes. We note that a previous paper prepared on behalf of the Actuarial Profession sets out the considerations to be taken into account in setting such best estimates. Essentially the calculation of Best Estimate claims provisions depends on all of: The availability of suitable credible data The application of appropriate calculation methods The application of judgement based on experience and knowledge of the business and market conditions. Our interpretation is that Proxy Methods may be required when either insufficient data is available to allow more sophisticated actuarial methods to be applied, or where companies do not have access to the necessary actuarial expertise, knowledge or experience required to apply more complex methods or to exercise judgement as to the appropriateness of different methods or the validity of the answers produced. Thus Proxy Methods are likely to be relatively simple methods which can be applied mechanically without a high degree of judgement. It is likely that for smaller companies and for companies which have not previous applied actuarial methods, there will be issues regarding the availability of sufficient data. For smaller companies (or indeed for small lines of business within larger companies), the volume of business may mean that the data available is sparse, or that the experience is volatile. For companies without previous actuarial involvement, there may be a lack of suitable historical data to use as the basis for projections, and the nature of the companys systems may make it difficult to reconstruct such historical data. In these circumstances, there will be a need for Proxy Methods, which should make use of such data as is available but may need to supplement it with the
use of suitable benchmark data. If insufficient benchmark data is available, there may be a need to arrange for the collection of such data to support the use of Proxy Methods. We believe that where Proxy Methods are utilised, the intention should still be to produce a Best Estimate of the provisions required. There will of course be a need to consider the credibility of the data available, and to supplement this with suitable benchmark data where necessary and feasible. We acknowledge that lack of appropriate data will increase the uncertainty associated with the projections, and there may be a tendency to take a more cautious view to compensate for this. However, unlike BAS, we believe that this uncertainty should be reflected in the risk margin rather than in the claims provisions themselves. (We also note that there is a need to consider how Proxy Methods for the calculation of risk margins can be developed, but that is outside the scope of this note.) We believe that ideally Best Estimates should be produced using the judgement of a suitably qualified and experienced person. Article 47 of the EC Draft Directive (quoted in section 5.1 of the BAS report) clearly envisages the existence of an actuarial function to provide this expertise. By virtue of their training and professional standards, we believe that suitably experienced qualified actuaries are ideally suited to exercise this judgement. Whatever methods, models or benchmarks are applied, the absence of experienced judgement and interpretation may render the results potentially misleading. A regulatory framework which caters for the setting of reserves in the absence of appropriate judgement is therefore exposed to failure and potential abuse. Most, if not all, reserving methods can however be used by anyone, with or without appropriate experience or judgement. Sometimes using judgement-free methods can provide useful objective benchmarks and can assist in illustrating the context in which judgement is being exercised, and the scale of uncertainty. However, regulatory reserves should not be based on judgement-free methods. We believe that the real issue is not so much proxy methods as proxy judgement. We believe that appropriate judgement can sometimes be exercised by experienced insurance professionals who have not qualified as actuaries. We believe that it would be practical and reasonable for a regulatory framework to provide transitional arrangements or scale-based exceptions allowing suitable non-actuaries to produce reserves using their experience and judgement.
Our Recommendations to CEIOPS In order for proxy methods to be a viable alternative to the full standard actuarial approach for calculating best estimates, we make the following recommendations: We recommend that a European benchmark database be created in time for the implementation of Solvency II. The benchmarks should include the aggregated premium, paid claim and incurred claim development data from the relevant classes of business. The data required to populate such a database would of course have to be
collected in such a way as to comply with European Competition Law. The most natural way for this would be via the statutory returns which would need to be designed accordingly. We recommend that on an annual basis aggregate market premium, paid claims and incurred claims development patterns, and average market loss ratios be calculated using full data and appropriate actuarial judgement and made available to the market as benchmarks. This could be financed by charging insurers and advisors a small fee for annual access, perhaps with insurers of below a minimum size being granted free access. We recommend that any insurance companies below a particular size, to be decided, very small companies, be granted perpetual exemption from the need to use a qualified actuary to produce actuarial best estimate reserves, provided they continue to qualify in terms of their size. We recommend that they have an alternative approach as set out below. We recommend that companies above the very small threshold but below a second threshold, small companies be granted a transitional period during which they can follow the alternative to an actuarial best estimate reserving basis, but after which they must revert to the actuarial best estimate basis. We recommend that transitional arrangements need to be set in place to collect appropriate data. The Recommended Alternative Approach We recommend that very small and small companies who opt for the alternative to actuarial best estimate reserving, must nominate their proxy reserving expert, and state that persons relevant skills and experience. That person must attend a course (say of 2 or 3 days, to be run centrally) and must provide evidence that they are suitably experienced to provide a proxy to actuarial judgement in the context of their company. The nominated expert will be required to perform and report on certain calculations such as chain ladder factors to ultimate, ultimate loss ratios and average claim sizes, using the centrally produced industry benchmarks described above, on the classes relevant to their company. This should not be positioned as directing or even encouraging the adoption of these benchmark results as their reserve estimates. The nominated expert must explain their own reserving figures in the context of the results using the industry benchmarks, setting out the similarities and differences and reasons for them. We believe that this is a practical way to provide a minimum standard, which should not be unduly onerous for small and very small companies, especially mono-lines.
UK Expert Group Report On General Insurance Proxy Methods
Prepared by GI Proxy Methods Expert Group Version 5 Published September 2007
INTRODUCTION
Solvency II is an EU project which (quoting from the ECs Framework for Consultation) aims to develop a new solvency system to be applied to life assurance, non-life insurance and reinsurance undertakings, which Member States and supervised institutions are able to apply in a robust, consistent and harmonised way. Solvency II will be built on three pillars (similar to Basel II): Pillar I: Pillar II: Pillar III: quantification of capital requirements supervisory review process market analysis of published data
One of the cornerstones of Pillar I is an Actuarial Best Estimate of technical provisions. In discussions between CEIOPS (the Committee of European Insurance and Occupational Pensions Supervisors) and the Groupe Consultatif (which represents actuarial associations in the European Union) it was observed that some companies in Europe do not have Actuarial Best Estimates of their technical provisions, either because of lack of knowledge or lack of data. It is envisaged that supervisors and actuarial associations will, where appropriate, encourage insurers to improve expertise and collect any necessary industry-wide data to allow Actuarial Best Estimates of technical provisions to be produced. In the interim, however, there is a need to find practical solutions (Proxy Methods), especially for smaller firms, where Actuarial Best Estimates are not available. Various member states have been asked to form Expert Groups to consider what Proxy Methods may be possible or appropriate in the absence of Actuarial Best Estimates. Against this background, the Financial Services Authority (FSA) asked the Board for Actuarial Standards (BAS) to set up an Expert Group to consider proxy methods for General Insurance (GI) technical provisions from a UK perspective. This report sets out the thinking of that Expert Group.
GI Proxy Methods Expert Group
12 September 2007
SUMMARY
Extent of the use of Proxy Methods in the UK We dont believe there is any significant use of Proxy Methods in the UK. The vast majority of UK insurance companies use actuaries in some capacity to help set their GI technical provisions. Types of Proxy Method In the rare instances where Proxy Methods have been used, they include: A. B. C. D. E. F. G. H. Applying expected loss-ratios (typically from Plan or Pricing / Underwriting assumptions) Judgement of experienced underwriters / claims staff Applying benchmark loss-ratios (from external data) Using benchmark claims development factors (from external data) Using benchmark survival ratios (typically for run-off / latent claims exposures) Case estimates alone (where known to be conservative, or for Catastrophe claims, say) Using IBNR ratios (as a percentage of case estimates or premium, for example) Application of statistical methods by non-actuaries
Details of these Proxy Methods are given in section 3. Availability of benchmark information The main likely source of benchmark information in the UK (outside Lloyds) is the annual FSA returns. These include loss-ratio and claim development information that can be used, with caution, to provide indicative loss-ratio or claims development factors. Lloyds of London have developed an in-house reserve benchmarking pack which syndicates can use to compare their reserves to market information in various ways. The Association of British Insurers publishes some high level market information on claims and profitability trends but this is not suitable to provide benchmarks for setting provisions. Advantages and Disadvantages of Proxy Methods In some situations Proxy Methods can be perfectly valid and may well be the sort of method that an actuary would use if they were involved in setting provisions. This is typically the case for classes of business which have no material, or credible, claims experience to act as a base (for example cyber risks or political risks), or classes of business that are being written for the first time by an insurer (for which, initially, market benchmark or Plan / expected loss-ratios may be the only guide). Setting provisions using the judgement of underwriting / claims staff for small classes of business, where those involved have a detailed knowledge and understanding of the business (and the cost of significant actuarial involvement may be disproportionate), may also be a perfectly reasonable approach. There are, however, significant risks in using market benchmark data or the nave application of standard claims development factors. The past is not necessarily a guide to the future! There are a whole host of reasons why benchmark ratios and factors may be inappropriate. These include changes to: the external legal environment; claims handling processes; features of the business such as excesses or deductibles; the mix or type of business being written; randomness of past claims experience. In short, methods based on the application of benchmark ratios / factors, be they marketderived or from other sources, may produce very unstable results. Such methods are best used to validate more sophisticated approaches and to inform judgements, rather than being used in isolation as a basis for setting provisions.
TABLE OF CONTENTS
SECTION
Introduction Summary 1. 2. 3. 4. 5. What weve done Decision-tree for Proxy Methods Types of Proxy Method Benchmarks Advantages and Disadvantages of Proxy Methods
PAGE NUMBER
2 3 5 6 8 12 15
Appendix I Appendix II Appendix III
Questions we asked consultancies about Proxy Methods Sample benchmark information from Lloyds of London Bibliography
1.
1.1
WHAT WEVE DONE

Who we are . The UK Expert Group comprises: Henry Johnson (and Jerome Kirk) / Lloyds of London James Upson / Financial Services Authority Julian Lowe / Aviva (and member of the Board for Actuarial Standards) (Chair) Kathryn Morgan / Co-operative Financial Services (when the Group formed) (Deputy-Chair) Louise Pryor / Board for Actuarial Standards Rachel Griffiths / Association of Mutual Insurers (AMI) Yvonne Braun / Association of British Insurers (ABI)
1.2
Who weve consulted with . The Expert Group has considerable experience of setting GI provisions in the UK (and overseas). However we thought wed supplement our experience by approaching a number of consultancies, namely Deloittes, PwC, KPMG, E&Y and Mazars. We were aware that theres potentially an element of self-selection if we speak to actuaries at consultancies. By definition, actuaries would only tend to deal with the sorts of insurance companies who deal with actuaries! However, the various consultancies do have experience of companies who dont routinely use actuaries to any great degree, so were able to give perspectives as to what methods and techniques they had seen in such companies. Although Mazars has an actuarial function, they also have a number of clients who dont use this (or other actuarial support), so were able to give some useful insight too. Finally, through the AMI, we spoke to Chief Executives of a number of smaller insurance companies, whose size doesnt warrant an in-house actuarial function. Our consultations revealed that nearly all UK insurance companies have some actuarial involvement in the setting of their technical provisions.
1.3
What benchmark information weve looked at . The FSA, ABI and Lloyds of London are all represented on our Expert Group. The relevant representatives provided information about the sort of benchmark information that is collected under their auspices, which is described in section 4.
2.
2.1
DECISION-TREE FOR PROXY METHODS

