truncating large claims

Hi all

In chapter 16, the reserving case study, one of the scenarios given is what adjustments might need to be made to a chain ladder approach if there is a change in the mix of business. The answer given is that we might need to sub-divide the data into rating groups, but that since this will reduce the stability of past patterns, we may be required to use lower cut off points to truncate and spread the effect of large claims.

I'm not 100% clear on the last point above -
1. How is the truncate-and-spread method actually implemented? I haven't seen it used in practice. Obviously you truncate the large claims, but where do you spread the truncated amount? Across all other cells in the same column of the triangle?
2. Is the objective of using a lower cut off point to increase the stability of the pattern? Will it really make that much of a difference?

Thanks

 

Still in chapter 16, another part of the question asks what adjustments to the chain ladder method are needed for the introduction in the last year of a new admin system, where all claims with no payments or correspondence for 12 months recorded as settled.

The answer given is that numbers of claims are not used in the chain ladder method, so we will not have to make any adjustment because this administrative change only affects numbers of claims.

I find this confusing - wouldn't unsettled claims have case reserves set against them, and once they are settled by the new admin system those reserves would be released? It seems to me that the introduction of the new admin system would not only affect claim numbers, and so it would have an effect on the CL.

I guess the answer could be that it has no effect on a paid-CL, but would have an effect on an incurred-CL.

 

I think there are two issues here. Firstly, volatility of data. Secondly appropriateness of run-off pattern. By truncating the data we'll reduce the volatility, and yes it could make quite a difference.

I've not seen the truncating method much in practice though. The problem is that large losses tend to develop differently, ie the second issue above. So it's more common to split out large losses and reserve for them separately. That would help solve both issues.