Risk Disclosures

Insurance liabilities and reinsurance assets

Calculation of IBNR and claims development

Amlin adopts a consistent process in the calculation of an adequate provision for insurance claim liabilities. The overriding aim is to establish reserves which are expected to be at least adequate and that there is consistency from year to year. Therefore the level of reserves are set at a level above the actuarial ‘best estimate’ position. However, there is a risk that, due to unforeseen circumstances, the reserves carried are not sufficient to meet insurance claim liabilities reported in future years on policy periods which have expired.

Process and methodology

The reserving process commences with the proper recording and reporting of claims information which is made up of paid and notified or outstanding claims. For our London Market business information is received through Xchanging (the London Market bureau) and, in the case of our UK commercial business and service companies, directly from policyholders. Claims records are maintained for each class by the underwriting year to which the policy incepts. For notified or outstanding claims a case reserve is established based on the views of underwriting management and claims managers, using external legal or expert advice where appropriate. This reserve is expected to be sufficient to meet the claim payment when it is finally determined. For some classes of business, particularly liability business, settlement may be several years after the initial notification of the claim, as it may be subject to complexities or court action. Underwriters and claims staff are responsible for setting case reserves for outstanding claims. For claims received from Xchanging, the market reserve is generally set by the leading underwriter but there are circumstances on larger claims where Amlin will post higher or lower case reserves than those notified. These cases are explained and discussed in reserving review.

To establish a provision for incurred but not reported (‘IBNR’) claims, the underwriting and claims teams use their experience and knowledge of the class of business to estimate the potential future development of each class for every year of account. The development period varies by class, by method of acceptance and is also determined by the deductible of each policy written. For casualty business the policy form will determine whether claims can be made on a claims made (as advised) or as a losses occurring (determined by date of loss) basis. This has a significant impact on the reporting period in which claims can be notified. In setting the IBNR provision estimates are made for the ultimate premium and ultimate gross claims value for each underwriting year. Allowance is then made for anticipated reinsurance recoveries to reach a net claims position. Reinsurance recoveries are calculated for outstanding and IBNR claims sometimes through the use of historical recovery rates and provisions are made as appropriate for bad debt or possible disputes.

To assist with the process of determining the reserves, triangulation statistics for each class are produced which show the historical development of premium, as well as paid and incurred losses, for each underwriting year, from inception to the date of review. Each class triangulation is also independently analysed by the internal actuarial team using actuarial software as appropriate. The aim of the actuarial exercise is to produce ‘best estimate’ ultimate premium and claims amounts which can be compared to the figures proposed by divisional management. Meetings are held in which executive management, actuarial staff and business management discuss claims issues and analyse the proposed and independently generated reserves to conclude the provision to be carried. These provisions are also reviewed annually by external actuaries who examine the work carried out and opine on the sufficiency of reserves.

Areas of uncertainty

The reserves established can be more or less than adequate to meet eventual claims arising. The level of uncertainty varies significantly from class to class but can arise from inadequate case reserves for known large losses and catastrophes or from inadequate provision for IBNR. The impact on profit before tax of a 1% variation in the total net claims reserves would be £11 million (2005: £11 million).

Large loss case reserves are determined through careful analysis of the individual claim, often with the advice of legal advisers. Liability claims arising from events such as the 11 September 2001 terrorist attacks in the United States is an example of a case where there continues to be some uncertainty over the eventual value of claims.

Property catastrophe claims such as earthquake or hurricane losses can take several months or years to develop as adjusters visit damaged property and agree claim valuations. Until all the claims are settled it requires an analysis of the area damaged, contracts exposed and the use of models to simulate the loss against the portfolio of exposure in order to arrive at an estimate of ultimate loss to the Company. There is uncertainty over the adequacy of information and modelling of major losses for a period of several months after a catastrophe loss. Account should also be taken of factors which may influence the size of claims such as increased inflation or a change in law.

The long tail liability classes, for which a large IBNR has to be established, represent the most difficult classes to reserve because claims are notified and settled several years after the expiry of the policy concerned. This is particularly the case for US liability written on a losses occurring basis.