When is a Proxy Method not a Proxy Method? As described in the Introduction, the cornerstone of Pillar I of Solvency II is an Actuarial Best Estimate of a companys technical liabilities. The potential need for Proxy Methods arose because of concerns that some companies do not have their technical provisions set with actuarial involvement. Hence we have taken this to mean that Proxy Methods only need to be considered for companies that do not use actuaries to help set their technical provisions. So, for example, some of the Proxy Methods we describe in section 3 may, in fact, be used by actuaries. For our purposes they are deemed to be Proxy Methods if they are applied by non-actuaries. Its also important to realise that whilst actuaries may be involved in setting provisions in most companies in the UK, the ultimate responsibility for setting provisions sits with the management of the insurer (based on actuarial advice as to what the underlying Best Estimate of the liabilities might be).
2.2
What is the decision-tree for using different types of Proxy Method? As noted above, our starting point is that any method used by an actuary constitutes an Actuarial Best Estimate. So in what follows we are only referring to instances where methods are applied by nonactuaries. Decision One: is this an entirely new class of business for both your company and the industry? Entirely new types of insurance wont have any historic data to act as a guide, nor will there be any industry benchmarks. If the answer to this question is yes, then the only source of guidance as to the level of reserves will be the expected or Plan loss-ratios, Proxy Method A in section 3. These expected or Plan loss-ratios should, one would hope, be based on market research and perhaps some underlying research and assumptions about the exposure to the risk of loss. An example of a relatively new type of insurance is cyber risk cover. This provides insurance for computer attacks by insiders or hackers that may include destruction or alteration of data, inability to access services, or downstream liability for attacks against other computers or networks from ones own systems. Such risks simply did not exist a couple of decades ago. If the answer to this question is no, then one can proceed to Decision Two. Decision Two: is this a new class of business for your company (but not the industry)? Companies may start writing new classes of business (for example Motor products having only previously written Household insurance). Whilst such classes of business may have been in existence in the market for many years, a company writing such business for the first time will not have any claims or profitability experience to act as a guide. In such instances Proxy Methods A, B or C may be appropriate. These either use the expected or Plan assumptions when the business was written, as per Decision One above, or supplement this with the judgement of staff (who may have experience of the business from other companies) or the application of industry benchmarks for typical levels of profitability (hence claims costs). Once such a new class has been written for a while, other Proxy Methods may be possible, once one has some credible historic claims data to act as a base. Classes may also be new but similar to existing products, in which case one might also have credible, relevant, data to allow other Proxy Methods to be used. If the answer to this question is also no, then one can proceed to Decision Three.
Decision Three: is there any credible historic claims data for your own company ? If the answer to this question is yes, then the class of business is not new to either your own company or the industry (from Decisions One and Two) and you have a credible amount of historic claims data. In this case there is enough information to perform a number of Proxy Methods. The initial potential Proxy Methods A, B and C may all still be used. In addition Proxy Methods D-H may also be used. Method D uses benchmark development patterns to apply to ones own claims data. For example assuming that 75% of claims have been paid after one year, a sensible allowance for future claims may be to take one third of the claims paid to date (that is, the missing 25%) as a provision. Method E is only intended to be used by companies in run-off, which typically may have exposure to asbestos-related or other types of latent claims. A typical Proxy Method in these circumstances may be the application of some standard, industry, survival ratios that estimate the future liability as a multiple of the rate of paying, or reporting, of claims. Methods F and G are similar to Method D in that they are implicitly making assumptions about likely future levels of claims given certain information about the claims paid or reported to date. For example Method F (just using case estimates) assumes that at the point the provisions are set that all claims for a period have been reported, no previously settled claims will be reopened and case estimates for reported claims will be more than adequate to pay for the final cost of claims (or that any surplus in the case estimates more than outweighs any future reporting, or reopening, of claims). Method H comprises doing the sort of projections and assessments that actuaries would do, but that this is being performed by non-actuaries. If the answer to Decision Three is no, then you are probably dealing with a type of business for which the claims experience is very sparse. For example a Catastrophe reinsurance contract with a high level of excess may be designed and priced such that it only expects to incur any claims once every twenty five years. It would therefore be reasonable to write such business for several decades and not receive any claims. Other more esoteric types of insurance may insure very rare events the possibility of Elvis being seen in your local fish and chip shop, for example. For such classes there may be a very small possibility of ever receiving any claims. In these circumstances one comes back full circle to Proxy Methods A, and possibly B, as the only reasonable approaches to arriving at expected claims costs.
3.
3.1
TYPES OF PROXY METHOD

What types of Proxy Method are there? From the Expert Groups own knowledge and the further input from a number of consultancies and smaller companies, we have categorised Proxy Methods into a number of types, Methods A-H below. Methods A and B can be described as Expectation and Judgement. Methods C-E can be described as Industry benchmarks. Methods F and G might be called Simple factor-based methods. Method H is Statistical techniques applied by non-actuaries. As previously noted, we are assuming all these Proxy Methods are applied by non-actuaries. Because there is considerable scope for the Proxy Methods to be unstable and unreliable, a further assumption we have made is that when insurers use Proxy Methods they should err on the side of caution. This is at odds with the intention of Solvency II (that an Actuarial Best Estimate should form the basis of Pillar I and be just that a Best Estimate with no deliberate caution or margin). The additional prudence reflects the lack of actuarial involvement and our understanding of the intention of CEIOPS regarding Proxy Methods. It is important to have some way of tracking how the claims experience develops if Proxy Methods are used. For example being able to split out the claims experience for prior periods to see whether the underlying assumptions made in setting provisions have been systematically too high, or too low, to assess the suitability of the Proxy Method. Where possible it would also be good practice to consider a number of Proxy Methods
3.2
Expectation and Judgement methods Method A: Applying expected loss-ratios When a company writes a class of business it will have made some initial, or Plan, assumptions about the likely level of claims experience. For classes of business which the company has been writing for a while, future loss-ratios may be based on past experience trended forwards to allow for claims inflation, rating action and changes in the mix or characteristics of the business. For entirely new classes of business (to the company and the industry), or classes of business with very sparse data, the company should have performed some underlying research into the likely exposure to claims. For example for high layers of Catastrophe reinsurance cover there is a considerable body of meteorological data on expected frequency and severity of windstorms and a number of companies happy to sell you their in-house models of windstorm exposures. For relatively new risks like cyber insurance, companies should have researched the underlying dynamics of the types of risk being insured and looked at whatever market research there is about the types of risk involved, or perhaps performed their own surveys or analysis. In the absence of past insurance experience or any credible claims data, using the assumption implicit in the pricing basis when the business was written may be the only sensible approach to setting provisions. Depending on the class of business, the provision may be set at this initial Plan, or expected, level and then reduced over time as any actual claims experience develops. For example one could draw down the initial provision by the amount of any payments made. Or if you thought all claims were likely to be reported over a period of N years, reduce the provision down to zero by 1/Nth per year over that N year period.
Method B: Judgement of experienced underwriters / claims staff The underwriters / claims staff may have worked in the industry for many decades and have an excellent knowledge of the companies own classes of business and market intelligence about claims or product developments. The application of such detailed knowledge and experience may be simply gut feel, or it may be combined with a more scientific, but often ad hoc, review of loss development triangles or other claims statistics. In smaller companies, or perhaps those with unusual niche products, this may be an entirely justifiable approach, particularly when the scale of the business would not warrant more detailed actuarial involvement (and the lack of actuarial involvement is compensated for by an element of prudence). 3.3 Industry benchmark methods Method C: Applying benchmark loss-ratios As well as a companys own Plan or expected loss-ratios, companies may also look to market information on loss-ratios as a guide. Such market information may be available through published results, friendly reinsurers (who might provide Quota Share reinsurance in cases where a company is expanding into a new market), or statutory returns (see section 4.2). Some companies may be trying to follow the prices of other companies, or use other companies rates as a guide, so may have Proxy market profitability information available by that route. The same considerations would then apply as Proxy Method A. The assumed market loss-ratio may be used as a guide to setting the companys provisions, noting the need for caution and prudence referred to earlier. Method D: Using benchmark claims development factors The application of this sort of benchmark information requires a little more sophistication. In the UK the FSA requires companies to submit FSA returns which show a considerable amount of information about the loss development of each company see section 4.2. This includes the amount and number of claims paid and reported for a series of years after the initial Accident (or Underwriting) year in which the business is written. Someone with a modicum of statistical ability and knowledge of general insurance could use the FSA returns to arrive at some default claims development assumptions, either based on a number of larger companies or for the industry as a whole. There are a number of companies who provide software to let you access FSA returns for the entire UK market, or some companies publish research analysing the FSA returns (investment analysts and consultancies). It could be that a company has its own claims development data, but the company is so small that, for a particular class, the developments may be easily distorted by large, or exceptional, claims experience. In such cases the typical development of larger, more credible, data may be used as a guide. Auditors or consultants may also be able to provide benchmark claims development factors. The essence of any claims development method is to use the past development of claims costs as a guide to the future. So, for example (as described in section 2.2), one may observe that in past periods on average 75% of claims have been paid after one year. This might mean that a sensible allowance for future claims may be to take one third of the claims paid to date (that is, the missing 25%) as a provision. There are, of course, many reasons why such an approach may be flawed, which is why the Solvency II model is to use actuaries to produce Best Estimates, taking account of the factors that may distort such methods see sections 3.5 and 5.2 for some more references to these distortions. It is good practice to produce estimates of future claims costs using a range of techniques. So, for example, one could use benchmark data to perform projections based on claims payment data, claims incurred data (payments plus case estimates) and project numbers of claims multiplied by average costs.
Method E: Using benchmark survival ratios As noted in section 2.2, this Proxy Method is only intended to be used by companies in run-off (or ongoing companies who have some business in run-off), which typically may have exposure to asbestos-related or other types of latent claims. A typical Proxy Method in these circumstances may be the application of some standard, industry, survival ratios that estimate the future liability as a multiple of the rate of paying, or reporting, of claims. For example if a company has exposure to UK asbestos claims, one could use the benchmark survival ratios from the UK Actuarial Professions GIRO paper: UK asbestos the definitive guide(1) as a sensible starting point to estimate future liabilities. This would involve taking the assumed underlying rate of reporting of, say, mesothelioma claims and multiplying this by the relevant survival ratio (in this case 68) to produce an undiscounted estimate of future claims costs. Other sources of benchmark survival ratios may be publications that refer to industry survival ratios (for example a rating agencys review of asbestos and latent claims exposures), or helpful information from ones own firm of auditors, or consultants, as to typically assumed survival ratios. 3.4 Simple factor-based methods Method F: Just booking case estimates alone If companies routinely set case estimates that are too high, simply booking case estimates may be a prudent Proxy Method (noting our prior assumption that CEIOPS intends Proxy Methods to err on the side of caution). Of course initially there may be delays in reporting of claims so this runs the risk that the provision for the current years business is too low. However if a business is in a fairly steady state, with case estimates for prior years comfortably too high, then the likely surplus for prior periods may more than outweigh any potential deficit due to late reporting of claims for the current year. It would be important for a company who uses this approach to monitor the development of claims for individual cohorts of business, to ensure that the assumption that case estimates are typically too high continues to hold good. Companies would have to be alert to any changes in case estimating philosophy, or changes in the nature of the business or the external claims environment, that may invalidate this assumption (see sections 3.5 or 5.2 for more details). Whilst the assumption that likely surpluses for past periods may more than outweigh the omission of late reported claims from current periods, should a company start to grow significantly, this assumption may also not hold good. Generally we do not regard this as a satisfactory Proxy Method. Method G: Using IBNR ratios This Proxy Method is a variation on Proxy Methods D and F. Some companies may have developed approximate assumptions over time that a suitable IBNR allowance might be a certain percentage of case reserves, or a certain percentage of premium. In some cases overseas there may be crude steers from local regulators that IBNR reserves should be at least X% of premium. The same concerns for Proxy Method F apply, with there being many reasons why such methods may not be reliable over time. Like Proxy Method F it would be important for companies to have ways of tracking this assumption, to see if it holds good, or spot sooner rather than later if the amount of IBNR required seems to be higher, or lower, each year.
10
3.5
Statistical techniques applied by non-actuaries Method H: statistical techniques applied by non-actuaries In some companies the analysis used to set provisions is performed by very experienced non-actuaries, who have a considerable amount of skill and expertise and who carry out their work in a thoroughly diligent and professional fashion. At the other extreme, it is possible for enthusiastic amateurs to pick up a text book and apply some simple chain-ladder techniques without any real appreciation of the dangers in doing so or the types of checks and balances that should accompany a high standard exercise to set provisions. It is not our intention to write a text book on how to perform actuarial techniques there are many publications which describe the steps involved and the potential pitfalls for the unwary. The UK Actuarial Profession has set out some considerations that should apply when estimating general insurance (2) provisions in their paper: Best Estimate provisions for General Insurance . The UK Actuarial Profession also recently sponsored some work to look at reserving practices in the UK which makes a number of recommendations as to good practices to employ when performing a reserving exercise: A change agenda for reserving. Report of the General Insurance Reserving Issues Taskforce (GRIT)(3). Any non-actuaries involved in setting provisions would be well advised to read these and similar publications to gain an appreciation of the sorts of consideration that should be borne in mind. Section 5 describes some of the issues to look out for with any method of estimating future claims costs that would apply particularly to any of the Proxy Methods described above.
11
4.
4.1
BENCHMARKS
Benchmarking data in the UK The Expert Group reviewed the types of benchmark information available in the UK. The main likely sources are the FSA returns, ABI data and the reserving benchmarking exercise performed by Lloyds of London. Each of these is described in more detail in the following sections. There may be other, softer, sources of benchmarking data, for example competitors report and accounts, industry publications looking at market loss-ratios or claims developments and personal knowledge of ones own staff. The various types of hard and soft benchmarking information may be used to help produce Proxy Methods B-D described in section 3. In the past insurers may have informally exchanged information to provide each other with technical assistance. Competition Law means it is now far harder to do so and insurers need to ensure they are not engaging in any anti-competitive behaviour when they share information.
4.2
FSA returns and its use for benchmarking For those not familiar with the UK insurance market, all insurers carrying on business in the UK (other than EEA firms subject to the insurance or reinsurance directives carrying on business through freedom of establishment or provision of services, other EEA firms availing themselves of treaty rights and nondirective friendly societies) must provide annual accounts and supplementary information in a standardised format to the UK regulator, the Financial Services Authority (FSA), in a document called the FSA return. The FSA returns are available to the public. Where it is unduly burdensome to provide the full detail in the returns (especially for smaller firms), the FSA may exempt the firm from providing some of the detailed information required. There are different types of return; a "Global return" reports the entire world-wide business of the firm, and (for most third country firms) branch return reports only in respect of the branch. The requirement as to whether a Global return, UK branch return or EEA branch return (or some combination of them) is provided depends on the location of the head office and the type of firm. There are a number of companies (such as Synthesys) which collate FSA returns data and sell software databases containing data from the returns of most or all insurers in the UK market. The FSA uses the returns as a key source of information to monitor the financial resources of an insurer and, for GI business, to assess retrospectively the adequacy of the insurer's claims provisions, both by reviewing the claims development of a firm over some years and by comparing the level of claims provisions between insurers. However, these analyses tend to be a starting point for discussion with the firms, recognising that there may be sound explanations for an apparent low level of claims provisions. Data in the returns is normally split into the following classes of business (no detailed information is provided separately for small classes of business): Accident & Health Motor (split between Personal Motor and Commercial Motor) Aviation Marine Goods in Transit Property (split between Household & Domestic and Commercial Property) Commercial Liability Financial Loss (split between Personal Financial Loss and Commercial Financial Loss) Miscellaneous Direct and Facultative Balance Direct and Facultative (lumping together any small classes) Treaty Reinsurance (split between Non-Proportional Treaty, Proportional Treaty and Miscellaneous) Balance treaty reinsurance (lumping together any small classes of treaty reinsurance)
12
One form which could be highlighted for its usefulness in benchmarking is Form 23 of the returns. In this Form, firms provide claims data split by class and accident year. For example, the ratio of Incurred But Not Reported claims (in column 6) to the corresponding reported claims (in column 5) could be used as a benchmark. Likewise, the claims ratio (in column 13) is of interest. It is also possible to use the development of claims from the beginning to the end of the year as benchmark development factors for a chain ladder exercise to help set provisions. However, any benchmarking against other firms contains a number of dangers, as mentioned in section 5.2. In addition, if you benchmark against only a few (or even one other) firms, it is possible that your benchmark is being distorted by special factors, such as a transfer of business, reinsurance arrangements or exceptional claims. The FSA has no plans at present to publish benchmark information to assist in setting claims provisions. 4.3 ABI data The ABI collates various statistics from its members. It publishes some of this aggregate data and more detailed information is available on the Members Only section of its website. The information covers a range of classes of business. By way of illustration for Motor it includes: Motor statistics This section contains statistics on premiums and claims, commission and expenses, change in provisions, equalisation reserves, underwriting result, operating ratios, and overseas data, for the UK and worldwide. Annual Motor Statistics This shows claims split by payment type with particular attention paid to bodily injury claims. Quarterly Motor Claims Market trends in the number of all claims and, in particular, the number and cost of all theft claims. Analysis Of UK Motor Market These reports look at the experience of the private and commercial motor markets over an eleven year period by analysing data extracted from companies statutory returns to the FSA. The level of information is such that its unlikely to be of any assistance for any benchmark Proxy Methods. 4.4 Lloyds of London reserve benchmarking exercise In the UK the words reserve / reserving are often used interchangeably with provision / provisioning. Whilst in the rest of the report we have generally referred to provisions, in this section we have followed Lloyds own nomenclature and referred to reserves / reserving. Lloyds undertakes an annual relative reserve benchmarking exercise. The main purpose of the exercise is to highlight and gain an understanding of the syndicates reserve levels within the market. The exercise can also be used to identify potential anomalous reserving positions that may need further investigation. For each syndicate a unique benchmark is constructed and compared to. The benchmark syndicate is calculated from the market data (excluding the syndicate in question) adjusted to reflect the underlying mix of business of the syndicate. The rebasing is required to ensure suitable correspondence between the syndicate and its benchmark and is conducted at a year of account/class of business level. Lloyds works with around 50 low level classes of business.
13
A number of Key Performance Indicators (KPIs) are calculated for both the syndicate and the benchmark over various time periods and brought together during the analysis. Reserve benchmarking necessarily includes large volumes of output as considering single or simple reserving measures can often be misleading. This in turn can lead to results becoming complex and difficult to interpret. To overcome this, Lloyds have extended the exercise to include the indexing of results which allows agents to be easily ranked for comparative purposes. The key points to consider when constructing the benchmark analysis are: The analysis must be conducted at a low enough level to allow for differences in behaviour between classes of business. The analysis focuses on relative reserve strength. It does not consider reserve adequacy. Numerous measures are required, preferably over time, to obtain a sound understanding of an underlying entitys reserves.
The exercise focuses on four key areas both gross and net of reinsurance: Reserve strength The absolute levels of reserves are considered. The driving principle is that the larger the reserves held, relative to payments and ultimates, then the stronger the reserves are. The four measures applied to estimate reserve strength are: reserves as a percentage of ultimate claims, IBNR as a percentage of ultimate claims, case reserves as a percentage of incurred claims and survival ratios. IBNR utilisation The proportion of the IBNR held at the start of a calendar year that is used (incurred) during the year shows how quickly the IBNR is being burnt. The slower the IBNR burn the stronger the reserves are judged to be. The three measures for IBNR utilisation are IBNR burn over each of the last 3 calendar years (2006, 2005 and 2004 for the current exercise). Reserving over time It is important to understand whether agents tend to increase or decrease their estimate of ultimate claims over time. Those that consistently increase will have tended to set weaker reserves initially and vice versa. The two measures for reserving over time are to compare the ultimate claims projected as at now (year-end 2006) for a given year of account to the ultimate claims set at the end of year 1 and the end of year 3. Quality of business Quality of business is important when considering reserve strength since better business will generally require lower nominal reserves. As the measures above look at nominal reserving, there needs to be some allowance for better quality business. If all the other measures considered are equal, better underlying business would suggest stronger reserves. The measure used for quality of business is paid loss ratio. Ultimate loss ratios are not used because these are dependent on reserve strength which may distort results. These four measures are ranked and combined to form a single overall Reserve Benchmarking Index (RBI). Separate RBIs are provided gross and net of reinsurance. The overall RBIs can then be used to directly compare the relative reserving of syndicates within the market. The results are compiled and fed back to the syndicates within the market. The analyses are presented at varying levels of granularity from single measures for a syndicate as a whole down to information on individual years of account over time. It would be possible to further break results down to a class of business level but this would include voluminous output that would become less useful. A copy of an example syndicate pack is included at Appendix II. This is based on a hypothetical syndicates results and is for illustration only.
14
5.
5.1
ADVANTAGES AND DISADVANTAGES OF PROXY METHODS

