Chapter 7: Analysis of experience Flashcards
Analysis of claims data on their own may be needed for: (3)
- estimating the cost of outstanding claims to set reserves
- monitoring the actual run-off of outstanding claims against estimated amounts
- monitoring the adequacy and use of reinsurance
Analysis of claims data along with data on exposure may be needed for: (5)
- comparing the relative profitability of various parts of the account
- reviewing premium rates, and for pricing new or amended products
- financial planning
- monitoring the insurer’s asset-liability position
- establishing the need for and values of reserves e.g. URR
Estimation of outstanding claims
The level of risk classification required for this depends on the account concerned.
The basic requirement is that there is a reasonable volume of stable, consistent claims data, from which projection patterns and trends can be deduced.
The data ideally needs to be divided into sufficiently homogeneous risk groups. However, this has to be balanced against the danger of creating data cells that have insufficient data in them to be credible for projection purposes.
Issues with the definition of a claim
If the claims department changes the definition of a claim, there is a danger that this will lead to distortions in an analysis unless it is identified within the data.
Issues with Administration
Statistics are also distorted by inevitable delays between the occurrence of an event and its appearance on office records, and by changes to the administration and processing of data.
The delays occur because claims are not reported immediately, and this delay in reporting is not necessarily constant.
Issues with Re-opened claims
Re-opened claims payments should be allocated to the risk group and claim year of the original claim.
If a separate claim is created for the additional amount involved, this could create distortions in the apparent overall experience of the subgroups concerned, the numbers of claims and hence the average cost per claim.
Issues with recoveries
A recovery from a reinsurer should be recorded separately on the same claim record as the gross claim payments.
This allows the analysis of claims both gross and net of reinsurance which, as well as being a statutory requirement, may also help produce more reliable projections.
This is especially important where the reinsurance takes some form of Excess of Loss cover, particularly where there are limited reinstatements.
Possible CAUSES of changes to the administration and processing of data (4)
- backlogs in processing
- computer breakdown
- public holidays
- changes in procedure
Effect of changes in the RATE at which items of claims exposure are processed
Affects claim payment patterns and the apparent in-force position of the portfolio.
Why is it important to be able to separate nil claims from those settled at some cost?
Because nil settlements (if not separated) will change the aggregate number of claims, but the aggregate cost will remain the same.
The effect of this will be to distort:
- claim frequency
- average cost per claim
Issues with partial payments
It is important - for each claim - to record each payment separately, with the date(s) of payment, and the status of the claim (settled or outstanding).
Why is it important to be able to distinguish between payments on settled claims and those still outstanding?
If not, the average cost of settled claims and the claim payment patterns will be distorted.
There is also the danger that, under some projection methods, the value of claims outstanding may be overestimated.
Issues with individual large claims
If left unadjusted in the aggregate data, individual large claims would unduly dominate the experience of the risk group.
Hence a report should be produced detailing all claims that exceed a certain size, so that such claims can be removed or truncated. The threshold points for reporting and truncating will differ according to the class involved and the purpose of the analysis. An example would be to truncate all large claims at say the 95th percentile.
Issues with concentrations of claims
As with individual large claims, it is important that concentrations of claims can be identified so that they can be dealt with separately from the main body of data.
In some cases, this may be possible from the individual claim records, by suitable use of such factors as class, date of claim, peril, type of claim, geographical area, etc.
However, this approach may not be perfect. Therefore, when it is apparent that an event of significant proportions has occurred, it is usual for the insurer to set up a catastrophe code for the event, which is then placed on the claim records for all claims arising from the event.
This will also assist with any reinsurance recovery.
3 Ways in which Exceptional claims may be treated
- If the claims are thought to be completely once-off, and not likely to be repeated in the future, exclude them altogether from the data and account for them elsewhere.
- If the claims are likely to occur occasionally, include only part of the cost of such claims, corresponding to the probability of their recurrence and the length of the periods being analysed and projected.
- Debit the cost of such claims against the specific rating group in which they occurred, but only up to a maximum amount per claim. Any excess above this cut-off amount would be spread over the experience of all other relevant rating groups.
How to adjust the data for a change in zero claims
This cannot be dealt with unless zero claims are recorded separately.
The way to handle these would normally be to exclude them since they will affect the frequency and average cost per claim.
Any change in the way these are recorded should also be noted.
How to adjust for a change in the type of claim
If such a change occurs, it can only be allowed for by projecting separately for the different types of claim.
E.g. if a contents claim was previously grouped under buildings type claims, and is now separated out, this would yield different results than previously.
How to adjust for a change in claim development patterns
If the data are to be used for a projection model, this can be a difficult problem to overcome.
The extent to which the data can be adjusted to make them suitable for projection purposes depends on the cause of the change.
How to adjust for a change in claim development patterns:
If it is due to a change in the mix of risks involved, or the types of claim occurring.
the data can be separated out into the relevant subgroups or claim types.