Question on Excess Credibility with Zero Experience

[scubatz65]

I was wondering if anyone had any helpful formulas/ material on applying credibility to individual insureds on excess layers even if the insured doesn't have actual experience in the layer, but is a large enough client that its experience should hold weight.

Most situations I will be coming across are with one insured and attempting to apply credibility to the insureds specific loss history coupled with industry standards.

Any bit of information would be greatly appreciated as I do not have much experience in working with credibility as I only sat for Exam 4 once.

[Cloister]

Originally Posted by scubatz65

I was wondering if anyone had any helpful formulas/ material on applying credibility to individual insureds on excess layers even if the insured doesn't have actual experience in the layer, but is a large enough client that its experience should hold weight.

Most situations I will be coming across are with one insured and attempting to apply credibility to the insureds specific loss history coupled with industry standards.

Any bit of information would be greatly appreciated as I do not have much experience in working with credibility as I only sat for Exam 4 once.

I've used a company's expected # of excess claims (usually using some form of size of loss distribution to generate that figure) rather than actual number of claims to figure into a credibility calculation.

[MountainHawk]

Originally Posted by Cloister

I've used a company's expected # of excess claims (usually using some form of size of loss distribution to generate that figure) rather than actual number of claims to figure into a credibility calculation.

If you are going to use claim counts as a basis for credibility, you should ALWAYS be using expected claim count, excess or no.

[Hiu]

the insured doesn't have actual experience in the layer？How long is the experience period? 1 years or 10 years? If it is 20 years, the things will be much different i'm afraid, aha

[Lucy]

Originally Posted by MountainHawk

If you are going to use claim counts as a basis for credibility, you should ALWAYS be using expected claim count, excess or no.

Same for claim dollars.

[Colymbosathon ecplecticos]

Originally Posted by Lucy

Same for claim dollars.

Yes, this is easily seen if you look at the derivation of classical credibility. The reasoning for other forms of credibility is similar.

[Glenn Meyers]

A couple of yoars ago I wrote an Actuarial Review article that describes a Bayesian approach to your question. The elevator description is that you start off with a set of parameterized distributions as a prior, use the data you have (it could consist of claims outside the layer in question), and then use Bayesian techniques to calculate the posteror weight of each distribution in your set. You then calcuate the cost of the layer in question using the posterior proabability weighed mixture of the distributions in your set.

Here is the link to the article.

http://www.casact.org/newsletter/ind...iewart&id=5554

Note that I did pretty well in the COTOR Challenges with variations of this approach.