Why would you use Proxy Methods ? As noted in section 2, we have assumed that Proxy Methods only need to be considered for insurers that do not use actuaries to help set their technical provisions. The clear steer from the development of Solvency II however is that an Actuarial Best Estimate should be at the heart of Pillar I. The latest EC Draft Directive, published on 10 July 2007, states (in Article 47) that: 1. Insurance and reinsurance undertakings shall provide for an effective actuarial function to undertake the following: (a) to coordinate the calculation of technical provisions; (b) to ensure the appropriateness of the methodologies and underlying models used as well as the assumptions made in the calculation of technical provisions; (c) to assess the sufficiency and quality of the data used in the calculation of technical provisions; (d) to compare best estimates against experience; (e) to inform the administrative or management body of the reliability and adequacy of the calculation of technical provisions; (f) to oversee the calculation of technical provisions in the cases set out in Article 80; (g) [The List goes on .] . Whilst all the above is in Draft, clearly it is expected that actuaries should be heavily involved in the calculation and oversight of GI technical provisions. However for small companies, or those writing niche products, the cost of significant actuarial involvement may be disproportionate. In such cases, as noted in section 3, we understand (and have assumed that) it is the intention of CEIOPS that Proxy Methods should err on the side of caution. It would be up to the management of the company, and the regulator, to form a view as to the capability of those involved in setting the provisions. Quoting from Article 47 again, in section 2 this describes the capabilities of those performing the actuarial role which include: sufficient knowledge of actuarial and financial mathematics and able where appropriate to demonstrate their relevant experience and expertise with applicable professional and other standards. This doesnt preclude non-actuaries being involved but notes the experience/expertise required and the need to have some sort of standards. As noted in section 3.2 under Proxy Method B, using the judgement of non-actuaries (typically underwriting / claims staff) to set provisions, may be entirely appropriate, subject to the requirements for experience / expertise and standards noted above.
5.2
What are the dangers in using Proxy Methods ? Many of these have been referred to already. The reason that the Draft EC Directive places such responsibility on actuaries is that mechanical or simplistic methods of setting provisions, applied without the necessary experience or understanding of the nature of the business, can produce very unreliable results. The starting premiss of any form of projection is that the past is, in some sense, a guide to the future. Actuaries then use their judgement and expertise to adjust for the various reasons why this is invariably not the case. In the absence of the application of this judgement and experience, Proxy Methods can be materially inaccurate. That is not to say that any one Proxy Method is inherently wrong or unsuitable. Just that Proxy Methods cannot be relied upon to provide a meaningful estimate of future claims costs without appropriate judgement and understanding.
15
Some of the many reasons why the assumptions underlying Proxy Methods may not hold good are listed below: Expected or Plan loss-ratios may be based on flawed assumptions. Few people set out and Plan to write unprofitable business but many of them manage to do so. Plans are generally more likely to be optimistic than pessimistic. Individual company experience, of both profitability and claims developments, may be markedly different from the market as a whole. Market benchmarks for loss-ratios or claims development factors may lead to reserves being significantly under- or over- stated. External legal, or claims environment factors (such as inflation, unemployment, market competitiveness) can change rapidly. For example legal decisions may increase or decrease typical amounts of injury claims, or worsening unemployment may considerably increase theft frequencies. New types of claim may also emerge. Past assumptions about claims development can become inappropriate in a short amount of time. Benchmark data would not reflect recent experience quickly enough. Claims development patterns may be significantly changed by internal factors such as new claims handling guidelines, new approaches to setting case estimates, the speed of paying claims or work state issues in the claims team. Internally (or externally) derived claims development patterns may need adjustment to reflect changes in claims handling processes. Internal product features may have changed which need to be reflected in any reserving exercise. For example higher or lower excesses or deductibles may dramatically change average costs or the number and speed of reporting of claims. Underwriters may have changed and the types of business being written may have changed. These changes could invalidate any assumptions about past claims or profitability experience. There may have been changes to a companys IT systems that mean that claims data is distorted, or changed, and past trends in claims data are no longer appropriate. As a generic observation, over-reliance on any one Proxy Method is inappropriate. Each may, at a point in time, produce sensible estimates. However changing circumstances may render its accuracy and validity of limited use. For this reason no one Proxy Method should be thought of as appropriate, rather a range of approaches should be used.
The list of potential pitfalls could go on for many pages. The UK Actuarial Professions Solvency II (2) (3) Best Estimate and GRIT papers, referred to in section 3.5 and for which further details are provided in Appendix III, provide a more exhaustive list of some of the considerations that may cause Proxy Methods, or any method of setting provisions, to produce misleading results.
16
APPENDIX I QUESTIONS WE ASKED CONSULTANCIES ABOUT PROXY METHODS

We asked each of the consultancies described in section 1 the following questions, either over the phone or by way of e-mail questionnaire: Q1. What proxy methods have you seen in the UK? Q2. What do you use as reasonableness checks? Q3. What issues do you see in using Proxy Methods? Q4. How many companies do you think use Proxy Methods - number and % of market?
17
APPENDIX II SAMPLE BENCHMARK INFORMATION FROM LLOYDS OF LONDON

Section 4.4 describes the Lloyds reserve benchmarking exercise. A sample showing the type of report they produce (excluding the detailed Appendices) is attached.
18
Agent Reserve Benchmarking Pack