The use of historical development data adjusted for known changes to wordings or the claims environment is fundamental to reserving these classes. It is used in conjunction with the advice of lawyers and third party claims adjusters on material single claims.

The allocation of IBNR to the reinsurance programme is an uncertain exercise as there is no knowledge of the size or number of future claims advices. The assumption over future reinsurance recoveries may be incorrect and unforeseen disputes could arise which would reduce recoveries made.

In the course of reserving the businesses provide a reserve for future events occurring to the existing portfolio. These provisions are removed in order to reflect GAAP accounting practice.

Dynamic Financial Analysis (DFA) modelling of risk

To improve our risk management capability, and our assessment of capital requirements, Amlin has developed a stochastic model to analyse the potential performance of the underwriting businesses. The output from the model includes a distribution of outcomes from reserves for prior written liabilities, investment performance and new business underwriting performance. The result is a combined view of the expected best estimate mean result and the range of possibilities around it. The analysis quoted below concentrates on underwriting assets and liabilities and thus excludes the investment performance of corporate assets.

The model requires the input of a large number of explicit parameters. Those inputs are based on many different sources of information including detailed historical data on premiums and claims, forecast income and exposures, estimated rating levels and catastrophe loss data from proprietary models applied to Amlin’s portfolio. It enables projection of an estimated mean ultimate loss ratio and the distribution of results around it. The model explicitly recognises diversification credit since class results are not all strongly correlated and thus individual classes are unlikely to all produce losses (or profits) in the same year. Due to the inherent uncertainty of predicting the key drivers of business performance, including in particular claims levels, individual runs of the model cannot be relied upon to accurately forecast outcomes. However, the output from many thousands of simulated results can provide a picture of the possible distribution of insurance business results. This output is useful in developing an understanding of the losses which may be borne by the business at various levels of profitability.

There are a large number of uncertainties and difficulties in achieving accurate results from the model. Some of the key issues are:

  • The model is based on business plan volume and rate expectations which may not be borne out in practice.
  • A significant change in the portfolio of business could result in the past not being a reliable guide to the future.
  • Changing external environmental factors may not be assessed accurately.
  • The potential for model error is significant in such a complex and developing discipline.
  • Key assumptions over levels of correlation between classes may over time prove to be incorrect.
  • Catastrophe model inputs, which estimate the severity and frequency of large catastrophes on the portfolio, may be incorrect.

Amlin Bermuda commenced underwriting late in 2005. As a result, the risk profile is changing in line with the growth of the business. The current assumptions as to the performance of the corresponding Syndicate 2001 classes of business have therefore been used as a proxy for the Amlin Bermuda portfolio.

The result reproduced below represents the modelled loss sustained by the business from a single 1 in 200 bad year i.e. at the 0.5 percentile. This probability is the calculation benchmark required by the FSA and Lloyd’s. However it does not represent the level of capital required for Amlin to support current and expected business levels, which should be considered over a longer period of modelling. Furthermore, Amlin is required to carry (larger) levels of capital which are sufficient in the eyes of rating agencies and clients. This is modelling for a single year’s outcome only. As explained above Group corporate assets are excluded from the analysis.

All figures in £m based on 2007 business plan forecast
 

Risk class after diversification
Underwriting (new business risk) (327)
Reserving (18)
Credit (reinsurance counterparty risk) (33)
Investment (market risk) 55
Diversified result (323)

Notes:

  1. Excluding any additional capital provision for operational risk.
  2. No dividend or tax is considered.
  3. Currency risk is not modelled explicitly.
  4. Non sterling amounts have been converted at Lloyd’s required rates, including for US Dollars $1.77 to £1.0.
  5. The strength of correlation between Amlin Bermuda and Syndicate 2001 has been estimated subjectively.
  6. These figures are based on work for the September 2006 ICA submission and thus include the previously projected year end asset and liability position.
  7. Investment income does not include group corporate assets.
  8. Within reserving risk no credit is taken for investment returns generated on assets backing the insurance liabilities (i.e. there is no discounting). Therefore the investment risk evaluation is shown as a positive, since, given the nature of assets held, even at the 1 to 200 level the modelled investment return is greater than zero.