Source in GRIT Paper

1.8 Our conclusions: Identifying where our reserving methods need to be enhanced

1.8.1 More sophisticated mathematical and statistical methodology need not be a priorty for actuaries at this stage. Rather the focus for enhancement and research should be in the following areas:

• better business understanding and adapting the actuarial methodology in the light of what has been learnt;
• allowing for the underwriting cycle;
• more focus on analysing and investigating the historic fit of the reserving model used and more use of diagnostics;
• whether extreme value theory has a role to play in reserving;
• practically, which methods do well in which circumstances;
• more focus on data quality

Terms of reference

General

1. The Working Party (WP) is constituted as a sub-committee of ROC, and will report to ROC. The chairman of the WP will be a member of ROC and will be responsible for coordination and communication between the WP and ROC.

2. The WP will report progress on a regular basis (eg monthly or bi-monthly) to ROC. This will include actions planned, issues identified, emerging conclusions, and progress against timetable.

3. The deliverables of the WP will include a report for GIRO 2008. A draft of this report and any other deliverables will be provided to ROC in draft for comment before wider distribution by the WP. In any reports or other deliverables or communications made by the WP, the WP should not claim to speak with the authority of ROC or the GIB without prior approval of ROC.

Specific

a. Investigate previous work done in this area make available literature review to point interested actuaries at relevant papers and presentations

b. Identification of the reserving methods that are going to be worked on prioritised list (at specific request of GIB this will include ICRFS)

c. Document the key strengths and likely failures of current reserving methods, in terms of the underlying assumptions behind the models and the circumstances when they are likely to fail, paying particular attention to those methods which have been deficient in practice.

d. Identify and document useful diagnostics for common reserving methods; setting out clearly what the diagnostic is meant to show and appropriate reponses when the diagnostic shows that the data is not a good fit

e. Obtain data for a variety of classes of business, and run a series of reserving exercises as the data progresses to see how well the model would have predicted the reserves at any point. The following list is not intended to be exhaustive but could include the following; for example

• relatively predictable short tail business
• business with a stable development pattern, but where case reserves are set conservatively and then release over time
• long tail business with significant reporting delays and delays in identifying quantum
• extremely volatile business (e.g. high layer excess where large claims appear, disappear, appear, disappear etc etc)
• data exhibiting the 'reserving cycle'
• data where there has been a change in underlying terms and conditions

As well as using 'real' data, it is likely that there will be value in performing these exercises on created data (how well do the different methods perform when there is perfect knowledge about the underlying business and process, and the only volatility in results is the known 'randomness' in claims experience)

f.  If time allows, discussion of some of the areas that the industry has struggled with in recent years (e.g, US Liability 97-01; Katrina) where there have been several attempts to get the answer 'right' and whether the above research suggests that any particular model would have been well suited the circumstances of each loss event.