Dummy Agent
Market Reserving and Capital
19
Disclaimer
No responsibility of liability is accepted by the Society of Lloyds, the Council, or any Committee of Board constituted by the Society of Lloyds or the Council or any of their respective members, officers, or advisors for any loss occasioned to any person acting or refraining from action as a result of any statement, fact, figure or expression of belief contained in this document or communication. The views expressed in the paper are Lloyds own. Lloyds provides the material contained in this document for general information purposes only. Lloyds accepts no responsibility, and shall not be liable for any loss which may arise from reliance upon the information provided.
GI PM EG Report version 5.doc
20
Contents
1 2 Introduction Executive Summary 4 5 6 6 7 8 8 9 10 10 12 12 14 16 18 20 22 23 24 27 27 27 27 28 28 29 30 30 30 30 31 31 31 32 34 34 35 36 36 37 38 41 44
3 Reserve Benchmarking Index 3.1 Gross Reserve Benchmarking Index for YOA 1993 2006 3.2 Net Reserve Benchmarking Index for YOA 1993 2006 4 Portfolio Summary 4.1 Market Share by Class of Business 4.2 Net Ultimate Claims and Reserves Analysis 5 Premiums and Claims Development Pattern 5.1 Premiums and Claims Development Patterns 6 Key Performance indicators (KPIs) 6.1 Reserve Strength KPIs 6.2 IBNR Utilisation KPIs 6.3 Reserving over time KPIs 6.4 Quality of Business 6.5 Reinsurance KPIs 6.6 Ultimate Loss Ratio 7 8 Anomalies Guide to Exhibits
9 Data and Methodology 9.1 Introduction 9.2 The Data 9.3 The Benchmark Portfolio calculation 9.4 Numerical Example of Benchmark Calculation 9.5 Deviations, the KPI and Index Calculations 9.6 The Ratio Formulae Appendix A - Definition of Key Performance Indicators A.1 Reserve Strength A.2 IBNR Utilisation A.3 Reserving Over Time A.4 Quality of Business A.5 Reinsurance A.6 Ultimate loss ratio Appendix B - Data Adjustments Appendix C - Reserve Benchmarking Index (Closed Years) C.1 Gross Reserve Benchmarking Index for Naturally Closed Years (1993 2003) C.2 Net Reserve Benchmarking Index for Naturally Closed Years (1993 2003) Appendix D Detailed KPI Analysis D.1 2006 As At Year Summary - Years of Account Grouped by Cycle D.2 2006 As At Year Summary - Years of Account Grouped by Naturally Open/Closed Years D.3 Detailed Summaries over Time - Years of Account Grouped by Cycle D.4 Detailed Summary - Years of Account Grouped by Naturally Open/Closed Years Appendix E List of Classes of Business with Risk Codes
21
1 Introduction
The Agent Reserve Benchmarking Pack has several important purposes. These include: to allow agents and MRC to gain an understanding of the agents reserve levels relative to the market; to be an internal tool for agents to help in analysing their reserves and reserve strength; to bring together comparative analysis carried out by MRC into one document; to identify agents whose reserving history appears consistently stronger or weaker than average; to act as a basis for discussion with agents about their reserving; to provide information to ICA reviewers.
Throughout the pack, data relating to your business is described as Own Portfolio. This includes: Agent Name Dummy Agent Syndicate(s) a, b, c
The general approach used in producing the pack is as follows: The agents Own Portfolio and a Market Portfolio Benchmark were compared. The Market Portfolio Benchmark is calculated from the market data for all other syndicates adjusted to reflect the underlying mix of business in the agents Own Portfolio. Details of the construction of the Benchmark are given in section 9.3; Key Performance Indicators (KPIs) have been calculated to give indicative measures of various items regarding reserve strength and nature of the business; The deviations of the KPIs have been ranked and then combined into various indices to give a constructive insight into relative reserve strengths for the agents, with the Reserve Benchmarking Index (RBI) as an overall summary. However, the RBI should be used in conjunction with its various components.
The data used in this pack is from the 2006 year end Solvency and Reserving Data (SRD) in respect of a particular agent and the market. Apart from where specifically stated, the data within the pack all relates to earned business only to avoid unnecessary distortions caused by annual accounting data items. The pack contains tables and charts detailing the statistics of the agents performance relative to the market benchmark performance. A guide to the exhibits in the pack is provided in Section 8. The guide contains a short description of the information in each section of the pack. Sections 2-6 contain the core analysis work, ranging from a summary level analysis to a more detailed level of data granularity. Sections 7-9 contain information concerning KPI anomalies, the guide to exhibits and calculation methods within the pack. Finally, the appendices contain definitions of KPIs, details of data adjustments, additional analysis at a more granular level and details of the business categorisations used within this pack. MRC contacts are available to discuss the content of the Pack and observations made within it with agents. Currency Units: All data within this pack is represented in converted m unless otherwise stated. Amounts were converted to Sterling using the year end exchange rates GBP1 = USD1.96 = CAD2.28 as set out in Market Bulletin Y3939. Naturally Closed Years: Some exhibits within this report compare the performance of year of account (YOA) categorised as naturally closed, MRC defines this as YOA prior to 2004 with 2006, 2005 and 2004 YOA being categorised as naturally open. Thus some open years are recorded in the naturally closed category.
22
2 Executive Summary
Key Performance Indicators (KPIs) have been calculated to give indicative measures of various items regarding reserve strength and the nature of the business. A Reserve Benchmarking Index (RBI) has been constructed using these KPIs for each YOA on both gross and net of reinsurance basis. These are represented in Chart 2.1. The quartiles measure the relative performance of the agent concerned against all other agents, 1st being highest and 4th the lowest quartile performance rankings. The RBI quartile for all YOA for both gross and net of reinsurance basis is shown in table 2.1. This table also contains the corresponding main components of the index. This index is shown in more detail in Section 3.
Table 2.1 Quartile Ranking of Index Relative quartile position for the RBI and its components of the Own Portfolio.
Gross
Net
RESERVE BENCHMARKING INDEX (RBI)

1. Reserve Strength 2. IBNR Utilisation 3. Reserving Over Time 4. Quality Of Business
Chart 2.1 Reserve Benchmarking Indices (RBI) as at Year end 2006 split by Year of Account The chart displays the RBI quartiles for each YOA on both gross and net of reinsurance basis.
1st Quartile 2nd Quartile 3rd Quartile 4th Quartile 1993 to 1996 1997 1998 1999 2000
Gross
2001 YOA
2002
Net
2003
2004
2005
2006
23
3 Reserve Benchmarking Index

The tables and charts below show the RBI quartiles for the Own Portfolio together with supporting information. The RBI is derived from four sub-indices; each in turn derived from KPI deviation statistics. These measures are reported for the period 1993-2006 combined, assessed as at year end 2006. For the KPI within each sub-index the table shows the Own Portfolio value, the Benchmark value and the deviation between them. The first chart of each section shows the quartile positions of the RBI and sub-indices for each YOA, and the second chart shows the same components measured at as dates 2001 to 2006. More detailed analysis by YOA can be found in section 6 while details of the calculation methodology can be found in Section 9. The corresponding analysis for naturally closed years (1993-2003) is shown in Appendix C.
3.1
Gross Reserve Benchmarking Index for YOA 1993 2006
Table 3.1.1 Key Performance Indicators (KPIs)

GROSS Reser ve Ben chmar kin g In dex
1. Reserve Strength
1.1 Reserve as % Ultimate Claims 1.2 IBNR as % Ultimate Claims 1.3 Case Reserve as a % of Incurred 1.4 Survival Ratio (years) 2.1 IBNR Burn (1 year) During 2006 2.2 IBNR Burn (1 year) During 2005 2.3 IBNR Burn (1 year) During 2004 3.1 Ultimate(@now) as % Ultimate(@YOA) 3.2 Ultimate(@now) as % Ultimate(@YOA+2) 4.1 Paid Loss Ratio
Quartile Portfolio Benchmark

31.6% 12.2% 22.1% 2.3 50.3% 47.4% 20.6% 97.3% 93.2% 62.9% 29.1% 12.8% 18.6% 2.3 35.8% 39.2% 30.4% 98.8% 99.4% 60.3%
Deviation
2.5% (0.6%) 3.4% 0.0 (14.5%) (8.1%) 9.8% 1.6% 6.1% (2.6%)
2. IBNR Utilisation
3. Reserving over time
4. Quality of business
Ot her KPIs
5. Reinsurance
N/A
5.1 % Premium Ceded 5.2 RI ULR 5.3 RI Ultimate as % of Gross Ultimate Claims 5.4 RI Reserve as % Gross Reserve 6.1 ULR
32.8% 97.3% 34.7% 28.1% 91.9%
30.5% 100.9% 36.2% 28.3% 85.0%
2.3% (3.6%) (1.5%) (0.1%) (6.9%)
6. Ultimate Loss Ratio
N/A
Chart 3.1.1 Gross RBI Index & Components by Year of Account

1st Quartile
2nd Quartile
3rd Quartile 4th Quartile 1993 to 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006
YOA
RBI Reserving over-time Reserve Strength Quality of Business Index IBNR Utilisation RBI All YOA
Chart 3.1.2 Gross RBI Index & Components by "as at Year End"
1st Quartile
2nd Quartile
3rd Quartile
4th Quartile 2001 RBI Reserving over-time 2002 2003 2004 2005 IBNR Utilisation 2006
As at Year
Reserve Strength Quality of Business Index
24
3.2
Net Reserve Benchmarking Index for YOA 1993 2006
Tables 3.2.1 Key Performance Indicators (KPIs)

Net Reser ve Benchmar king Index
1. Reserve Strength
1.1 Reserve as % Ultimate Claims 1.2 IBNR as % Ultimate Claims 1.3 Case Reserve as a % of Incurred 1.4 Survival Ratio (years) 2.1 IBNR Burn (1 year) During 2006 2.1 IBNR Burn (1 year) During 2005 2.1 IBNR Burn (1 year) During 2004 3.1 Ultimate(@now) as % Ultimate(@YOA) 3.2 Ultimate(@now) as % Ultimate(@YOA+2) 4.1 Paid Loss Ratio
Quartile Portfolio Benchmark

34.8% 16.2% 22.3% 3.1 44.5% 40.6% 24.0% 96.5% 100.7% 58.2% 32.7% 16.1% 19.8% 3.4 29.5% 33.5% 28.0% 94.4% 99.5% 52.5%
Deviation
2.1% 0.1% 2.5% (0.2) (15.0%) (7.1%) 4.0% (2.1%) (1.3%) (5.7%)
2. IBNR Utilisation
3. Reserving over time 4. Quality of business
Ot her KPIs
6. Ultimate Loss Ratio
N/A
6.1 ULR
89.2%
78.0%
(11.3%)
Chart 3.2.1 Net RBI Index & Components by Year of Account
1st Quartile 2nd Quartile 3rd Quartile 4th Quartile 1993 to 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006
YOA
RBI Reserving over-time Reserve Strength Quality of Business Index IBNR Utilisation RBI All YOA
Chart 3.2.2 Net RBI Index & Components by "as at Year End"
1st Quartile
2nd Quartile
3rd Quartile
4th Quartile 2001 RBI Reserving over-time 2002 2003 2004 IBNR Utilisation 2005 2006
As at Year
Reserve Strength Quality of Business Index
25
4
4.1
Portfolio Summary
Market Share by Class of Business
The tables and charts below show the mix of classes of business and the market share of the Own Portfolio for each year of account measured by the gross cumulative written premium as at year end 2006. Table 4.1.1 Own Portfolio Mix by Lloyds High-Level Class of Business
This table shows for each YOA, the mix of the agents business expressed by the Lloyds high level class of business.
YOA Property (D&F)
Property Treaty 7.5% 5.0% 4.4% 5.4% 4.4% 7.5% 8.4% 2.8% 5.0% 7.7% 10.5% 24.2% 30.5% 35.0%
Casualty 14.4% 5.0% 7.5% 9.3% 22.9% 7.4% 8.2% 9.3% 12.4% 10.8% 14.2% 10.3% 9.3% 6.7%
Casualty Treaty 0.5% 0.5% 0.8% 0.8% 1.1% 1.6% 1.5% 0.8% 0.8% 3.6% 4.0% 4.7% 6.8% 7.5%
Aviation 5.4% 7.5% 5.6% 4.6% 6.6% 10.1% 10.6% 5.6% 4.3% 6.5% 9.0% 9.4% 7.4% 7.3%
Marine 31.0% 36.8% 49.9% 56.0% 36.4% 39.6% 33.1% 33.1% 32.3% 26.3% 23.9% 18.6% 16.0% 18.8%
Energy 10.3% 14.0% 15.8% 16.7% 13.8% 15.0% 13.9% 15.0% 10.6% 5.3% 6.7% 6.6% 7.0% 7.4%
UK Motor 6.0% 11.2% 9.5% 0.0% 2.1% 6.0% 7.1% 7.9% 7.4% 4.9% 0.0% #N/A #N/A #N/A
Overseas Motor 0.4% 0.2% 0.2% 0.0% 0.2% 0.7% 2.2% 3.9% 3.7% 1.8% 0.5% 0.3% 0.5% 0.5%
Accident & Health 13.1% 14.8% 0.0% 0.1% 2.3% 2.0% 4.0% 5.9% 6.2% 4.4% 6.8% 6.8% 5.7% 3.7%
1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006
11.4% 4.9% 6.3% 7.0% 10.2% 10.2% 11.0% 15.8% 17.3% 28.6% 24.5% 19.1% 16.8% 13.1%
Chart 4.1.1 - Own Portfolio Mix by Lloyds High-level Class of Business

100% 90% 80% % Own Portfolio 70% 60% 50% 40% 30% 20% 10% 0% 1993 1994 1995 1996 1997 1998 1999 YOA Property (D&F) Aviation Overseas Motor Property Treaty Marine Accident & Health Casualty Energy Market Share by GCWP Casualty Treaty UK Motor 2000 2001 2002 2003 2004 2005 2006 0.0% 4.0% 2.0% 6.0% 10.0% 8.0% 12.0%
Table 4.1.2 Own Portfolio Market Share Position by Lloyds High-Level Class of Business
This table shows the classes of business for which the Own Portfolio has significant market share for each YOA. It also shows the Own Portfolio aggregate market share across all classes of business.
YOA Market Share by GCWP Property (D&F) Property Treaty Casualty Casualty Treaty Aviation Marine Energy UK Motor Overseas Motor Accident & Health
1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006
9.6% 5.4% 3.4% 2.4% 3.7% 2.8% 3.0% 3.1% 3.5% 4.5% 4.6% 5.5% 6.4% 6.9%
TOP5 TOP20 TOP20 TOP20 TOP20 TOP20 TOP20 TOP5 TOP10 TOP10 TOP10 TOP10
TOP5 TOP20 TOP20 TOP20 TOP20 TOP20 TOP20
TOP20 TOP10 TOP5 TOP5 TOP5
TOP5 TOP20 TOP20 TOP20 TOP10 TOP20 TOP20 TOP20 TOP20 TOP20 TOP20 TOP20 TOP20 TOP20
TOP10 TOP10 TOP10 TOP20 TOP20 TOP20 TOP10 TOP10 TOP10 #VALUE! #VALUE! #VALUE!
TOP20
Market Share
TOP5 TOP5
TOP20 TOP20 TOP10 TOP10 TOP10 TOP20 TOP20 TOP20 TOP10
TOP10 TOP10 TOP10 TOP5 TOP5 TOP5 TOP5
26
4.2
Net Ultimate Claims and Reserves Analysis
The tables and charts below show the components of the net ultimate claims and reserves for each YOA, and how the reserves are split by class of business. Table 4.2.1 Components of Agents Net Ultimate Claims This table shows the components of the Own Portfolio net ultimate claims for each YOA (in converted million).
YOA 1993 to 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 Paid 551 120 137 194 216 230 108 108 291 281 16 Outstanding 22 10 17 28 52 44 35 42 88 267 42 IBNR (2) 6 6 11 8 35 39 59 95 173 127 Ultimate 571 137 160 233 277 308 182 210 474 721 184
Total
2,253
646
556
3,456
Note: Ultimate here is the ultimate claims arising from earned premiums. No allowance has been made for claims arising from unearned premium.
Chart 4.2.1 Net Reserves by High Level Line This chart shows the net reserves for YOA 1993 to 2006 as at year end 2006 split within the pie by Lloyds high level class of business and the largest component of the first pie is analysed into Lloyds Low Level classes of business in the column chart
Energy 8.8% Casualty Treaty 9.9%
Aviation 6.1% Accident & Health 3.3% Other 1.1% NM General Liability (US direct) 6.6%
Property Treaty 14.8% Casualty 21.1%
Professional Indemnity (US) 5.5% Medical Malpractice 2.6% Professional Indemnity (non-US) 1.6% Employers Liability/ WCA (US) 1.4% Other 3.5%
Property (D&F) 16.4% Marine 18.5%
27
5 Premiums and Claims Development Pattern

5.1 Premiums and Claims Development Patterns
The tables and charts in this section show the development of the net ultimate claims and the net ultimate premiums for each year of account from the Own Portfolio data. Both the ultimate premium and claims positions given in section 5 include unearned provisions.
Table 5.1.1 Net Ultimate Premium Development (m)

This table shows the net ultimate premium for each YOA as at each year end. This gives an indication of how well the future premiums were estimated at each development year and the level of confidence in the current estimate.
YOA 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006
'As at' 1993 1994 1995 1996 1997 1998 1999 389 455 458 458 458 459 460 227 254 267 269 270 270 126 148 173 174 173 47 96 102 107 70 105 177 71 112 136
Year 2000 2001 2002 2003 2004 2005 2006 459 450 451 451 450 449 449 266 266 267 267 266 265 265 173 172 173 173 172 172 172 107 107 107 107 106 107 107 178 178 180 180 177 178 178 112 111 111 108 107 108 108 138 137 152 149 148 148 148 147 175 181 170 174 180 175 219 243 230 224 225 230 329 331 341 336 336 430 385 384 380 514 586 567 438 495 562
Table 5.1.2 - Net Ultimate Claims Development (m)

This table shows the net ultimate claims for each YOA as at each year end. This gives an indication of how well the claims reserves were estimated at each development year. Decreasing ultimate values can indicate previous strong reserves or a weakening in reserving basis.
YOA 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006
'As at' 1993 1994 1995 1996 1997 1998 1999 59 146 330 316 311 255 272 60 175 176 167 166 165 86 83 100 93 87 29 62 77 73 53 88 154 75 130 153
Year 2000 2001 2002 2003 2004 2005 2006 282 266 269 281 277 285 267 166 166 167 161 161 166 161 85 86 84 81 80 79 80 73 71 70 70 70 69 69 164 181 147 131 130 132 133 138 147 152 156 158 153 160 166 204 227 231 227 231 233 150 244 262 255 282 276 278 324 332 308 292 291 308 246 219 172 164 182 371 235 211 212 480 482 478 788 753 443
28
Chart 5.1.1 - Net Ultimate Premium Development

This chart shows the development of the Own Portfolio Net ultimate premium for each YOA. The figures used in this chart are shown in Table 5.1.1
700 600 500 400 m 300 200 100 0 1 2 3 1993 2000 4 1994 2001 5 6 7 8 9 D e v e lo p m e n t Y e a r 1996 2003 1997 2004 10 11 1998 2005 12 1999 2006 13 14
1995 2002
Chart 5.1.2 - Net Ultimate Claims Development

This chart shows the development of the Own Portfolio Net ultimate claims for each YOA. The figures used in this chart are shown in Table 5.1.2
900 800 700 600 m 500 400 300 200 100 0 1 2 3 1993 2000 4 1994 2001 5 6 7 8 9 D e v e lo p m e n t Y e a r 1996 2003 1997 2004 10 1998 2005 11 12 1999 2006 13 14
1995 2002
29
6 Key Performance indicators (KPIs)

The tables within this section show a range of reserving KPIs for the Own Portfolio and the Benchmark Portfolio for each YOA. In tables that show deviations between the Own Portfolio and the Benchmark Portfolio these are colour coded with green indicating a favourable deviation and red an unfavourable deviation. For details of the measures and formulas, used to calculate the results below, please refer to Section 9 and Appendix A.
6.1
Reserve Strength KPIs
Table 6.1 - Reserve Strength Percentile Rank

YOA Gross Percentile Rank 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 1993 to 2006
Percentile Rank
Table 6.1.1 - Reserve as % of Ultimate Claims

YOA Gross Portfolio Benchmark Deviation Portfolio Net Benchmark Deviation 1993 4.0% 3.2% 0.8% 4.7% 3.1% 1.5% 1994 3.1% 3.6% (0.5%) 3.0% 3.6% (0.6%) 1995 3.1% 3.8% (0.7%) (4.1%) 4.0% (8.2%) 1996 6.8% 5.6% 1.2% 8.2% 7.0% 1.1% 1997 7.3% 10.3% (3.0%) 11.8% 12.5% (0.7%) 1998 9.4% 8.6% 0.8% 14.0% 11.3% 2.7% 1999 18.1% 11.6% 6.5% 16.8% 15.4% 1.4% 2000 28.1% 20.1% 8.0% 21.8% 19.4% 2.5% 2001 29.3% 26.3% 2.9% 25.4% 28.6% (3.2%) 2002 38.4% 33.1% 5.4% 40.5% 32.6% 7.9% 2003 46.4% 42.2% 4.2% 48.4% 43.4% 5.0% 2004 37.6% 34.5% 3.1% 38.6% 38.7% (0.2%) 2005 52.9% 50.9% 2.0% 61.0% 60.8% 0.2% 2006 91.8% 88.2% 3.7% 91.5% 87.5% 4.1% 1993 to 2006 31.6% 29.1% 2.5% 34.8% 32.7% 2.1%
Table 6.1.2 - IBNR as % of Ultimate Claims

YOA Gross Portfolio Benchmark Deviation Portfolio Net Benchmark Deviation 1993 0.4% 0.8% (0.4%) 0.4% 1.0% (0.5%) 1994 0.2% 0.9% (0.7%) 0.0% 0.9% (0.8%) 1995 0.7% 0.9% (0.2%) 1.1% 1.1% (0.0%) 1996 1.6% 1.6% 0.0% 2.4% 2.3% 0.1% 1997 1.3% 2.6% (1.3%) 2.0% 3.8% (1.8%) 1998 2.0% 2.4% (0.4%) 3.6% 3.8% (0.2%) 1999 5.8% 3.6% 2.2% 4.9% 5.7% (0.8%) 2000 2.6% 5.2% (2.6%) 3.3% 6.7% (3.5%) 2001 9.3% 9.0% 0.3% 11.3% 11.4% (0.1%) 2002 17.8% 14.9% 2.9% 21.3% 15.5% 5.8% 2003 24.8% 22.0% 2.9% 28.4% 21.8% 6.5% 2004 16.8% 16.5% 0.3% 20.0% 20.0% (0.0%) 2005 18.1% 19.3% (1.1%) 23.9% 26.5% (2.6%) 2006 69.0% 67.5% 1.5% 68.8% 66.4% 2.4% 1993 to 2006 12.2% 12.8% (0.6%) 16.2% 16.1% 0.1%
Table 6.1.3 Case reserves as % of Incurred Claims

YOA Gross Portfolio Benchmark Deviation Portfolio Net Benchmark Deviation 1993 3.7% 2.4% 1.2% 4.3% 2.2% 2.1% 1994 2.9% 2.7% 0.2% 2.8% 2.7% 0.1% 1995 2.5% 2.9% (0.5%) 2.7% 3.0% (0.3%) 1996 5.3% 4.1% 1.3% 5.9% 4.8% 1.0% 1997 6.1% 7.9% (1.8%) 7.5% 9.1% (1.6%) 1998 7.5% 6.3% 1.2% 10.8% 7.7% 3.1% 1999 13.1% 8.3% 4.8% 12.5% 10.3% 2.2% 2000 26.2% 15.6% 10.5% 19.4% 13.5% 5.9% 2001 22.0% 19.0% 3.0% 16.0% 19.4% (3.4%) 2002 25.1% 21.3% 3.8% 24.3% 20.2% 4.1% 2003 28.7% 25.9% 2.8% 27.9% 27.6% 0.4% 2004 25.0% 21.6% 3.4% 23.2% 23.4% (0.1%) 2005 42.4% 39.1% 3.3% 48.7% 46.6% 2.1% 2006 73.7% 63.6% 10.1% 72.9% 62.8% 10.1% 1993 to 2006 22.1% 18.6% 3.4% 22.3% 19.8% 2.5%
Table 6.1.4 - Survival Ratio (measured in years)

YOA Gross Portfolio Benchmark Deviation Portfolio Net Benchmark Deviation 1993 10.8 12.6 (1.8) 13.8 15.4 (1.6) 1994 44.3 26.3 18.0 20.1 (78.1) 98.2 1995 3.8 7.1 (3.4) (2.1) 13.0 (15.1) 1996 6.0 8.1 (2.1) 6.8 4.4 2.4 1997 1.7 3.4 (1.7) 2.4 4.2 (1.9) 1998 4.0 4.7 (0.7) 3.9 4.8 (0.8) 1999 9.6 5.3 4.3 6.2 7.5 (1.3) 2000 9.2 6.6 2.6 4.9 4.8 0.0 2001 6.9 5.6 1.4 4.3 5.3 (0.9) 2002 4.4 4.3 0.1 4.2 4.6 (0.3) 2003 5.9 5.4 0.4 6.1 6.1 0.0 2004 1.8 1.7 0.1 2.5 2.3 0.1 2005 1.3 1.4 (0.0) 2.1 2.3 (0.2) 2006 11.3 7.5 3.8 10.8 7.0 3.8 1993 to 2006 2.3 2.3 0.0 3.1 3.4 (0.2)
Net
30
Chart 6.1 - Reserve Strength Percentile Rank

1st Quartile 2nd Quartile 3rd Quartile 4th Quartile
1993 to 1996 1997 1998 1999 2000 2001 Gross 2002 Net 2003 2004 2005 2006
YOA
Chart 6.1.1 Deviations of Reserve as a % of Ultimate Claims Ratios

9% 7% 5% 3% 1% (1%) (3%) (5%) 1993 to 1996 1997 1998 1999 2000 YOA 2001 Gross 2002 Net 2003 2004 2005 2006
Chart 6.1.2 Deviations of IBNR as a % of Ultimate Claims Ratios

9% 7% 5% 3% 1% (1% ) (3% ) (5% ) 1 9 9 3 to 1 9 9 6 1997 1998 1999 2000 YOA 2001 G ros s 2002 Ne t 2003 2004 2005 2006
Chart 6.1.3 Deviations of Case reserves as % of Incurred Claims Ratios

15% 10% 5% 0% (5% ) 1 9 9 3 to 1 9 9 6 1997 1998 1999 2000 YOA 2001 G ros s 2002 Ne t 2003 2004 2005 2006
Chart 6.1.4 Deviations of Survival Ratios

3 .0 1 .0 ( 1 .0 ) ( 3 .0 ) ( 5 .0 ) 1 9 9 3 to 1 9 9 6 1997 1998 1999 2000 YOA 2001 G ros s 2002 Ne t 2003 2004 2005 2006
31
6.2
IBNR Utilisation KPIs
The IBNR utilisation is a measure of how much of the IBNR at the beginning of a year has been incurred during that year. IBNR utilisation gives an idea of how quickly reserves are being used. The tables in this section show IBNR utilisation during 2004, 2005 and 2006 for each YOA. The calculations in section 6.2 include unearned provisions.
Table 6.2 - IBNR Utilisation Percentile Rank

Percentile Rank
Table 6.2.1 - IBNR Burn (1 year) during 2006

YOA Gross Portfolio Benchmark Deviation Portfolio Net Benchmark Deviation 1993 (696%) (11%) 685% (1461%) (2%) 1460% 1994 (63%) (35%) 28% (204%) (69%) 135% 1995 28% (41%) (70%) 86% (49%) (134%) 1996 (23%) (12%) 11% (32%) 19% 51% 1997 9% (12%) (21%) 77% (4%) (81%) 1998 77% (16%) (93%) 157% (17%) (174%) 1999 8% 1% (7%) 21% (5%) (25%) 2000 15% 4% (12%) 57% 17% (41%) 2001 18% 12% (5%) 42% 15% (28%) 2002 17% 20% 3% 58% 15% (44%) 2003 25% 20% (6%) 19% 17% (2%) 2004 43% 29% (14%) 36% 24% (12%) 2005 65% 48% (17%) 53% 39% (14%) #N/A #N/A 2006 #N/A #N/A 1993 to 2005 50% 36% (15%) 44% 29% (15%)

YOA Gross Portfolio Benchmark Deviation Portfolio Net Benchmark Deviation 1993 (133%) 11% 144% 330% (63%) (393%) 1994 108% (34%) (142%) 276% (59%) (335%) 1995 (23%) (7%) 17% (20%) (21%) (1%) 1996 20% (34%) (54%) 34% (97%) (131%) 1997 21% 12% (9%) 0% 16% 16% 1998 (54%) (52%) 2% 32% (2%) (35%) 1999 21% 25% 4% 57% 31% (25%) 2000 47% 5% (43%) 41% 8% (33%) 2001 29% 30% 1% 33% 23% (10%) 2002 50% 21% (29%) 2% 21% 19% 2003 21% 24% 3% 19% 23% 3% 2004 69% 60% (10%) 53% 47% (5%) #N/A #N/A #N/A #N/A 2005 #N/A #N/A 2006 #N/A #N/A 1993 to 2004 47% 39% (8%) 41% 34% (7%)

YOA Gross Portfolio Benchmark Deviation Portfolio Net Deviation 1993 (3%) (42%) (39%) 185% (336%) 1994 35% 30% (5%) 1968% (59%) (2027%) 1995 (24%) (12%) 13% (13%) (5%) 9% 1996 17% 12% (5%) (4%) 6% 10% 1997 (7%) (33%) (27%) 20% (40%) (60%) 1998 23% 13% (10%) (0%) 20% 20% 1999 28% 17% (11%) (2%) 14% 16% 2000 4% 30% 26% 65% 38% (27%) 2001 22% 40% 17% 26% 32% 5% 2002 15% 37% 22% 33% 33% (0%) 2003 26% 31% 5% 22% 30% 8% #N/A #N/A #N/A #N/A #N/A #N/A 2004 #N/A #N/A 2005 #N/A #N/A 2006 #N/A #N/A 1993 to 2003 21% 30% 10% 24% 28% 4%
Net
Benchmark (151%)
32
Chart 6.2 - IBNR Utilisation Percentile Rank

1993 to 1996 1997 1998 1999 2000 2001 Gross 2002 Net 2003 2004 2005 2006
YOA
Chart 6.2.1 Deviations of IBNR Burn during 2006 Ratios

100% 50% 0% (50%) (100%) 1993 to 1996 1997 1998 1999 2000 YOA 2001 Gross 2002 Net 2003 2004 2005 2006

20% 0% (20%) (40%) (60%) (80%) (100%) 1993 to 1996 1997 1998 1999 2000 YOA 2001 Gross 2002 Net 2003 2004 2005 2006

20% 0% (20%) (40%) (60%) (80%) (100%) 1993 to 1996 1997 1998 1999 2000 YOA 2001 Gross 2002 Net 2003 2004 2005 2006
33
6.3
Reserving over time KPIs
Ultimate projections can be compared as at the end of 2006 to the corresponding values at historic valuation dates. This is a measure of how the reserves have changed over time. The tables in this section show how the ultimate held now compare to those set at the end of the YOA in question and to those set at the valuation two years after the YOA in question. The calculations in section 6.3 include unearned provisions.
Table 6.3 - Reserving over time Percentile Rank

Percentile Rank
Table 6.3.1 - Reserving over time (Ult Now/Ult as at YOA)

YOA Gross Portfolio Benchmark Deviation Portfolio Net Benchmark Deviation 1993 68.1% 95.7% 27.6% 94.7% 96.0% 1.3% 1994 96.9% 97.3% 0.4% 96.5% 97.2% 0.7% 1995 88.4% 94.7% 6.4% 93.4% 94.7% 1.2% 1996 93.2% 93.3% 0.1% 94.1% 93.9% (0.3%) 1997 82.1% 105.0% 22.9% 81.1% 103.3% 22.2% 1998 99.6% 100.8% 1.1% 116.0% 103.9% (12.0%) 1999 116.7% 115.4% (1.3%) 140.0% 117.2% (22.8%) 2000 198.2% 155.1% (43.1%) 184.6% 133.0% (51.6%) 2001 101.0% 104.7% 3.7% 95.1% 110.1% 15.0% 2002 78.5% 80.4% 1.9% 74.1% 77.9% 3.8% 2003 66.2% 70.6% 4.5% 57.2% 67.2% 10.1% 2004 113.0% 99.4% (13.6%) 99.7% 89.1% (10.6%) 2005 97.0% 94.3% (2.7%) 95.5% 89.6% (5.9%) 2006 1993 to 2005 97.3% 98.8% 1.6% 96.5% 94.4% (2.1%)
Table 6.3.2 - Reserving over time (Ult Now/Ult as at YOA+2)

YOA Gross Portfolio Benchmark Deviation Portfolio Net Benchmark Deviation 1993 68.1% 95.7% 27.6% 94.7% 96.0% 1.3% 1994 96.9% 97.3% 0.4% 96.5% 97.2% 0.7% 1995 88.4% 94.7% 6.4% 93.4% 94.7% 1.2% 1996 93.2% 93.3% 0.1% 94.1% 93.9% (0.3%) 1997 82.1% 105.0% 22.9% 81.1% 103.3% 22.2% 1998 99.6% 100.8% 1.1% 116.0% 103.9% (12.0%) 1999 110.9% 105.0% (5.9%) 114.3% 102.6% (11.6%) 2000 96.9% 96.8% (0.1%) 105.8% 102.9% (2.9%) 2001 100.0% 103.4% 3.4% 100.0% 103.5% 3.5% 2002 109.8% 97.0% (12.8%) 105.8% 96.9% (8.9%) 2003 102.1% 97.2% (4.8%) 100.3% 95.9% (4.4%) #N/A #N/A #N/A #N/A 2004 2005 #N/A #N/A 2006 #N/A #N/A 1993 to 2003 93.2% 99.4% 6.1% 100.7% 99.5% (1.3%)
Net
34
Chart 6.3 - Reserving over time Percentile Rank

1993 to 1996 1997 1998 1999 2000 2001 Gross 2002 Net 2003 2004 2005 2006
YOA
Chart 6.3.1 Deviations of Reserving over time (Ult Now/Ult as at YOA) Ratios
25% 15% 5% (5%) (15%) (25%) (35%) (45%) (55%) 1993 to 1996 1997 1998 1999 2000 YOA 2001 Gross 2002 Net 2003 2004 2005 2006
Chart 6.3.2 Deviations of Reserving over time (Ult Now/Ult as at YOA+2) Ratios
25% 20% 15% 10% 5% 0% (5%) (10%) (15%) 1993 to 1996 1997 1998 1999 2000 YOA 2001 Gross 2002 Net 2003 2004 2005 2006
35
6.4
Quality of Business
The paid loss ratios can give an indication of how profitable the business written has been. The tables and charts in this section show the paid loss ratios for each YOA.
Table 6.4.1 Quality of Business Percentile Rank

Percentile Rank
Table 6.4.2 Paid Loss Ratio

YOA Gross Portfolio Benchmark Deviation Portfolio Net Benchmark Deviation 1993 59.8% 52.9% (6.9%) 56.8% 55.5% (1.3%) 1994 65.8% 55.5% (10.3%) 58.9% 54.3% (4.7%) 1995 42.0% 47.5% 5.5% 44.7% 49.6% 4.9% 1996 59.0% 66.4% 7.5% 59.4% 63.4% 4.0% 1997 86.0% 96.7% 10.7% 67.6% 87.7% 20.0% 1998 149.4% 130.1% (19.2%) 127.3% 113.4% (13.9%) 1999 123.2% 132.4% 9.1% 130.8% 117.8% (12.9%) 2000 129.1% 114.9% (14.3%) 123.8% 111.9% (11.8%) 2001 105.0% 94.9% (10.2%) 99.7% 85.2% (14.6%) 2002 29.2% 33.6% 4.5% 32.3% 39.4% 7.1% 2003 25.1% 29.1% 4.1% 28.6% 31.2% 2.5% 2004 54.3% 50.4% (3.9%) 51.8% 44.7% (7.0%) 2005 74.6% 68.7% (5.9%) 60.7% 39.9% (20.9%) 2006 3.2% 5.2% 2.1% 5.1% 6.3% 1.2% 1993 to 2006 62.9% 60.3% (2.6%) 58.2% 52.5% (5.7%)
Net
36
Chart 6.4.1 Quality of Business Percentile Rank

1993 to 1996 1997 1998 1999 2000 2001 Gross 2002 Net 2003 2004 2005 2006
YOA
Chart 6.4.2 Deviations of Paid Loss Ratios

25% 15% 5% (5%) (15%) (25%) 1993 to 1996 1997 1998 1999 2000 YOA 2001 2002 2003 Gross Net 2004 2005 2006
37
6.5
Reinsurance KPIs
Please note that the following KPIs are for information only, and are not part of the Reserve Benchmarking Index (RBI), the notation RI denotes reinsurance. The level of reinsurance cover can be an indication of the reliance on reinsurance. The level of reinsurance acceptable will vary depending on the type of business written and on the attitude towards risk of the agent. Desirable levels of reinsurance cover are a matter of opinion so the deviations have not been coloured here and are just to provide an indication of the level of reinsurance cover compared to the market.
Table 6.5.1 - % Premium Ceded

YOA Portfolio Benchmark Deviation 1993 36.6% 28.8% 7.8% 1994 33.2% 28.7% 4.5% 1995 32.3% 28.4% 3.9% 1996 30.8% 35.6% (4.7%) 1997 24.2% 36.0% (11.8%) 1998 43.1% 44.4% (1.3%) 1999 36.8% 43.6% (6.8%) 2000 36.0% 38.5% (2.4%) 2001 38.4% 37.2% 1.2% 2002 31.2% 33.6% (2.3%) 2003 29.0% 29.3% (0.4%) 2004 18.8% 26.9% (8.1%) 2005 37.5% 23.3% 14.2% 2006 38.4% 24.0% 14.4% 1993 to 2006 32.8% 30.5% 2.3%
Table 6.5.2 - RI ULR

YOA Portfolio Benchmark Deviation 1993 67.1% 48.1% 18.9% 1994 82.3% 60.8% 21.5% 1995 44.2% 43.5% 0.7% 1996 60.1% 74.3% (14.1%) 1997 143.3% 121.4% 21.9% 1998 187.0% 160.6% 26.4% 1999 139.1% 163.4% (24.4%) 2000 217.2% 151.5% 65.7% 2001 172.0% 144.7% 27.3% 2002 32.2% 34.0% (1.9%) 2003 25.5% 39.0% (13.5%) 2004 99.0% 87.9% 11.1% 2005 162.8% 265.7% (102.9%) 2006 4.2% 24.7% (20.6%) 1993 to 2006 97.3% 100.9% (3.6%)
Table 6.5.3 - RI ultimate as % of Gross Ultimate Claims

YOA Portfolio Benchmark Deviation 1993 39.4% 25.4% 14.0% 1994 40.3% 30.4% 9.9% 1995 32.9% 25.0% 7.9% 1996 29.3% 37.5% (8.2%) 1997 37.3% 40.5% (3.2%) 1998 48.9% 50.1% (1.1%) 1999 34.0% 47.5% (13.5%) 2000 43.6% 40.6% 3.0% 2001 44.5% 41.8% 2.7% 2002 21.2% 22.7% (1.5%) 2003 15.8% 22.7% (6.9%) 2004 21.4% 30.7% (9.3%) 2005 38.6% 44.2% (5.7%) 2006 4.1% 13.4% (9.3%) 1993 to 2006 34.7% 36.2% (1.5%)
Table 6.5.4 - RI Reserve as % Gross Reserve

YOA Portfolio Benchmark Deviation 1993 29.1% 26.1% 3.0% 1994 42.3% 30.7% 11.6% 1995 189.3% 20.0% 169.3% 1996 15.7% 21.3% (5.6%) 1997 (1.1%) 28.0% (29.1%) 1998 23.5% 34.4% (10.9%) 1999 38.9% 30.6% 8.2% 2000 56.2% 42.7% 13.5% 2001 51.7% 36.7% 15.0% 2002 17.0% 23.8% (6.8%) 2003 12.2% 20.4% (8.2%) 2004 19.3% 22.3% (3.0%) 2005 29.2% 33.4% (4.2%) 2006 4.4% 14.1% (9.6%) 1993 to 2006 28.1% 28.3% (0.1%)
38
Chart 6.5.1 Deviations of % Premium Ceded Ratios

15% 10% 5% 0% (5%) (10%) (15%) 1993 to 1996 1997 1998 1999 2000 2001
YOA
2002
2003
2004
2005
2006
Chart 6.5.2 Deviations of RI ULR Ratios

50% 0% (50%) (100%) 1993 to 1996 1997 1998 1999 2000 2001
YOA
2002
2003
2004
2005
2006
Chart 6.5.3 Deviations of RI Ultimate as % of Gross Ultimate Claims Ratios

10% 5% 0% (5%) (10%) (15%) 1993 to 1996 1997 1998 1999 2000 2001
YOA
2002
2003
2004
2005
2006
Chart 6.5.4 Deviations of RI Reserve as % Gross reserve Ratios

20% 10% 0% (10%) (20%) (30%) 1993 to 1996 1997 1998 1999 2000 2001
YOA
2002
2003
2004
2005
2006
39
6.6
Ultimate Loss Ratio
Please note that the following KPI is for information only, and is not part of the Reserve Benchmarking Index (RBI). The ultimate loss ratio can give an indication of how profitable the business written has been. The tables and charts in this section show the ultimate loss ratios for each YOA.
Table 6.6.1: Ultimate Loss Ratio

YOA Gross Portfolio Benchmark Deviation Portfolio Net Benchmark Deviation 1993 62.3% 54.6% (7.7%) 59.6% 57.3% (2.3%) 1994 67.9% 57.6% (10.3%) 60.8% 56.3% (4.5%) 1995 43.3% 49.3% 6.0% 42.9% 51.6% 8.7% 1996 63.3% 70.4% 7.1% 64.7% 68.2% 3.5% 1997 92.8% 107.8% 15.0% 76.7% 100.1% 23.5% 1998 164.8% 142.3% (22.5%) 148.0% 127.8% (20.2%) 1999 150.5% 149.8% (0.7%) 157.1% 139.2% (17.9%) 2000 179.6% 143.7% (35.9%) 158.3% 138.8% (19.5%) 2001 148.5% 128.8% (19.7%) 133.8% 119.3% (14.5%) 2002 47.4% 50.2% 2.9% 54.3% 58.4% 4.1% 2003 46.8% 50.4% 3.6% 55.5% 55.1% (0.4%) 2004 87.1% 77.0% (10.0%) 84.3% 73.0% (11.3%) 2005 158.3% 139.8% (18.6%) 155.6% 101.5% (54.1%) 2006 39.1% 44.4% 5.3% 60.9% 50.6% (10.3%) 1993 to 2006 91.9% 85.0% (6.9%) 89.2% 78.0% (11.3%)
Chart 6.6.1 Deviations of Ultimate Loss Ratios

25% 15% 5% (5%) (15%) (25%) (35%) (45%) (55%) 1993 to 1996 1997 1998 1999 2000 YOA 2001 Gross 2002 Net 2003 2004 2005 2006
40
Anomalies
The tables below detail data elements in the Own Portfolio that have been identified as potential contributors to significant KPI deviations between the Own Portfolio and Benchmark, and are identified as possible anomalies in the data. All KPIs shown are as at 2006. Any feedback on the cause of these anomalies would assist our understanding of the position. Table 7.1 Anomalies at the Whole Account Level
Anomalies found at the Whole Account Level which have been identified by large deviations from the Benchmark.
Own Portfolio Deviation from Benchmark Agent Calc (000s) Ultimate Ultimate Premium Claims (000s) (000s)
YOA
Ratio
Benchmark
Table 7.2 Anomalies at the Lloyds Low Level Class of Business

The anomalies found at the low level classes of business which have been identified by large deviations from the Benchmark.
Own Portfolio Deviation from Benchmark Agent Calc (000s) Ultimate Premium (000s) Ultimate Claims (000s)
Class of Business
YOA
Ratio
Benchmark
41
8 Guide to Exhibits
The following text briefly describes the content of the exhibits in the Agent Reserve Benchmarking Pack. Section 3: Reserve Benchmarking Index Section 3.1: Gross Reserve Benchmarking Index for YOA 1993 -2006, as at year end 2006 This table shows KPI statistics, values are shown for the Own Portfolio and the deviation of this measure from that calculated for the Benchmark portfolio. This shows the index components for YOA 1993-2006, as at year end 2006 This chart shows the index components measured at year ends 2001-2006 for aggregated applicable YOA.
Table 3.1.1 Chart 3.1.1.
Chart 3.1.2
Section 3.2: Net Reserve Benchmarking Index for YOA 1993 -2006, as at year end 2006 Table Chart The same format as those described above for section 3.1. The same format as those described above for section 3.1.
Section 4: Portfolio Summary Section 4.1: Market Share by line of business This table shows for each YOA, for each of 10 high level business categories the mix of the agents business within each YOA, measured by the gross cumulative written premium as at year end 2006. This table shows for each YOA, for each of 10 high level business categories the position of the Own Portfolio by market share for each business category within each YOA. Values shown will be Top 20, Top 10, or Top 5. It also shows the Own Portfolio aggregate market share across all business categories. Shows data from Table 4.1.1 in chart form and additionally shows the market share of the Own Portfolio by YOA.
Table 4.1.1
Table 4.1.2
Chart 4.1.1
Section 4.2: Net Ultimate Claims and Reserves Analysis Table 4.2.1 This table shows the components of the Own Portfolio net ultimate claim values for each YOA This chart shows the Own Portfolio net reserve value as at year end 2006 split within the first pie by high level class of business and then within the column graph the largest component of the first pie is analysed into the Lloyds Low Level classes of business.
Chart 4.2.1
Section 5: Premium and Claim Development Patterns Section 5.1: Premium and Claim Development Patterns Table 5.1.1 Table 5.1.2 Chart 5.1.1 Chart 5.1.2 This table shows for each YOA the development of the Own Portfolio Net ultimate premium. This table shows for each YOA the development of the Own Portfolio Net ultimate claim. This chart shows the data from table 5.1.1 This chart shows the data from table 5.1.2
42
Section 6: Key Performance Indicators Section 6.1: Reserve Strength KPIs The tables within this section show a range of reserving statistics for the Own Portfolio and the Benchmark Portfolio for each YOA. In tables that show deviations between the Own Portfolio and the Benchmark Portfolio these are colour coded with green indicating a favourable deviation and red an unfavourable deviation. For details of the formulas used to calculate the results, please refer to Section 9 and Appendix A. Tables Charts The statistics in this section give an indication of the reserve strength The charts within this section illustrate the figures within the tables.
Section 6.2: IBNR Utilisation KPIs The tables are in the same format as Section 6.1 Tables Charts The statistics in this section give an indication of how quickly the IBNR is being converted into incurred claims. The charts within this section illustrate the figures within the tables.
Section 6.3: Reserving over time KPIs The tables are in the same format as Section 6.1 Tables Charts The statistics in this section give an indication of how the estimated ultimate claims have changed over time. The charts within this section illustrate the figures within the tables.
Section 6.4: Quality of Business The tables are in the same format as Section 6.1 Tables Charts The statistics in this section give an indication of the quality of business from claims paid. The charts within this section illustrate the figures within the tables.
Section 6.5: Reinsurance KPIs The tables are in the same format as Section 6.1 The KPIs in this section are for information only and are not part of the Reserving Benchmark Index. Tables Charts The statistics in this section relate to reinsurance use and performance. The charts within this section illustrate the figures within the tables.
Section 6.6: Ultimate Loss Ratio The tables are in the same format as Section 6.1 The KPIs in this section are for information only and are not past of the Reserving Benchmark Index. Tables Charts The statistics in this section show the ultimate loss ratio. The charts within this section illustrate the figures within the tables.
43
Section 7: Anomalies The table details data elements relating to the Own Portfolio that have been identified as being potential contributors to significant KPI deviations between the Own Portfolio and the Benchmark Portfolio. These are at the whole account level and are as at year end 2006. This is a class of business level summary in the same format as the whole account level summary.
Table 7.1 Table 7.2
Section 9: Data and Methodology This section details the data used, the benchmark calculation and the index calculations.
Appendix A: Definition of Key Performance Indicators This section gives the formulae used to calculate the KPIs used in the pack.
44
Appendix B: Data Adjustments These tables show the data removed from the analysis due to the data failing checks applied by MRC, these checks are described within Section 9. The tables summarise the data removed for as at each year end from 2000-2006, for each YOA on both a gross and net basis.
Tables
Appendix C: Reserve Benchmarking Index (Closed Years) Section C.1: Gross Reserve Benchmarking Index for Naturally Closed YOA, as at year end 2006 The same format as those described above for section 3.1. The same format as those described above for section 3.1. Section C.2: Net Reserve Benchmarking Index for Naturally Closed YOA, as at year end 2006 The same format as those described above for section 3.1. The same format as those described above for section 3.1.
Appendix D: Detailed KPI Analysis Section D.1 2006 as at year Summary -YOA Grouped by Cycle These tables show the KPIs of Section 6 reported as at year end 2006, for YOA grouped to reflect the underwriting cycle, namely 1993-96, 1997-2001, 2002-2004, and then 2005 and 2006 individually.
Tables
Section D.2 2006 as at year Summary -YOA Grouped by Naturally Closed/Open Years These tables show the KPIs of Section 6 reported as at year end 2006, for YOA grouped to reflect the definitions of naturally closed/open years as defined previously within the Purpose section of this pack, namely 1993-03, and then 2004, 2005 and 2006 individually.
Tables
Section D.3 Detailed Summary over Time YOA Grouped by Cycle These tables show the KPIs of Section 6 reported as at year ends 2000-2006 , for YOA grouped to reflect the underwriting cycle, namely 1993-96,1997-2001,2002-2004, and then 2005 and 2006 individually.
Tables
Section D.4 Detailed Summary over Time YOA Grouped by Naturally Closed/Open Years These tables show the KPIs of Section 6 reported as at year ends 2000-2006, for YOA grouped to reflect the definitions of naturally closed/open years as defined previously within the Purpose section of this pack, namely 1993-03, and then 2004, 2005 and 2006 individually.
Tables
Appendix E: List of Classes of Business with Risk Codes This is a summary showing the complete list of classes of business used in creating this pack. The risk codes used for each class of business are listed.
45
9 Data and Methodology

9.1
a) b) c) d)
Introduction
The data used The benchmark calculation The Index calculations The KPI (ratio) formulae
This section provides a description of:
9.2
The Data
This pack utilises the 2006 year end Solvency and Reserving Data (SRD) as the primary data source. This data set is collected annually at risk code level and covers all non-life business on both gross and net of reinsurance basis. Each year incremental data is reported by syndicates and these returns have been aggregated historically to provide cumulative information. The ratios and other measures have been adapted to use the earned position that arises from the annual accounting treatment unless otherwise stated. The figures within this report are necessarily dependant on the quality and accuracy of the underlying data. The SRD also contains potentially anomalous data entries which can distort the calculations for a particular agent. These anomalies relate to negative value data for cumulative premium and claim data, which may or may not be valid but would not be expected to generally occur. For the purpose of our analysis we consider data at a Lloyds Low Level class of business (itself an aggregation of risk codes as detailed in Appendix E) and have analysed the occurrence of negative value data at this level. To avoid such data distorting the analysis we have adjusted data from the agent SRD dataset entries as follows Any data record (for each syndicate, at a low level class, for both gross and net separately) with a significant negative ultimate claim value will have the following components set to zero: paid claims, outstanding claims reserve, IBNR, and ultimate claims. Any data record (for each syndicate, at low level class, for both gross and net separately) with a significant negative ultimate premium value will have the following components set to zero: premium. For these adjustments significant is defined as over 5% of the value of the corresponding market cell. These adjustments in total represent -2.71% of the market total net of reinsurance premiums and 0.83% of the market total net ultimate claim values (as such the quantum of the adjustment is, in our opinion, tolerable).
A summary of the adjustment made for the syndicate(s) included within this pack can be found in Appendix B of this report.
46
9.3
The Benchmark Portfolio calculation
To assess an Agents performance against that of the market, a benchmark is required that takes account of the different mixes of business written by different Syndicates. To construct the benchmark, the market has been segmented into low level classes of business. A list of these classes is shown in Appendix E. The Agent/Syndicates portfolio mix across classes is calculated based on a proportional split of gross ultimate premiums. The benchmarks portfolio mix is calibrated to be equal to the Agent/Syndicate separately for each year of account, each as at year. To measure the benchmarks performance, a notional benchmark syndicate is constructed. This notional syndicate shares the same portfolio mix of the Agent/Syndicate being benchmarked, but exhibits market average performance within each year of account and class of business. This is done by scaling up or down the markets financial results such as claims and reserves according to the mix of business written by the Agent/Syndicate under review. A numerical example of this process is given below. In general, when calculating the benchmark, the Own Portfolio data is excluded from the market data.
9.4
Numerical Example of Benchmark Calculation
Syndicate X has written 200m of business in a particular year of account across 3 classes of business. This is split as follows: 100m (50%) of Employers Liability, 60m (30%) of Professional Indemnity and 40m (20%) of Public Liability. It estimates that its ultimate claims will be 150m, 60m and 15m for each of these three classes respectively. This equates to an overall ULR of 112.5% (225m / 200m). Overall, the rest of the market (excluding Syndicate X) has written 2bn (40%) of Employers Liability, 2bn (40%) of Professional Indemnity and 1bn (20%) of Public Liability. Overall estimated ultimate claims are 3bn, 2bn and 500m for each of these three classes respectively. This equates to loss ratios of 150%, 100% and 50% respectively, or an overall ULR of 110% (5.5bn / 5bn). The two ULRs of 112.5% and 110% are not directly comparable as Syndicate X has written proportionally more Employers Liability than the market a class that has produced a poor result at a market level. We therefore create a benchmark that has written the same mix of business as Syndicate X (i.e. 50% of Employers Liability, 30% of Professional Indemnity and 20% of Public Liability) and has had market average experience (i.e. loss ratios of 150%, 100% and 50% respectively). We can then calculate an overall, weighted average ULR for the notional syndicate as 115%. This is calculated as follows: Gross Benchmark Premium = 100m + 60m + 40m (the same as the syndicate) = 200m Gross Benchmark Claims = 100m*150% + 60M*100% + 40M*50% = 230m Gross Benchmark ULR = 230m / 200m = 115% Whereas initially the Syndicate appeared to have a higher loss ratio than the market, when allowing for the syndicates business mix, it can be seen that the Syndicate actually out-performed the market, achieving equal loss ratios in Employers Liability and Professional Indemnity to the market and achieving a more profitable underwriting result than the market in Public Liability. The weightings have been calculated separately for Gross and Net of reinsurance KPIs. For Net calculations, the benchmark Net of reinsurance premium is set as follows (for a given class of business): Net benchmark premium = Gross Benchmark Premium * (Market net premium/ Market Gross Premium) All KPIs have been calculated using this premium weighting approach. This also needs to be borne in mind when interpreting the results.
47
9.5 9.5.1
Deviations, the KPI and Index Calculations Deviations
For a given ratio we calculate a deviation measure comparing the ratio value for the Own Portfolio to that for the Benchmark. Deviations are either measured as Benchmark Own Portfolio or as Own Portfolio Benchmark. The choice of calculation approach varies from measure to measure but the aim is that a deviation will be positive if the Own Portfolio ratio compares favourably with the Benchmark Portfolio and negative if it does not. For example, consider the measure Ultimate loss ratio, the deviation here is measured as Benchmark Own Portfolio, this will give a positive value if the Own Portfolio ultimate loss ratio is less than that of the Benchmark portfolio and a negative value otherwise. Within this report the deviations are colour coded with green indicating a favourable deviation and red an unfavourable deviation.
9.5.2
Index Calculations
To compare the level of deviation between agents the deviation is converted into a relative value, for example if the deviation is measured as Benchmark Ratio Own Portfolio Ratio this is converted to a relative deviation by dividing the result by the absolute value of the Benchmark Ratio. Note that the deviations shown within the pack are the nominal deviations rather than the relative deviations used to inform the index. The range of relative deviation values across all agents can be ranked for a particular measure. This provides the means to calculate the rank of that agents relative deviation compared to all other agents relative deviation measures. To calculate an index value for a particular year, the ranks of various KPIs are combined into an aggregate score, which itself is then ranked and converted into a percentile value. Similarly the Reserve Benchmarking Index (RBI) is calculated by summing the ranks of the four component indices, again ranking this value and calculating the equivalent percentile. Please note that although unusual, it is possible that the ranking of the RBI may appear to be different to the average of the various KPIs or the four component indices. The RBI values are calculated using the following combinations of ratios shown in Table 9.5.2.1
48
Table 9.5.2.1 Index Component KPIs (ratios) Reserves as % of Ultimate Claims IBNR as % Ultimate Claims Case reserves as % Incurred Survival Ratio Reserve Strength IBNR burn during 2006 IBNR burn during 2005 IBNR burn during 2004 IBNR Utilisation Ultimate (@now)/Ultimate(@YOA) Reserving Over Time Ultimate (@now)/Ultimate(@YOA+2) 1 1 1 1 1 Weighting 2 1 1 1
Quality of Business Paid loss ratio
Index
Component Indices Reserve Strength IBNR Utilisation
Weighting 2 1 1 1
Reserve Benchmarking Index
Reserving Over Time Quality of Business
To calculate the RBI for all years of account (1993-2006) combined, the overall relative deviations for the KPIs are calculated for the combined period 1993-2006. The same process described above is then applied to obtain the rankings and corresponding indices including the RBI for this combined period. As previously mentioned, the ranking of the RBI may appear to be different to the average of the various KPIs or four component indices. Example Index Calculation: Benchmark KPI: 80% Own portfolio KPI: 70%
Deviation measured as Benchmark Own Portfolio Deviation = 80%-70% = 10% Relative Deviation = 10%/80% = 12.5% Suppose the rank of this measure across all agents is 5 where the lower the rank the better, if this is a rank of 5 out of 50 agents then (assuming no duplicate rankings) the equivalent percentile is calculated as 4/(4+45) = 8.1% which in turn would be assigned a quartile grading of 1st Quartile.
49
9.6
The Ratio Formulae
Appendix A sets out the formulae for each KPI (ratio) included in the pack.
50
APPENDIX III BIBLIOGRAPHY

(1) UK Asbestos the definitive guide, J Lowe et al, GIRO 2004, available on the UK Actuarial Professions website at: http://www.actuaries.org.uk/files/pdf/proceedings/giro2004/Lowe.pdf (2) Best Estimate Provisions for General Insurance, Solvency 2 working group chaired by K Morgan, Autumn 2006, available on the UK Actuarial Professions website at: http://www.actuaries.org.uk/files/pdf/general_insurance/best_estimates_paper.pdf (3) A change agenda for reserving. Report of the general insurance reserving issues Taskforce (GRIT), Tony Jones et al, March 2006, available on the UK Actuarial Professions website at: http://www.actuaries.org.uk/files/pdf/sessional/sm20060327.pdf
51

Actuarial Best Estimates and Proxy Methods

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Actuarial Best Estimates and Proxy Methods

Uploaded by

Copyright:

Available Formats

General Insurance Reserving Actuarial Best Estimates and Proxy Methods

Our comments will be in the same context.

UK Expert Group Report On General Insurance Proxy Methods

Prepared by GI Proxy Methods Expert Group Version 5 Published September 2007

GI Proxy Methods Expert Group

Appendix I Appendix II Appendix III

WHAT WEVE DONE

DECISION-TREE FOR PROXY METHODS

TYPES OF PROXY METHOD

ADVANTAGES AND DISADVANTAGES OF PROXY METHODS

APPENDIX I QUESTIONS WE ASKED CONSULTANCIES ABOUT PROXY METHODS

APPENDIX II SAMPLE BENCHMARK INFORMATION FROM LLOYDS OF LONDON

Agent Reserve Benchmarking Pack

GI PM EG Report version 5.doc

GI PM EG Report version 5.doc

GI PM EG Report version 5.doc

RESERVE BENCHMARKING INDEX (RBI)

GI PM EG Report version 5.doc

3 Reserve Benchmarking Index

Gross Reserve Benchmarking Index for YOA 1993 2006

Table 3.1.1 Key Performance Indicators (KPIs)

Quartile Portfolio Benchmark

3. Reserving over time

32.8% 97.3% 34.7% 28.1% 91.9%

30.5% 100.9% 36.2% 28.3% 85.0%

2.3% (3.6%) (1.5%) (0.1%) (6.9%)

6. Ultimate Loss Ratio

Chart 3.1.1 Gross RBI Index & Components by Year of Account

GI PM EG Report version 5.doc

Net Reserve Benchmarking Index for YOA 1993 2006

Tables 3.2.1 Key Performance Indicators (KPIs)

Quartile Portfolio Benchmark

3. Reserving over time 4. Quality of business

Chart 3.2.1 Net RBI Index & Components by Year of Account

GI PM EG Report version 5.doc

Chart 4.1.1 - Own Portfolio Mix by Lloyds High-level Class of Business

TOP5 TOP20 TOP20 TOP20 TOP20 TOP20 TOP20

TOP20 TOP10 TOP5 TOP5 TOP5

TOP20 TOP20 TOP10 TOP10 TOP10 TOP20 TOP20 TOP20 TOP10

TOP10 TOP10 TOP10 TOP5 TOP5 TOP5 TOP5

GI PM EG Report version 5.doc

Net Ultimate Claims and Reserves Analysis

Energy 8.8% Casualty Treaty 9.9%

Property Treaty 14.8% Casualty 21.1%

Property (D&F) 16.4% Marine 18.5%

GI PM EG Report version 5.doc

5 Premiums and Claims Development Pattern

Table 5.1.1 Net Ultimate Premium Development (m)

Table 5.1.2 - Net Ultimate Claims Development (m)

GI PM EG Report version 5.doc

Chart 5.1.1 - Net Ultimate Premium Development

Chart 5.1.2 - Net Ultimate Claims Development

GI PM EG Report version 5.doc

6 Key Performance indicators (KPIs)

Reserve Strength KPIs

Table 6.1 - Reserve Strength Percentile Rank

Table 6.1.1 - Reserve as % of Ultimate Claims

Table 6.1.2 - IBNR as % of Ultimate Claims

Table 6.1.3 Case reserves as % of Incurred Claims

Table 6.1.4 - Survival Ratio (measured in years)

GI PM EG Report version 5.doc

Chart 6.1 - Reserve Strength Percentile Rank

Chart 6.1.1 Deviations of Reserve as a % of Ultimate Claims Ratios